International Journal of Advances in Applied Sciences (IJAAS) Volume 8, issue 3, Sep. 2019 by ijaas_journal

ISSN: 2252-8814

IJAAS

International Journal of

Advances in Applied Sciences

Advances in Applied Sciences (IJAAS) is a peer-reviewed and open access journal dedicated to publish significant research findings in the field of applied and theoretical sciences. The journal is designed to serve researchers, developers, professionals, graduate students and others interested in state-of-the art research activities in applied science areas, which cover topics including: chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences.

Editor-in-Chief: Qing Wang, National Institute of Advanced Industrial Science and Technology (AIST), Japan Co-Editor-in-Chief: Chen-Yuan Chen, National Pingtung University of Education, Taiwan, Province of China Bensafi Abd-El-Hamid, Abou Bekr Belkaid University of Tlemcen, Algeria Guangming Yao, Clarkson University, United States Habibolla Latifizadeh, Shiraz (SUTECH) University, Iran, Islamic Republic of EL Mahdi Ahmed Haroun, University of Bahri, Sudan

Published by:

Institute of Advanced Engineering and Science (IAES) Website: http://iaescore.com/journals/index.php/IJAAS/ Email: info@iaesjournal.com, IJAAS@iaesjournal.com

Information for Authors International Journal of Advances in Applied Sciences (IJAAS) is an interdisciplinary journal that publishes material on all aspects of applied and theoretical sciences. The journal encompasses a variety of topics, including chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences. Submission of a manuscript implies that it contains original work and has not been published or submitted for publication elsewhere. It also implies the transfer of the copyright from the author to the publisher. Authors should include permission to reproduce any previously published material.

Ethics in publishing For information on Ethics in publishing and Ethical guidelines for journal publication (including the necessity to avoid plagiarism and duplicate publication) see http://www.iaescore.com/journals/index.php/IJAAS/about/editorialPolicies#sectionPolicies

Paper Submission You must prepare and submit your papers as word document (DOC or DOCX). For more detailed instructions and IJAAS template please take a look and download at: http://www.iaescore.com/journals/index.php/IJAAS/about/submissions#onlineSubmissions The manuscript will be subjected to a full review procedure and the decision whether to accept it will be taken by the Editor based on the reviews. Manuscript must be submitted through our on-line system: http://www.iaescore.com/journals/index.php/IJAAS/ Once a manuscript has successfully been submitted via the online submission system authors may track the status of their manuscript using the online submission system.

IJAAS

International Journal of

Advances in Applied Sciences

A novel multilevel inverter with reduced switch count C. R. Balamurugan, K. Vijayalakshmi

171-175

Odd generalized exponential log logistic distribution: A new acceptance sampling plans based on percentiles Srinivasa Rao Gadde, K. Rosaiah, D. C. U. Sivakumar, K. Kalyani

176-183

Cache optimization cloud scheduling (COCS) algorithm based on last level caches K. Vinod Kumar, Ranvijay

184-194

On approach for homogeneity increasing of films grown from the gas phase with account natural convection and changes in the rate of chemical interaction between materials Evgeny L. Pankratov

195-203

Improved speed response of DC motor via intelligent techniques Hassan Farahan Rashag

204-207

Theory and development of magnetic flux leakage sensor for flaws detection: A review Nor Afandi Sharif, Rizauddin Ramli, Abdullah Zawawi Mohamed, Mohd Zaki Nuawi

208-216

Active power loss reduction by opposition based kidney search algorithm Kanagasabai Lenin

217-224

A novel impedance source fed H-type flying capacitor multilevel inverter C. R. Balamurugan, P. Vijayakumar, T. Sengolrajan

225-231

Performance analysis of security framework for software defined network architectures K. A. Varun Kumar, D. Arivudainambi

232-242

Compact planar ultrawideband MIMO antenna for wireless applications P. Pavithra, A. Sriram, K. Kalimuthu

243-250

Responsibility of the contents rests upon the authors and not upon the publisher or editors.

IJAAS

Vol. 8

No. 3

pp. 171-250

September 2019

ISSN 2252-8814

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 171~175 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp171-175



171

A novel multilevel inverter with reduced switch count C. R. Balamurugan, K. Vijayalakshmi Department of EEE, Karpagam College of Enigneering, India

Article Info

ABSTRACT

Article history:

This paper presents a multilevel inverter with reduced number of switches to produce a five level output. PWM technique (pulse width modulation) has been used to trigger the MLI switches. It gives reduced harmonic. This proposed topology is connected with R-load and RL-load. Four signals are generated for switching on the multilevel inverter (MLI) switches by comparing four level triangular waveform with sine wave. In this proposed topology two switches are reduced from the conventional Cascaded five level inverter. The simulation analysis has been done by MATLAB/SIMULINK.

Received Apr 23, 2019 Revised May 25, 2019 Accepted Jul 14, 2019 Keywords: CMLI Five level PWM Reduced switch count

Corresponding Author: C. R. Balamurugan, Department of EEE, Karpagam College of Enigneering, Myleripalayam Village, Othakkal Mandapam, Tamil Nadu 641032, India. Email: crbala2017@kce.ac.in

INTRODUCTION A multilevel inverter is capable of generating higher level in the output voltage waveform. It gives desired AC output voltage from several DC sources. MLI is used in high current and high voltage applications. When the level of output voltage increases, it reduces the total harmonic distortion in the output voltage waveform. Multilevel inverters are used in the areas where we need high power such as power conditioning, industrial motor drives and power grid. Surya Suresh [1] demonstrates the new topology has the advantage of its reduced number of devices compared to conventional topology and can be extended to any number of levels. The harmonic reduction is achieved by selecting proper switching angles. Divya subramaniyan [2] proposed inverter provides higher quality output with lower power loss compared to other conventionl inverters. Sivagamasundari [3] based on analysis of cascaded multilevel inverter using hybrid PWM method. The proposed topology reduces the number of power switches when compared to the conventional cascaded inveter. Vinod kumar, et al., [4] designed single phase five level inverter is simulated using multicarrier PWM and conventional method. The THD in output voltage of MLI using multicarrier PWM is less compared to conventional method. Panchal, et al., [5] proposed a three phase fivel level cascaded inverter and modified cascade inverter for asynchronous motor. From simulation analysis modified cascade multilevel inverter is very beneficial over the cascaded MLI. Praveen [6] proposed a new topology with low number of switches to increase the levels of output voltage. Proposed work of single phase cascade inverter output voltage total harmonic distortions is reduced and efficiency is improved. Dubey [7] presents design and analysis of cascade H-bridge multilevel inverter using sinusoidal pulse width modulation technique. From this analysis quality of power improved by using multilevel inverter. Lavanya Raj [8] proposed a five level inverter with multicarrier pulse width modulation technique and embedded matlab s-function. Total harmonic distortion is low for multicarrier PWM. The THD can be further reduced by using

Journal homepage: http://iaescore.com/online/index.php/IJAAS

172



ISSN: 2252-8814

s-function. Balamurugan, et al., [9] deals with different modulation strategies of MLI. Ankit Dubey [10] proposed a various multilevel inverter topologies and control methods. 2.

CASCADED MULTILEVEL INVERTER A single phase cascaded inverter is shown in Figure 1. It is proposed to develop a desired AC output voltage from several DC sources. This is the most common type of inverter and use separate DC source. Each separate DC source is connected to a H-bridge inverter. Each inverter can generate three different output voltage levels +Vdc, 0, -Vdc. +Vdc can be obtained by switching on S1 and S4. To obtain –Vdc switches S2 and S3 are turned on wheares 0 can be obtained by turning on four switchs S1, S2, S3 and S4. The number of voltage levels in a CHB inverter can be obtained from m= (2N+1). Where N is the number of H-bridge. In CHB inverter the voltage levels is always an odd number. It requires less number of components among all multilevel inverter to achieve the same number of voltage levels. Switching states of cascaded multilevel inverter is shown in Table 1. Table 1. Switching states of cascaded five level inverter Voltage (V0) 0 Vdc 2Vdc -Vdc -2Vdc

S1 0 1 1 0 0

S2 1 0 1 1 0

S3 0 0 0 0 1

S4 1 1 0 1 1

S5 1 0 1 1 0

S6 0 1 1 0 0

S7 1 0 0 0 1

S8 0 1 0 1 1

S1 Vdc1 S3

V0 S5

Vdc2 S7

Figure 1. Cascaded multilevel inverter 3.

PROPOSED MULTILEVEL INVERTER Figure 2 shows the proposed five levels multilevel inverter. Due to higher number of switching devices, the inverter circuit size and cost also increases. In this paper the proposed topology provides higher number of output voltage levels with reduce number of switches. In this topology five levels of output +2Vdc, Vdc, 0, -Vdc and -2Vdc are obtained by one module. Here (m-1)/2 number of DC sources are used to produce m levels in the output voltage. In this inverter, PWM technique is employed to achieve high quality output voltage. Four signals are generated by comparing four level shifted triangular waveforms with single sine wave. Switching states of five level inverter with reduced number of switches is shown Table 2. Table 2. Switching states of five level inverter with reduced number of switches Voltage (V0) 0 Vdc 2Vdc -Vdc -2Vdc

S1 0 0 0 1 1

S2 0 1 1 0 0

S3 0 0 1 0 0

S4 0 0 0 1 0

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 171 – 175

S5 0 1 1 0 0

S6 0 0 0 0 1

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

173

Figure 2. Proposed multilevel inverter with reduced number of switches 4.

SIMUATION AND RESULT ANALYSIS The simulation circuit was developed for proposed inverter with R and RL load. The simulated output voltage is shown for only one sample value of ma=1. For simulation analysis, the following parameter values are used: Vdc1=Vdc2=100V, R=100ohms, L=3mH F, fc=2000Hz and fm=50Hz. 4.1. Proposed circuit with R load It consists of five switches with two DC sources. The load can be assumed as R-load. The proposed multilevel inverter with R-load circuit has been shown in Figure 3. Figure 4 and Figure 5 show the output voltage and harmonic spectrum of proposed inverter with R-load. Table 3 shows the measurement across multilevel inverter for R-load.

Vdc1

Vdc2

D4 S3

S4 D3

Figure 3. Proposed multilevel inverter with R-load

A novel multilevel inverter with reduced switch count (C. R. Balamurugan)

174



ISSN: 2252-8814

Figure 4. Output voltage of the proposed inverter with R-load

Figure 5. Harmonic spectrum of output voltage of the proposed inverter with R-load

Table 3. Measurement across multilevel inverter for R-load ma 1 0.9 0.8 0.7 0.6

Peak value (V) 199.8 180 160 140 120.2

RMS value (V) 141.3 127.3 113.2 98.96 85.02

DC component 0.2137 0.2624 0.0098 0.4624 0.3832

THD (%) 26.33 32.87 37.38 40.46 41.65

FFT analysis has been done for generated output voltage by using MATLAB. Total harmonic distortion observed for this R-load to be 26.33% as shown in Figure 5. 4.2. Proposed circuit with RL load The load can be assumed as RL-load. The proposed multilevel inverter with RL-load circuit has been shown in Figure 6. Figure 7 and Figure 8 display the output voltage and harmonic spectrum of proposed inverter with RL-load. For switching on the MLI switches PWM scheme has been used. Four signals are generated by comparing four level triangular waveform with sine wave. FFT analysis has been done in the MATLAB for generated output voltage waveform. Total harmonic distortion observed to be 29.40% as shown in Figure 8. Table 4 shows the measurement across multilevel inverter for RL-load.

D2 R

Vdc1

Vdc2 D4

S4 D3

L S6

Figure 6. Proposed multilevel inverter with RL-load

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 171 – 175

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

Figure 7. Output voltage of the proposed inverter with RL-load

175

Figure 8. Harmonic spectrum of output voltage of the proposed inverter with RL-load

Table 4. Measurement across multilevel inverter for RL-load ma 1 0.9 0.8 0.7 0.6

Peak value (V) 199 179.1 158.8 137.9 117.6

RMS value (V) 140.7 126.6 112.3 97.49 83.15

DC component 0.0556 0.1913 0.5075 0.4161 0.3341

THD (%) 29.40 35.58 41.39 47.54 51.47

CONCLUSION In this paper the simulation results of single phase five level inverter with R and RL load are analysed by MATLAB/SIMULINK. The output voltage and THD spectrum of various load are obtained. From Table 3 and Table 4 the proposed inverter with R-load gives less %THD and its give better performance than RL-load. REFERENCES [1] K. Surya, Suresh K., and Vishnu Prasad M, “Performance and Evaluvation of New Multilevel Inverter Topology,” International Journal of Advances in Engineering and Technology, vol. 3(2), pp. 485-494, 2012. [2] Divya Subramaniyan and Rebiya Rasheed, “Five Level Cascaded H-Bridge Multilevel Inverter Using Multicarrier Pulse Width Modulation Technique,” International Journal of Engineering and Innovative Technology, vol. 3(1), pp. 438-441, 2013. [3] Sivagamasundari M.S. and Malba Mary P., “Analysis of Cascaded Five Level Multilevel Inverter Using Hybrid Pulse Width Modulation,” International Journal of Emerging Technology and Advanced Engineering, vol. 3(4), pp. 55-58, 2013. [4] Vinod Kumar P., “Single Phase Cascaded Multilevel Inverter Using Multicarrier PWM Technique,” ARPN Journal of Engineering and Applied Sciences, vol. 8(10), pp. 796-799, 2013. [5] Tejas M. Panchal, “Simulation of Cascaded and Modified Cascaded H-Bridge Multilevel Inverter for 3-Phase Asynchronous Motor,” International Journal of Recend Trends in Engineering and Technology, vol. 11, pp. 404-414, 2014. [6] Praveen B. R., “Design and Implementation of Cascade H-Bridge Multilevel Inverter,” International Journal of Latest Research in Engineering and Technology, pp. 129-134, 2016. [7] Akanksha Dubey and Ajay Kumar Bansal, “Cascaded H-Bridge Multilevel Inverter,” IJCTA, vol. 9, no. 7, pp. 3029-3036, 2016. [8] Lavanya Raj K. and Karthik N., “Cascaded Five Level Inverter Switching Sequence Optimization Using SFunction,” International Journal of Engineering Development and Research, vol. 4(3), pp. 190-194, 2016. [9] Balamurugn C. R., “A Review on Modulation Strategies of Multilevel Inverter,” International Journal of Electrical Engineering and Computer Science, vol. 3(3), pp. 681-705, 2016. [10] Ankit Dubey and Rakesh Singh Lodhi, “A Review on Cascade H-Bridge Multilevel Inverters,” International Journal of Emerging Technology and Advanced Engineering, vol. 6(12), pp. 231-235, 2016.

A novel multilevel inverter with reduced switch count (C. R. Balamurugan)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 176~183 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp176-183



176

Odd generalized exponential log logistic distribution: A new acceptance sampling plans based on percentiles Gadde Srinivasa Rao1, K. Rosaiah2, D. C. U. Sivakumar3, K. Kalyani4 1

Department of Statistics, The University of Dodoma, Tanzania Department of Statistics, Acharya Nagarjuna University, India

2,3,4

Article Info

ABSTRACT

Article history:

In this paper, acceptance sampling plans are developed for the odd generalized exponential log logistic distribution based on percentiles when the life test is truncated at a pre-specified (pre-determined) time. The minimum sample size necessary to ensure the specified life percentile is obtained under a given consumer’s risk. The operating characteristic values of the sampling plans as well as the producer’s risk are presented. One example with real data set is also given as an illustration.

Received Apr 1, 2019 Revised Jul 14, 2019 Accepted Aug 2, 2019 Keywords: Acceptance sampling Consumer’s risk OC curve Producer’s risk Truncated life tests

Corresponding Author: Gadde Srinivasa Rao, Department of Statistics, The University of Dodoma, P.O. Box: 259, Tanzania. Email: gaddesrao@yahoo.com

INTRODUCTION Acceptance sampling is ‘the middle of the road’ approach between no inspection and 100% inspection. The objective of acceptance sampling is not to estimate the quality of the lot, but to decide whether or not the lot is likely to be acceptable. The applications stanch from real life scenarios: if every bullet was tested in advance prior to war, no bullet is at hand for the time of action and if no bullet is tested, then malfunctions may occur in the war with disastrous results. The selection of a sample from a lot or consignment and the outcome of the products totally depend on the characteristics collected from this sample which was described by [1]. This procedure is called as acceptance sampling plan (ASP) or ‘lot sentencing’. In mass production, a sample is taken at random and tested on the basis of the quality characteristics, ASP is used to accept or reject a submitted lot. An ASP is a specified plan that establishes the minimum sample size to be used for testing. In most ASPs for a truncated life test, the foremost issue is to determine the minimum sample size from a lot under consideration. Traditionally, the lot of items is accepted when the life test indicates that the average life of items exceeds the specified one, otherwise it is rejected. For any industries, the objective is to reducing the cost and test time, a truncated life test may be conducted to obtain the smallest sample size to ensure a certain average life time/percentile lifetime of items, for a given acceptance number c, the number of failures observed does not exceed when the life test is terminated at a pre-assigned time. The decision is to accept the lot if a pre-determined average lifetime/percentile lifetime can be reached with a pre-determined high probability which provides protection to consumer. Therefore, the life test is ended at the time the failure is observed or at the pre-assigned time, whichever is earlier. For such a truncated life test and the associated decision rule; we are focused in obtaining the smallest sample size to arrive at a decision. Journal homepage: http://iaescore.com/online/index.php/IJAAS

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



177

In the past few decades, much effort has gone into the investigation of acceptance sampling plans under a truncated life test. The truncated life tests in the exponential distribution were first considered by [2]. Truncated life tests are deliberated by many authors for various distributions: for example [3-7]. The ASPs could be used for the quantiles and derived the formulae for generalized Birnbaum-Saunders distribution and Marshall-Olkin extended Lomax distribution was proposed by [8, 9]. ASP based on truncated life tests for log-logistic distribution and exponentiated Fréchet distribution was proposed by [10, 11]. The design of ASPs based on the population mean under a truncated life test is considered by all the authors who are in the above. For a skewed distribution, the median represents a better quality parameter than the mean was suggested by [12]. On the other hand, for a symmetric distribution, mean is preferable to use as a quality parameter. ASPs based on the truncated life tests to Birnbaum-Saunders distribution and Burr type XII for percentiles was considered by [13, 14] and they proposed that the ASPs based on mean may not satisfy the requirement of engineering on the specific percentile of strength or breaking stress. When the quality of a specified low percentile is concerned, the ASPs based on the population mean could pass a lot which has the low percentile below the required standard of consumers. Furthermore, a small diminution in the mean with a simultaneous small intensification in the variance can result in a significant downward shift in small percentiles of interest. This means that a lot of products could be accepted due to a small decrease in the mean life after inspection. But the material strengths of products are deteriorated significantly and may not meet the consumer’s expectation. Therefore, engineers should pay more attention to the percentiles lifetimes than the mean life in life testing applications. Moreover, most of the employed life distributions are not symmetric. Actually, percentiles provide more information regarding a life distribution than the mean life does. When the life distribution is symmetric, the 50 th percentile or the median is equivalent to the mean life. Several authors developed the acceptance sampling plans based on percentile. ASPs from truncated life tests based on the log-logistic and inverse Rayleigh distributions, Marshall – Olkin extended Lomax distribution, Linear Failure Rate distribution, Half Normal distribution, Gompertz distribution for percentiles were developed by [15-20]. ASPs based on median life for Fréchet distribution was discussed by [21]. An ASPs from truncated life tests based on the weighted exponential distribution was considered by [22]. New acceptance sampling plans based on percentiles for exponentiated Fréchet distribution was constructed by [23]. An ASPs based on percentiles for Odds exponential log logistic distribution (OELLD) was discussed by [24]. New ASPs based on life tests for Birnbaum–Saunders distributions was considered by [25]. Acceptance sampling for attributes via hypothesis testing and the hyper-geometric distribution was developed by [26]. Acceptance sampling based on life tests from some specific distributions was constructed [27]. These reasons we are motivate to develop ASPs based on the percentiles, since odd generalized exponential log logistic distribution, we prefer to use the percentile point as the quality parameter, and it will be denoted by 𝑡 . The rest of the paper is organized as follows: In Section 2, we describe concisely the odd generalized exponential log logistic distribution. In Section 3, the design of proposed acceptance sampling plan for lifetime percentiles under a truncated life test is presented. In Section 4, we present the description of the proposed plan and obtain the necessary results. An example with real data set and comparison of the proposed sampling scheme with the OELLD is also given as an illustration. Finally, conclusions are made in Section 5. 2.

THE ODD GENERALIZED EXPONENTIAL LOG LOGISTIC DISTRIBUTION In this section, we provide a brief summary about the odd generalized exponential log logistic distribution (OGELLD). The OGELLD was introduced and studied quite extensively by [28]. The probability density function (pdf) and cumulative distribution function (cdf) of OGELLD respectively are given as follows 𝑓(𝑡; 𝜎, 𝜆, 𝜃, 𝛾) =

𝐹(𝑡; 𝜎 , 𝜆, 𝜃, 𝛾) = 1 − 𝑒

1−𝑒

for 𝑡 > 0, 𝜎, 𝜆 > 0, 𝜃, 𝛾 > 1

, t > 0, 𝜎, 𝜆, 𝜃 and 𝛾 > 1

(1)

(2)

where 𝜎, 𝜆 are the scale parameters and 𝜃, 𝛾 are shape parameters respectively. The 100q-th quantile of the OGELLD is given as 𝑡 = 𝜎𝜂 , where 𝜂 = −𝜆 𝑙𝑛 1 − 𝑞

(3)

Odd generalized exponential log logistic distribution: A new acceptance sampling … (Gadde Srinivasa Rao)

178



ISSN: 2252-8814

Hence, for the fixed values of 𝜆 = 𝜆 , 𝜃 = 𝜃 and 𝛾 = 𝛾 , the quantile tq given in (3) is the function of scale parameter 𝜎 = 𝜎 , that is 𝑡 ≥ 𝑡 ⇔ 𝜎 ≥ 𝜎 , where

𝜎 =

= 𝑡

−𝜆 𝑙𝑛 1 − 𝑞

(4)

= 𝑡 𝜂

Note that 𝜎 also depends on 𝜆 , 𝜃 and 𝛾 , to construct the acceptance sampling plans for the OGELLD ascertain 𝑡 ≥ 𝑡 , equivalently that 𝜎 exceeds 𝜎 . 3.

THE ACCEPTANCE SAMPLING PLAN The problem considered is that of finding the minimum sample size necessary to ensure a percentile lifetime of the product, when the life test is terminated at a pre-assigned time 𝑡 and when the observed number of failures does not exceed a given acceptance number c. The decision procedure is to accept a lot only if the specified percentile lifetime can be established with a pre-assigned high probability α, which provides protection to the consumer. The life test experiment gets terminated at the time at which (𝑐 + 1) failure is observed or at quantile time 𝑡 , whichever is earlier. The probability of accepting lot based on the number of failures from a sample under a truncated life test at the test time schedule 𝑡 is given by 𝑃 ( 𝑝) = ∑

𝑛 𝑝 (1 − 𝑝) 𝑖

(5)

where n is the sample size, c is the acceptance number and p is the probability of getting a failure within the life test schedule, 𝑡 . If the product lifetime follows an OGELLD, then𝑝 = 𝐹(𝑡 ; 𝜎, 𝜆 , 𝜃 , 𝛾 ). Usually, it would be convenient to express the experiment termination time 𝑡 as 𝑡 = 𝛿 𝑡 for a constant 𝛿 and the targeted 100q-th lifetime percentile, 𝑡 . Suppose 𝑡 is the true 100q-th lifetime percentile. Then, p can be rewritten as

𝑝 = 1 − 𝑒𝑥𝑝 −

(6)

= 1 − 𝑒𝑥𝑝 −

In order to obtain the proposed design parameters of the proposed plan, we prefer the approach based on two points on the Operating Characteristic (O.C) curve by considering the Type I and Type II errors (i.e., producer’s and consumer’s risk). In our methodology, the quality level is intended through the ratio of its percentile lifetime to the true lifetime, 𝑡 ÷ 𝑡 . These ratios are very useful for the producer to give the better quality of products. Meanwhile the producer’s perspective, the probability of lot acceptance should be at least 1 − 𝛼 at acceptable reliability level (ARL), 𝑝 . Therefore, the producer demands that a lot should be accepted at various levels, say 𝑡 ÷ 𝑡 = 2,4,6,8 in (5). Whereas the consumer’s viewpoint, the lot is rejection should be at most 𝛽 at the lot tolerance reliability level (LTRL), 𝑝 . However, the consumer considers that a lot should be rejected when 𝑡 ÷ 𝑡 = 1. From (5), we have 𝑃 (𝑝 ) = ∑

𝑛 𝑝 (1 − 𝑝 ) 𝑖

≥ 1−𝛼

(7)

𝑃 (𝑝 ) = ∑

𝑛 𝑝 (1 − 𝑝 ) 𝑖

≤𝛽

(8)

where 𝑝 and 𝑝 are given by

𝑝 = 1 − 𝑒𝑥𝑝

and 𝑝 = 1 − 𝑒𝑥𝑝

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 176 – 183

𝜂 𝛿

(9)

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

179

The proposed plan parametric quantities for different values of parameters 𝜆, 𝜃 and 𝛾 are constructed. Given the producer’s risk 𝛼 = 0.05 and termination time schedule 𝑡 = 𝛿 𝑡 with 𝛿 = 1.0, 1.5, 2.0 and 2.5 the four parameters of the proposed plan under the truncated life test at the pre-specified time, 𝑡 with 𝜃 = 2, 𝛾 = 2 and 𝜆 = 0.5, 1.0, 1.5, 2.0 are obtained according to the consumer’s confidence levels 𝛽 = 0.25, 0.10, 0.05, 0.01 for 50th percentile and the O.C. values are also obtained and the results are framed in Table 1 to Table 4. The proposed plan parameters are presented in Table 1 to Table 4 for 𝜃=2, 𝛾 = 2 and 𝜆 = 0.5, 1.0, 1.5, 2.0 with 50th percentiles, whereas Table 5 shows the plan parameters for 𝜆 = 10.7592, 𝜃 = 2.4083 and 𝛾=1.3177 at 50th percentile. On clear observation, we noticed from Table 1 to Table 4 that the percentile ratio increases, the sample size ‘n’ decreases. Table 1. Minimum sample size necessary to assert the 50th percentile life and the corresponding O.C values of OGELLD for 𝜆 = 0.5, 𝜃 = 2.0, 𝛾 = 2.0 𝑡 𝛿 =1.0 𝛿 =1.5 𝑡 c n 𝑃 (𝑝 ) c n 𝑃 (𝑝 ) c 0.25 2 10 25 0.9574 8 14 0.9531 9 4 3 10 0.9658 3 7 0.9672 3 6 2 7 0.9811 1 3 0.9513 2   8 1 5 0.9609 0.9718 1 0.10 2 14 37 0.9578 13 24 0.9600 12 4 4 14 0.9701 4 10 0.9678 3 6 2 9 0.9604 2 6 0.9652 2     8 0.9814 0.9840 1 0.05 2 17 47 0.9512 16 31 0.9505 15 4 5 18 0.9749 4 11 0.9504 4 6 3 13 0.9765 3 9 0.9771 2  8 2 11 0.9670 2 7 0.9741 0.01 2 19 4 6 25 0.9593 6 16 0.9739 5 6 4 19 0.9773 3 11 0.9518 3    8 3 17 0.9763 0.9813 The upward arrow (↑) indicates the same values as the cell above. 𝛽

𝛿 =2 n 𝑃 (𝑝 ) 13 0.9539 5 0.9810 4 0.9806 3 0.9513 18 0.9519 6 0.9549 5 0.9580 3 0.9513 23 0.9534 8 0.9660 5 0.9580  0.9807 30 0.9506 11 0.9563 8 0.9622  0.9858

c 8 2 1  13 3 2  16 4 2  5 3 2

n 10 3 2  17 5 4  21 7 4  9 7 5

𝛿 =2.5 𝑃 (𝑝 ) 0.9526 0.9653 0.9513 0.9723 0.9522 0.9582 0.9641 0.9839 0.9618 0.9588 0.9641 0.9839 0.9619 0.9534 0.9646

Table 2. Minimum sample size necessary to assert the 50th percentile life and the corresponding O.C values of OGELLD for 𝜆 = 1.0, 𝜃 = 2.0, 𝛾 = 2.0 𝑡 𝛿 =1.0 𝛿 =1.5 𝑡 𝑃 (𝑝 ) 𝑃 (𝑝 ) c n c n c 0.25 2 4 12 0.9697 4 7 0.9605 6 4 1 5 0.9835 1 3 0.9757 1  6 0 2 0.9622 0 1 0.9576     8 0.9786 0.9759 0 0.10 2 5 17 0.9587 5 9 0.9637 6 4 1 7 0.9673 1 4 0.9544 2     6 0.9928 0.9898 1 8 0 4 0.9576 0 2 0.9524 0 0.05 2 6 21 0.9616 7 13 0.9714 8 4 1 8 0.9576 1 4 0.9544 2     6 0.9906 0.9898 1    8 0.9969 0 2 0.9524 0.01 2 8 30 0.9610 8 16 0.9590 9 4 2 14 0.9805 2 7 0.9789 2 6 1 11 0.9822 1 5 0.9835 1      8 0.9940 0.9945 The upward arrow (↑) indicates the same values as the cell above. 𝛽

𝛿 =2 𝑃 (𝑝 ) n 8 0.9648 2 0.9747  0.9945 1 0.9576 8 0.9648 4 0.9858 3 0.9843 1 0.9576 11 0.9673 4 0.9858 3 0.9843  0.9948 13 0.9539 5 0.9687 4 0.9702  0.9898

c 7 2 1  11 2 1  11 2 1  14 2 1 

n 8 3 2  13 3 2  13 3 2  17 4 3 

𝛿 =2.5 𝑃 (𝑝 ) 0.9634 0.9867 0.9871 0.9957 0.9645 0.9867 0.9871 0.9957 0.9645 0.9867 0.9871 0.9957 0.9598 0.9561 0.9644 0.9877

Odd generalized exponential log logistic distribution: A new acceptance sampling … (Gadde Srinivasa Rao)

180



ISSN: 2252-8814

Table 3. Minimum sample size necessary to assert the 50th percentile life and the corresponding O.C values of OGELLD for 𝜆 = 1.5, 𝜃 = 2.0, 𝛾 = 2.0 𝑡 𝛿 =1.0 𝛿 =1.5 𝑡 c n 𝑃 (𝑝 ) c n 𝑃 (𝑝 ) c 0.25 2 2 7 0.9720 3 5 0.9760 4 4 0 2 0.9706 0 1 0.9529 1     6 0.9910 0.9852 0      8 0.9962 0.9936 0.10 2 3 12 0.9715 4 7 0.9786 6 4 1 7 0.9956 1 3 0.9935 1 6 0 4 0.9821 0 2 0.9706 0      8 0.9924 0.9873 0.05 2 3 13 0.9621 4 8 0.9560 6 4 1 8 0.9942 1 4 0.9875 1 6 0 5 0.9777 0 2 0.9706 0      8 0.9905 0.9873 0.01 2 4 19 0.9603 6 12 0.9731 8 4 1 11 0.9890 1 5 0.9798 1  6 0 7 0.9690 0 3 0.9563     8 0.9867 0.9810 0 The upward arrow (↑) indicates the same values as the cell above. 𝛽

𝛿 =2 n 𝑃 (𝑝 ) 5 0.9688 2 0.9893 1 0.9662  0.9852 8 0.9648 2 0.9843 1 0.9662  0.9852 8 0.9648 3 0.9702 1 0.9662  0.9852 11 0.9673 3 0.9702  0.9966 2 0.9706

c 8 1  0 8 1  0 8 1  0 12 1  0

n 9 2  1 9 2  1 9 2  1 14 2  1

𝛿 =2.5 𝑃 (𝑝 ) 0.9596 0.9667 0.9960 0.9719 0.9596 0.9667 0.9960 0.9719 0.9596 0.9667 0.9960 0.9719 0.9524 0.9667 0.9960 0.9719

Table 4. Minimum sample size necessary to assert the 50th percentile life and the corresponding O.C values of OGELLD for 𝜆 = 2.0, 𝜃 = 2.0, 𝛾 = 2.0 𝑡 𝛿 =1.0 𝛿 =1.5 𝑡 𝑃 (𝑝 ) 𝑃 (𝑝 ) c n c n c 0.25 2 1 5 0.9576 3 5 0.9847 4 4 0 3 0.9837 0 1 0.9748 1     6 0.9966 0.9945 0      8 0.9989 0.9982 0.10 2 2 9 0.9792 3 6 0.9630 4 4 0 4 0.9783 0 2 0.9503 1     6 0.9955 0.9891 0      8 0.9986 0.9964 0.05 2 2 11 0.9632 3 6 0.9630 6 4 0 5 0.9730 0 2 0.9503 1     6 0.9944 0.9891 0      8 0.9982 0.9964 0.01 2 3 17 0.9728 4 9 0.9520 6 4 0 7 0.9624 1 4 0.9963 1   6 0.9922 0 3 0.9837 0      8 0.9975 0.9947 The upward arrow (↑) indicates the same values as the cell above. 𝛽

𝛿 =2 𝑃 (𝑝 ) n 5 0.9688 2 0.9951 1 0.9837  0.9945 5 0.9688 2 0.9951 1 0.9837  0.9945 8 0.9648 2 0.9951 1 0.9837  0.9945 8 0.9648 3 0.9860 2 0.9677  0.9891

c 9 1 0  9 1 0  9 1 0  9 1 0 

n 10 2 1  10 2 1  10 2 1  10 2 1 

𝛿 =2.5 𝑃 (𝑝 ) 0.9582 0.9789 0.9631 0.9872 0.9582 0.9789 0.9631 0.9872 0.9582 0.9789 0.9631 0.9872 0.9582 0.9789 0.9631 0.9872

Table 5. Minimum sample size necessary to assert the 50th percentile life and the corresponding O.C values of OGELLD for 𝜆 = 10.7592, 𝜃 = 2.4083, 𝛾 = 1.3177 𝑡 𝛿 =1.0 𝛿 =1.5 𝑡 c n 𝑃 (𝑝 ) c n 𝑃 (𝑝 ) c 0.25 2 2 7 0.9831 3 5 0.9817 4 4 0 3 0.9692 0 1 0.9637 1     6 0.9913 0.9896 0      8 0.9965 0.9958 0.10 2 2 9 0.9643 3 6 0.9565 4 4 0 4 0.9591 1 3 0.9961 1   6 0.9884 0 2 0.9794 0      8 0.9953 0.9916 0.05 2 3 13 0.9795 3 6 0.9565 6 4 1 8 0.9971 1 3 0.9961 1 6 0 5 0.9856 0 2 0.9794 0      8 0.9942 0.9916 0.01 2 4 19 0.9806 4 8 0.9674 6 4 1 11 0.9944 1 4 0.9925 1  6 0 7 0.9799 0 3 0.9692     8 0.9918 0.9874 0 The upward arrow (↑) indicates the same values as the cell above. 𝛽

𝛿 =2 n 𝑃 (𝑝 ) 5 0.9688 2 0.9927 1 0.9747  0.9896 5 0.9688 2 0.9927 1 0.9747  0.9896 8 0.9648 2 0.9927 1 0.9747  0.9896 8 0.9648 3 0.9792  0.9981 2 0.9794

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 176 – 183

c 9 1 0  9 1 0  9 1 0  9 1 0 

n 10 2 1  10 2 1  10 2 1  10 2 1 

𝛿 =2.5 𝑃 (𝑝 ) 0.9599 0.9740 0.9501 0.9792 0.9599 0.9740 0.9501 0.9792 0.9599 0.9740 0.9501 0.9792 0.9599 0.9740 0.9501 0.9792

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



181

DESCRIPTION OF METHODOLOGY FOR PROPOSED PLAN WITH REAL DATA EXAMPLE 4.1. Description of the proposed plan Let us assume that the producer desires to implement a proposed plan for assuring that the 50th percentile life of the products under inspection is at least 1000 hours when 𝛽 = 0.10 at the percentile ratio 𝑡 ÷ 𝑡 = 2. He desires to run this experiment 1000 hrs. From the past data, if it is observed that the lifetime of the item follows OGELLD with 𝜆 = 𝜃 = 𝛾 = 2. The optimal plan from Table 4 or specified requirements such as, 𝛽=0.10, 𝜆 = 𝜃 = 𝛾=2, 𝑡 ÷ 𝑡 = 2 and 𝛿 = 1.0 is obtained as 𝑛 = 9 and c=2 with the acceptance probability is 0.9792. Most of the life testing with ASPs for various life time distributions available in the literature is based on one point on the OC curve approach for assuring mean or percentile lifetime. But in this study, we have designed sampling plans based on two-points on the OC curve approach for assuring percentile lifetime of the products under OGELLD. 4.2. Real data example The following real data set corresponds to an uncensored data set from [29-32] on breaking stress of carbon fibres (in Gba). We describe the proposed plan for this data set: 0.39, 0.81, 0.85, 0.98, 1.08, 1.12, 1.17, 1.18, 1.22, 1.25, 1.36, 1.41, 1.47, 1.57, 1.57, 1.59, 1.59, 1.61, 1.61, 1.69, 1.69, 1.71, 1.73, 1.80, 1.84, 1.84, 1.87, 1.89, 1.92, 2.00, 2.03, 2.03, 2.05, 2.12, 2.17, 2.17, 2.17, 2.35, 2.38, 2.41, 2.43, 2.48, 2.48, 2.50, 2.53, 2.55, 2.55, 2.56, 2.59, 2.67, 2.73, 2.74, 2.76, 2.77, 2.79, 2.81, 2.81, 2.82, 2.83, 2.85, 2.87, 2.88, 2.93, 2.95, 2.96, 2.97, 2.97, 3.09, 3.11, 3.11, 3.15, 3.15, 3.19, 3.19, 3.22, 3.22, 3.27, 3.28, 3.31, 3.31, 3.33, 3.39, 3.39, 3.51, 3.56, 3.60, 3.65, 3.68, 3.68, 3.68, 3.70, 3.75, 4.20, 4.38, 4.42, 4.70, 4.90, 4.91, 5.08, 5.56 We show a rough indication of the goodness of fit for our model by plotting the density (together with the data histogram) for the data shows that the OGELLD is a good fit in Figure 1 and also goodness of fit is emphasized with Q-Q plot, displayed in Figure 1. The maximum likelihood estimates of the parameters of OGELLD for the breaking stress of carbon fibres are 𝜆 = 10.7592, 𝜃 = 2.4083 and 𝛾 = 1.3177 and the K-S test and found that the maximum distance between the data and the fitted of the OGELLD is 0.0644 with p-value is 0.8006. Therefore, the four-parameter OGELLD provides good fit for the breaking stress of carbon fibres.

Figure 1. The density plot and Q-Q plot of the fitted OGELLD for the strength data Let us suppose that it is desired to develop the single ASP to satisfy that the 50 th percentile lifetime is greater than breaking stress of carbon fibres 0.35 through the experiment to be completed by breaking stress of carbon fibres 0.35. Let us fix that the consumer's risk is at 25% when the true 50th percentile is breaking stress of carbon fibres 0.35 and the producer's risk is 5% when the true 50 th percentile is breaking stress of carbon fibres 0.70. Since 𝜆 = 10.7592, 𝜃 = 2.4083 and 𝛾 = 1.3177, the consumer's risk is 25%, 𝛿 = 1.0 and 𝑡 /𝑡 = 2, the minimum sample size and acceptance number given by n =9 and c =2 from Table 5. Thus, the design can be implemented as follows. Select a sample of 9 breaking stress of carbon fibres, we will accept the lot when no two or more failure occurs before breaking stress of carbon fibres 0.70. According to this proposed plan, the breaking stress of carbon fibres could have been accepted because there is only one failure before the termination time of breaking stress of carbon fibres 0.70. Odd generalized exponential log logistic distribution: A new acceptance sampling … (Gadde Srinivasa Rao)

182



ISSN: 2252-8814

4.3. Comparison of distributions In Table 6, we compare the plan parameters of the proposed sampling plan with the odds exponential log –logistic distribution (OELLD) which was studied by [24] when 𝛽 = 0.05 and for various levels of 𝛿 = 1.0, 1.5, 2.0, 2.5. The sample size for the OGELLD is smaller as compared to OELLD for 50th percentiles. Table 6. Comparison between OGELLD and OELLD for the combination of 𝜆 = 𝜃 = 𝛾 = 2 𝛽 = 0.05 𝑡 /𝑡 2 4 6 8

𝛿 = 1.0 n 11 5 5 5

OGELLD 𝛿 = 1.5 𝛿 n 6 2 2 2

= 2.0 n 8 2 1 1

𝛿 = 2.5 n 10 2 1 1

𝛿 = 1.0 n 21 8 8 8

𝛿 = 1.5 n 13 4 4 2

OELLD 𝛿 = 2.0 n 11 4 3 3

𝛿 = 2.5 n 13 3 2 2

CONCLUSION In this manuscript, we established the single ASPs based on the OGELLD percentiles when the life test is truncated at a pre-fixed time. To ensure that the life quality of products exceeds a specified one in terms of the percentile life, the ASPs based on percentiles can be used. We have designed sampling plans based on two-points on the OC curve approach for assuring percentile lifetime of the products. Some tables are provided for practical use in industry and also proposed plan illustrated with real data set. We fitted the proposed OGELLD curve for the above data which is shown in the following graphs. The proposed sampling scheme is illustrated with a real data set and results shows that our methodology performs well as compared with existing sampling plans. ACKNOWLEDGEMENTS The authors are deeply thankful to the editor and reviewers for their valuable suggestions to improve the manuscript. REFERENCES [1] H. F. Dodge and H. G. Romig, Sampling inspection tables – single and double sampling, 2nd Edition. John Wiley and Sons, New York, 1959. [2] B. Epstein, “Truncated life tests in the exponential case,” Annals of Mathematical Statistics, vol. 25, pp. 555-564, 1954. [3] M. Sobel and J. A. Tischendrof, “Acceptance sampling with new life test Objectives,” Proceedings of Fifth National Symposium on Reliability and Quality Control, Philadelphia, Pennsylvania, pp.108-118, 1959. [4] S.S. Gupta and P.A. Groll, “Gamma distribution in acceptance sampling based on life tests,” J. Amer. Statist. Assoc., vol. 56, pp.942-970, 1961. [5] R. R. L. Kantam and K. Rosaiah, “Half logistic distribution in acceptance sampling based on life tests,” IAPQR Transactions, vol. 23(2), pp.117-125, 1998. [6] A. Baklizi and A.E.K. EI Masri, “Acceptance sampling based on truncated life tests in the Birnbaum-Saunders model,” Risk Analysis, vol. 24(6), p. 1453, 2004. [7] T.-R. Tsai and S.-J. Wu, “Acceptance sampling based on truncated life tests for generalized Rayleigh distribution,” Journal of Applied Statistics, vol. 33(6), pp. 595-600, 2006. [8] N. Balakrishnan, V. Leiva and J. Lopez, “Acceptance sampling plans from truncated life tests based on the generalized Birnbaum-Saunders distribution,” Communication in Statistics-Simulation and Computation, vol. 36(3), pp. 643-656, 2007. [9] G. S. Rao, M. E. Ghitany and R.R.L. Kantam, “Acceptance sampling plans for Marshall-Olkin extended Lomax distribution,” International Journal of Applied Mathematics, vol. 21(2), pp. 315-325, 2008. [10] R. R. L. Kantam, K. Rosaiah and G. S. Rao, “Acceptance sampling based on life tests: Log-logistic models,” Journal of Applied Statistics, vol. 28(1), pp. 121-128, 2001. [11] A.D. Al-Nassar and A.I. Al-Omari, “Acceptance sampling plan based on truncated life tests for exponentiated Fréchet distribution,” Journal of Statistics and Management Systems, vol. 16(1), pp. 13-24, 2013. [12] Shanti S. Gupta, “Life test sampling plans for normal and log-normal distribution,” Technometrics, vol. 4, pp. 151- 175, 1962. [13] Y. L. Lio, T.-R. Tsai, and S.-J. Wu, “Acceptance sampling plan based on the truncated life test in the Birnbaum Saunders distribution for percentiles,” Communications in Statistics-Simulation and Computation, vol. 39, pp. 119-136, 2009.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 176 – 183

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



183

[14] Y.L. Lio, T.-R. Tsai, and S.-J. Wu, “Acceptance sampling plans from truncated life tests based on the Burr type XII percentiles,” Journal of the Chinese Institute of Industrial Engineers, vol. 27(4), pp. 270-280, 2010. [15] G.S. Rao and R.R. L. Kantam, “Acceptance sampling plans from truncated life tests based on the log-logistic distributions for percentiles,” Economic Quality Control, vol. 25(2), pp. 153–167, 2010. [16] G.S. Rao, R.R. L. Kantam, K. Rosaiah, and J. Pratapa Reddy, “Acceptance sampling plans for percentiles based on the inverse Rayleigh distribution,” Electron. J. App. Stat. Anal., vol. 5(2), pp. 164-177, 2012. [17] G.S. Rao, “Acceptance Sampling Plans for percentiles based on the Marshall-Olkin extended Lomax distribution,” International Journal of Statistics and Economics, vol. 11(2), pp. 83-96, 2013. [18] B. Rao, M. Ch. Priya, and R. R. L. Kantam, “Acceptance Sampling Plans for Percentiles Assuming the Linear Failure Rate Distribution,” Economic Quality Control, vol. 29(1), pp. 1-9, 2013. [19] B. Rao, Ch. Srinivasa Kumar, and K. Rosaiah, “Acceptance Sampling Plans from Life Tests Based on Percentiles of Half Normal Distribution,” Journal of Quality and Reliability Engineering, vol. 2013, 2013. [20] W. Gui and S. Zhang, “Acceptance sampling plans based on truncated life tests for Gompertz distribution,” Journal of Industrial Mathematics, vol. 2014, 2014. [21] S. Balamurali, M. Aslam, and M. S. Fallah Nezhad, “An acceptance sampling plan under Fréchet distribution assuring median life,” Research Journal of Applied Sciences, Engineering and Technology, vol. 6(24), pp. 4519-4523, 2013. [22] W. Gui and M. Aslam, “Acceptance sampling plans based on truncated life tests for weighted exponential distribution,” Communications in Statistics-Simulation and Computation, vol. 46(3), pp. 2138-2151, 2017. [23] G. S. Rao, K. Rosaiah, M. Sridhar Babu, and D. C. U. Sivakumar, “A New Acceptance Sampling Plans based on Percentiles for Exponentiated Fréchet Distribution,” Stochastics and Quality Control, vol. 31(1), pp. 37-44, 2016. [24] G. S. Rao, K. Rosaiah, K. Kalyani, and D. C. U. Sivakumar, “A New Acceptance Sampling Plans based on Percentiles for Odds Exponential Log Logistic Distribution,” The Open Statistics & Probability Journal, vol. 7, pp. 45-52, 2016. [25] M. Aslam, C.H. Jun, and M. Ahmad, “New acceptance sampling plans based on life tests for Birnbaum–Saunders distributions,” Journal of Statistical Computation and Simulation, vol. 81(4), pp. 461-470, 2011. [26] R. W. Samohyl, “Acceptance sampling for attributes via hypothesis testing and the hyper geometric distribution,” Journal of Industrial Engineering International, vol. 14(2), pp. 395-414, 2018. [27] M. Aslam and M. M. Ali, Acceptance Sampling Based on Life Tests from Some Specific Distributions. In: Testing and Inspection Using Acceptance Sampling Plans. Springer, Singapore, pp. 41-90, 2019. [28] K. Rosaiah, G. S. Rao, D. C. U. Sivakumar, and K. Kalyani, “The Odd Generalized Exponential Log Logistic Distribution,” International Journal of Mathematics and statistics Invention, vol. 4(5), pp. 21-29, 2016. [29] M. D. Nichols and W. J. Padgett, “A Bootstrap control chart for Weibull percentiles,” Quality and Reliability Engineering International, vol. 22, pp. 141-151, 2006. [30] A. I. Al-Omari. “Acceptance sampling plan based on truncated life tests for three parameter Kappa distribution,” Economic Quality Control, vol. 29(1), pp. 53-62, 2014. [31] N. Balakrishnan and H. J. Mailk, “Best linear unbiased estimation of location and scale parameters of the loglogistic distribution,” Comm. Statist, Theory & Meth., vol. 16, pp. 3477-3495, 1982. [32] P. R. Tadikamalla and N. L. Johnson, “Systems of frequency curves generated by the transformation of logistic variables,” Biometrika, vol. 69, pp. 461-465, 1982.

Odd generalized exponential log logistic distribution: A new acceptance sampling … (Gadde Srinivasa Rao)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 184~194 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp184-194



184

Cache optimization cloud scheduling (COCS) algorithm based on last level caches K. Vinod Kumar, Ranvijay Department of Computer Science and Engineering, Motilal Nehru National Institute of Technology, India

Article Info

ABSTRACT

Article history:

Recently, the utilization of cloud services like storage, various software, networking resources has extremely enhanced due to widespread demand of these cloud services all over the world. On the other hand, it requires huge amount of storage and resource management to accurately cope up with everincreasing demand. The high demand of these cloud services can lead to high amount of energy consumption in these cloud centers. Therefore, to eliminate these drawbacks and improve energy consumption and storage enhancement in real time for cloud computing devices, we have presented Cache Optimization Cloud Scheduling (COCS) Algorithm Based on Last Level Caches to ensure high cache memory Optimization and to enhance the processing speed of I/O subsystem in a cloud computing environment which rely upon Dynamic Voltage and Frequency Scaling (DVFS). The proposed COCS technique helps to reduce last level cache failures and the latencies of average memory in cloud computing multi-processor devices. This proposed COCS technique provides an efficient mathematical modelling to minimize energy consumption. We have tested our experiment on Cybershake scientific dataset and the experimental results are compared with different conventional techniques in terms of time taken to accomplish task, power consumed in the VMs and average power required to handle tasks.

Received Apr 23, 2019 Revised Jun 18, 2019 Accepted Jul 3, 2019 Keywords: Cloud computing Dynamic voltage and frequency scaling Energy consumption Last level cache

Corresponding Author: K. Vinod Kumar, Department of Computer Science and Engineering, Motilal Nehru National Institute of Technology, Allahabad, 211004-India. Email: vinodkgpt@gmail.com

INTRODUCTION In the field of next upcoming generation of computing platform, the application of big data is most emerging application because there is a huge increment in the creation of data and the storage space. As per the research done in 2012, the ever-lasting increment of data has converted a particular dataset of few terabytes data to the several petabytes [1]. The several features like large capacity, bulky distributed datasets and high velocity is mainly considered in the applications of Big Data which requires the diverse processing schemes such that optimization strategies and an accurate decision making can be enabled [2]. In actual world a huge amount of data is generated in many fields like medical, surfing on internet, telecommunication, pharmaceutical, technology of information and the business. Moreover, in recent time, cloud computing applications has taken immense rise in real world due to its various facilities provided by the cloud providers like ‘pay-as-you-go’ scheme, massive promotions, easy to use, large connectivity with massive number of subscribers. An application of the cloud computing is distributed type application which allows the user to provide the services as per there demand through the internet [3]. The main reason for the drastic growth of cloud computing is because its application saves huge amount of computational time and capacity required for storing the data and also for accessing the several resources. The different resources which are provided by the cloud providers (i.e. amazon, Microsoft etc.) are Journal homepage: http://iaescore.com/online/index.php/IJAAS

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



185

available in the form of VMs (virtual machines) within Infrastructure-as-a-Service (IaaS) model [4]. In a cloud computing environment, the time required to execute any of the operations on the virtual machines is clearly relied upon the number of instructions handled by 𝑉𝑀 which is in some millions and processing power which is in million instructions per second per core. However, the execution time is also dependent on the criticality levels of various functions as it can be varied for every function. Thus, an efficient scheduling is highly appreciated to execute millions of instructions at a time in a cloud environment. Therefore, an efficient task scheduling can be utilized to reduce the processing time in numerous functions of cloud computing environment. Furthermore, various adaptive control methods like dynamic voltage and frequency scaling (𝐷𝑉𝐹𝑆), dynamic speed scaling (𝐷𝑆𝑆) and dynamic frequency scaling (𝐷𝐹𝑆) are introduced by different researchers in recent time to conserve energy and shield environment. Dynamic voltage and frequency scaling (𝐷𝑉𝐹𝑆) is one of the most promising technology which can adaptively optimize power by scaling frequency and voltage. Other energy optimization technologies which are available in the market are like 𝐷𝐹𝑆, 𝐷𝑆𝑆 and 𝐷𝑆𝑃. All these techniques provide energy-conserving scheduling by diminishing the supply frequency and voltage of the cloud computing environment when numerous jobs are processing adaptively [5-8]. Currently, various chip manufacturing companies provide built-in-processors in integration with 𝐷𝑉𝐹𝑆 technologies to speed up performance of their network and reduce power consumption in various cloud computing systems like Intel utilizes Intel SpeedStep processor [9], 𝐴𝑅𝑀 utilizes intelligent energy manager [10] and 𝐴𝑀𝐷 company uses Power Now processors. The processors which works on higher frequencies can provide decent performance. However, at the same time, they consume immense amount of energy in cloud computing environment. Thus, to ensure high performance and lower energy consumption, 𝐷𝑉𝐹𝑆 is the most suitable technique currently. The energy can be conserved in the cloud computing devices in two ways such as by scaling voltage and frequency either at task slack period or while processing external peripherals. Therefore, a huge amount of energy can be conserved using 𝐷𝑉𝐹𝑆 technologies. Thus, numerous researchers are concerned about the lack of energy conservation in cloud computing environment and hence numerous power efficient technologies are introduced by different researchers to protect environment by enormous amount of power dissipation and enhance the performance of the environment. However, these power efficient technologies require enormous amount of interaction cost between inter-processors. Moreover, these technologies provide insufficient results and energy consumption of cloud computing devices cannot be reduced more due to enormous amount of memory utilization. Furthermore, cache memory optimization is an essential factor to reduce further energy consumption in the cloud computing environment. In recent years, the speed of cloud computing processors and density of main memory has taken immense growth as well as the utilization of I/O sub-systems has extremely enhanced. Specifically, the growth of I/O subsystems in the applications like multimedia and networking has further enhanced the demand of storage capacity. Even though the processing speed of I/O memory sub-systems is highly enhanced, it cannot fulfill the demand of computer sub-systems. Therefore, storage sub-systems are the one of the performances limiting factor and even can be a reason for high energy consumption in cloud computing environment. Therefore, to reduce energy consumption and enhance performance of I/O subsystems in cloud computing environment, the optimization of cache is an essential factor. Furthermore, the storage devices in which cache memory is utilized are termed as fast storage subsystems. However, the capacity of these fast storage sub-systems is limited and hence, replace methods should be utilized to enhance the efficiency of storage devices and cache memory. Moreover, some of the well-known cache replace algorithms utilizes 𝐹𝐼𝐹𝑂, 𝐿𝐹𝑈 𝑎𝑛𝑑 𝐿𝑅𝑈 and their relevant technologies. The cache hit ratio is utilized to enhance the performance of cloud computing environment which rely upon the information reference confinedity. The performance of I/O subsystem is clearly depending on the processing speed of storage sub-systems which need to be similar in all the storage devices. However, this often does not happen as the processing speed of storage devices varies system to system. Therefore, there is a need of a technique, which maintains the cache hit ratio and enhance the accessing speed of the network by optimizing cache memory. Moreover, energy consumption and higher memory utilization can be minimized by reducing the traffic on shared cache memories and memory handlers. The additional energy consumption and slower processing in the cloud computing environment is due to missing of last level caches (𝐿𝐿𝐶) on shared storage sub-systems. Therefore, these problems need to be focused soon so that the performance and memory capacity of cloud computing devices get enhanced. Thus, here, we have adopted a Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) Algorithm Based on Last Level Caches to ensure high cache memory Optimization and to enhance the processing speed of I/O subsystem in a cloud computing environment which rely upon Dynamic Voltage and Frequency Scaling (𝐷𝑉𝐹𝑆). The proposed cache Optimization technique helps to minimize the mismanagement of last level caches and to identify the behaviors of cache patterns. This technique decreases congestion on shared cache memories and relocates memory sub-systems dynamically. The memory capacity Cache optimization cloud scheduling (COCS) algorithm based on last level caches (K. Vinod Kumar)

186



ISSN: 2252-8814

of cloud computing environment can be enhanced by sharing caches between cores dynamically. The experimental outcomes verify that the proposed Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) Algorithm can provide higher performance in terms of energy consumption, memory capacity and efficiency in cloud computing environment even in the worst-case scenarios. This paper is presented in following sections which are as follows. In section 2, we describe about the related work to cache memory and their importance and drawbacks in existing techniques and in section 3, we described our proposed Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) methodology. In section 4, experimental results and evaluation shown and section 5 concludes our paper. 2.

RELATED WORK In recent years, the concerns related to enormous memory utilization and performance in cloud computing environment have taken immense growth. Thus, performance enhancement and proper management of storage sub-systems is highly critical need. Moreover, to enhance the storage in cloud computing environment, the optimization of cache memory. Therefore, a widespread literature survey is introduced on proper utilization of storage sub-systems and energy aware scheduling algorithms and their link with 𝐷𝑉𝐹𝑆 in a multi-core heterogeneous cloud computing environment. In [11], an algorithm for the efficient management of shared caches and their effective partitioning is presented to reduce the accessing of main memory in cloud computing environment. This technique helps to minimize the arithmetic and addressing operations. The architecture of last level cache is introduced to reduce loop tiling problems. However, the balancing of energy consumption and performance is a challenging task. In [12], a cache energy Optimization technique is introduced by interchanging from high speed LI cache to L1 low speed cache using the DVFS application. This model provides a cache hierarchy to model a low-power cache algorithm. This method helps to reduce congestion and also decrease extra latency of main memory. However, a proper cache switching modelling is required to utilize this technique in real time. In [13], an efficient cache architecture is presented for DVFS-enabled devices top reduce the cache overhead in the cloud computing environment. This technique is capable of handling faults by changing associativity adaptively in the network. This technique also helps to eliminate redundant information present in the network. In [14], an efficient caching technique is introduced to compare performance of the network with various state-of-art-techniques. The static cache and redis cache technique help to reduce congestion in the cloud computing network and to provide proper resource utilization. In [15], an accurate memory frequency scaling strategy is introduced using Graphics Processing Units (𝐺𝑃𝑈𝑠) based on dynamic voltage and frequency scaling (𝐷𝑉𝐹𝑆) technique to reduce energy consumption and enhance performance of the network. This strategy optimizes L2 cache, shared memory and L1 cache memory. In [16], various cache bypassing methods are introduced based on CPU and GPU cores and compared with each other. This technique helps to enhance the capacity of cache memory and decrease energy consumption in larger caches. However, the problems like high congestion in the network and miss rate can be enhanced while using cache by-passing techniques. In [17], a cache by-passing technique is presented for various mobile SOCs which are clock domain and 𝐷𝑉𝐹𝑆 enabled. This technique helps to enhance performance of the network and energy conservation in 𝐷𝑉𝐹𝑆-enabled CPUs. Here, last level caches are directly accessed to enhance energy conservation and performance of the system. In [18], a novel cache Optimization technique is presented based on Dynamic voltage and frequency scaling to enhance the reliability of the cloud computing network. This technique helps to enhance performance and capacity of the system. However, overhead of the network is very high using this technique. Various researchers have introduced different cache memory Optimization techniques in above literatures. However, very few methods can be utilized in real-time due to various problems like high overhead, high energy consumption, slower performance and unable to reduce cache memory [11, 12, 14, 1719]. Thus, we have adopted a Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) Algorithm Based on Last Level Caches to ensure high cache memory Optimization and to enhance the processing speed of I/O subsystem in a cloud computing environment based on Dynamic voltage and Frequency Scaling (𝐷𝑉𝐹𝑆) technique. 3.

PROPOSED CACHE OPTIMIZATION CLOUD SCHEDULING (COCS) ALGORITHM MODELLING This section provides detailed modelling for the proposed Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) technique. Here, we introduce a Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) Algorithm Based on Last Level Caches to ensure high cache memory Optimization and to enhance the processing speed of I/O subsystem in a cloud computing environment based on Dynamic voltage and Frequency Scaling (𝐷𝑉𝐹𝑆) technique. The architecture of the proposed Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) technique is Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 184 – 194

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



187

presented in Figure 1. Here, we provide efficient modeling to reduce energy consumption and enhance capacity and efficiency of the cloud computing network. The proposed cache Optimization technique migrates cloud virtual machines to reduce the final Last Level Cache (𝐿𝐿𝐶) failures in the cloud computing network. This technique contains confined and generalized scheduling stages. Here, Figure 1 demonstrates the proposed cache Optimization technique which identifies the behavior of various caches in every 𝑉𝑀 from different working nodes and gathers all the VMs to decrease the final cache failures and the latencies of average memory in the cloud computing network. Figure 1 defines the complete architecture of the proposed 𝐶𝑂𝐶𝑆 model. In each working node, 𝐿𝐿𝐶 failures are checked to measure the performance of the system. The monitoring system measures the 𝐿𝐿𝐶 failures per VM and transfer it to the cloud computing scheduler. The generalized scheduling is depending on the status of VMs from each node. The proposed 𝐶𝑂𝐶𝑆 Algorithm utilizes the measured information of cloud computing VMs. Initially, the 𝐶𝑂𝐶𝑆 technique locates VMs on working nodes which takes CPU and memory for every node to enhance the behavior of cache memories.

Figure 1. Architecture diagram of proposed Cache Optimization Cloud Scheduling (COCS) model 3.1. Modeling of cache-aware cloud scheduling This section provides a detailed modeling for proposed cache-aware cloud scheduling technique. Initially, the proposed cache Optimization techniques takes conflicts of shared cache memories. Then, migrates cloud virtual machines to reduce the final Last Level Cache (𝐿𝐿𝐶) failures in the cloud computing network. This technique contains confined and generalized scheduling stages. All the VMs are gathered together to at each working node and planned to domains of mutual caches. The optimization of VMs is presented to enhance the network bandwidth and network capacity. The cloud simulator scheduler is utilized to reallocate VMs at the nodes where 𝐿𝐿𝐶 failures in a computing network of generalized stage. Here, Algorithm 1 is demonstrated in Table 1 which describes the proposed 𝐶𝑂𝐶𝑆 Algorithm based on Last level caches. This cache aware scheduling algorithm distributed into two stages like one is confined stage and other one is generalized stage. All the VMs at every working nodes are classified by LLC failures and then gathered together according to their missed last level caches in the domains of shared caches. The virtual machines with the most LLC failures fit in group 1, next VM with most 𝐿𝐿𝐶 in group 2. Similarly, the virtual machines with the least LLC failures fits in group 1 and next VM with least 𝐿𝐿𝐶 in group 2. This scheme helps to allocate all the VMs in either of the group. The cloud scheduler realizes two types of nodes present in the cloud computing network such as node with highest 𝐿𝐿𝐶 failures and node with least 𝐿𝐿𝐶 failures in the generalized phase. If the difference between 𝐿𝐿𝐶 failures is more than threshold value then both the VMs are swapped using cache optimization technique. The cloud scheduler performs two-stage scheduling progressively by decreasing the final 𝐿𝐿𝐶 failures in a cloud computing network after a regular interval of time. Algorithm 1 Cache Optimization Cloud Scheduling (COCS)Algorithm Step 1: N_L= <n_x1,….,n_x2> // LLC miss of every computing node Step 2: W_L= <w_x1,….,w_x2> // LLC miss of VMs in each working nodes // confined stage Step 3: for every working node j from 1 to y do // identify LLC failures in for every VM in working node j Step 4: 𝑛𝑥 ⇐ 𝑐𝑜𝑙𝑙𝑙𝑒𝑐𝑡𝑠 (𝑗) Cache optimization cloud scheduling (COCS) algorithm based on last level caches (K. Vinod Kumar)

188



ISSN: 2252-8814

Step 5: 𝑊𝕃 ⇐ 𝑠𝑜𝑟𝑡 𝑛𝑥 // disperse VMs with LLC failures Step 6: disperse (𝑊𝕃 ) Step 7: end for // Generalized stage // identify working nodes with largest and least LLC failures Step 8: largeNode ⇐ 𝑠𝑒𝑎𝑟𝑐ℎ 𝑙𝑎𝑟𝑔𝑒𝑁𝑜𝑑𝑒(𝑁𝕃 ) Step 9: leastNode ⇐ 𝑠𝑒𝑎𝑟𝑐ℎ 𝑙𝑒𝑎𝑠𝑡𝑁𝑜𝑑𝑒(𝑁𝕃 ) // search VMs which contains maximum and minimum LLC failures Step 10: largeVM ⟸ 𝑠𝑒𝑎𝑟𝑐ℎ 𝑙𝑎𝑟𝑔𝑒𝑠𝑡𝑉𝑀(𝑙𝑎𝑟𝑔𝑒𝑁𝑜𝑑𝑒) Step 11: leastVM ⇐ 𝑠𝑒𝑎𝑟𝑐ℎ 𝑙𝑒𝑎𝑠𝑡𝑉𝑀 (𝑙𝑒𝑎𝑠𝑡𝑛𝑜𝑑𝑒) Step 12: if 𝑇 < 𝑙𝑎𝑟𝑔𝑒𝑁𝑜𝑑𝑒𝕃𝕃ℂ − 𝑙𝑒𝑎𝑠𝑡𝑁𝑜𝑑𝑒𝕃𝕃ℂ 𝑡ℎ𝑒𝑛 Step 13: interchange (𝑙𝑎𝑟𝑔𝑒𝑉𝑀, 𝑙𝑒𝑎𝑠𝑡𝑉𝑀 ) Step 14: end if 3.2. Modelling for minimizing energy consumption using COCS In this section efficient mathematical modelling is presented to minimize energy consumption in multicore cloud computing systems using proposed cache optimization cloud scheduling technique. This technique helps to minimize energy consumption in cloud devices using cache minimization concept by eliminating dynamic constrained minimization problem. The energy consumption 𝑃(𝑡) occurs in any multicore cloud computing processor is expressed as (1)

𝑃(𝑡) = 𝑒(𝑡)𝑀 ,

Where, 𝑃(𝑡) represents the energy consumption in a cloud computing processor in the 𝑡 managing period. Here, 𝑒(𝑡) represents power consumption of a cloud computing multi-core processor which is linked with both frequencies of the core 𝐵 and size of the current 𝐿2 cache which remains a constant for the frequencies of the core 𝐵 , size of the current L2 cache and task-load of the system and remain unchanged throughout in every managing period whereas 𝑀 is managing period to release multiple illustrations of every task during 𝑡 managing period. Here, 𝑣 (𝑡) represents the core (𝐵 ) utilization in a 𝑡 managing period which depends on the statistics produced by an operating system. Here, the cloud computing processor consists of two cache levels L1 and L2 which are shared with homogenous cores in multi-core shared architecture. In the proposed model each cloud computing processors support DVFS-enabled cores which conserve high amount of energy. The cache memory is divided into various tasks. The level 2 cache memory divider can be denoted as 𝑎 (𝑡) for a core size 𝐵 . The maximum core frequency for a core size 𝐵 can be expressed as 𝕗 ↑ (𝑡). Then, energy consumption 𝑃(𝑡) of the cloud computing multi-core processor can be expressed as ( )|

( )|

min

∑

𝑚𝑖𝑛

𝑃(𝑡)

,𝕗 ( )|

[𝑉 − 𝑣 (𝑡)] ,

(2) (3)

𝑅↓, ≤ 𝕗 (𝑡) ≤ 𝑅↑, 𝑤ℎ𝑒𝑟𝑒, (1 ≤ 𝑘 ≤ 𝑖)

(4)

∑

(5)

𝑎 (𝑡) ≤ 𝐴

Where, 𝑉 is a Utilization points of sets in which 𝑉 = [𝑉 , … … . , 𝑉 ] for a frequency range of 𝑅↑, , , 𝑅↓, for every core 𝐵 and overall size of L2 caches cloud computing processors is denoted by 𝐴 whereas {𝑎 (𝑡)|1 ≤ 𝑘 ≤ 𝑖 } represents the size of cache memory partition and {𝕗 (𝑡)|1 ≤ 𝑘 ≤ 𝑖} denotes the frequency of cores in a 𝑡 managing period to reduce the difference between core utilization 𝑣 (𝑡) and Utilization points of sets (𝑉 ). Here, the (2) represents the least energy consumption in any cloud computing multicore processor due to constant power generation 𝑒(𝑡) for the 𝑡 managing period. Here, the (3) represents shows that the frequency of CPU usually lies in the acceptable range for every core suing the proposed 𝐶𝑂𝐶𝑆 technique. The change in the frequency of any cloud computing device depends on the processors utilized. The (4) represents summation of each divided cache memory which is equivalent to the overall cache memory of the cloud computing processor. For every core of the processor, the difference between core utilization 𝑣 (𝑡) and Utilization point of sets (𝑉 ) is minimized using the proposed cache Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 184 – 194

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

189

optimization technique by changing size of cache partition and frequency of cores. However, these processes decrease the speed of the cloud computing processors. Therefore, to enhance the speed and performance of the system a dynamic model is introduced to maintain a link between managing 𝑣 (𝑡) and manipulated 𝑎 (𝑡) factors and core frequency 𝕗 (𝑡) in the 𝑡 managing period. Initially, for a core 𝐵 , the proposed dynamic model provides a link between 𝑏 (𝑡), task operating time 𝑀 and both manipulated factors 𝕗 (𝑡) in the 𝑡 managing period and 𝑎 (𝑡). Then, 𝑏 (𝑡) can be of operates in various manners like frequency dependent and independent segments and can be represented as 𝑏 (𝑡) = 𝑖 . 𝕗 (𝑡)

(6)

+ 𝑠 (𝑡),

Where, 𝑖 . 𝕗 (𝑡) is a segment which rely upon frequency whereas 𝑠 (𝑡) represents the a segment which is independent of frequency for an operating time 𝑀 due to operating speed of I/O devices does not rely upon the frequencies of core. These I/O devices does not participate at the time of executing task. Then, the reserved cache memory for a task operating time 𝑀 is denoted as 𝑠 (𝑡) which defines a strong link between size of cache memory and the number of cache failures. Then, a link between 𝑠 (𝑡) and 𝑎 (𝑡) and allocated cache size for a multi-core architecture 𝐵 is expressed as 𝑠 (𝑡) =

𝐷 𝑎 (𝑡) + 𝐻 0 ≤ 𝑎 (𝑡) ≤ 𝑋 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑎 (𝑡) ≥ 𝑋

(7)

Where, 𝑋 is the size of operating set within a task operating time 𝑀 whereas 𝐷 and 𝐻 are the specified task factors. Here, the (7) describes that using the proposed COCS technique, whenever size of operating set 𝑋 is larger than 𝑎 (𝑡), and then the size of cache memory enhances and can lead to a minimum operating time. Similarly, whenever size of operating set 𝑋 is smaller than 𝑎 (𝑡), then cache failure rate becomes high and cannot be handled by assigning an extra cache memory. Then, to manage realtime task, the link between overall independent frequency and operating time of every task for multi-core processor 𝐵 and size of overall cache 𝑎 (𝑡) allocated to core 𝐵 can be expressed as 𝑠 (𝑡) =

∑ 𝐷

𝑎 (𝑡) + ∑ 𝐻 0 ≤ 𝑎 (𝑡) ≤ 𝑋 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑎 (𝑡) ≥ 𝑋

(8)

Where, 𝐷 = 𝐷 𝑎 (𝑡). 𝑎 (𝑡) and 𝑋 = ∑ 𝑋 . Here, the (8) represents the summation of (7) for each task on a multi-core processor 𝐵 . Then, the proposed COCS technique helps to minimize interference between shared caches resources of various cores and can be expressed as ℎ (𝑡) = ∑ 𝑖

𝑞

. 𝕗 (𝑡)

+ ∑ 𝐷

𝑞 𝑎 (𝑡) + ∑ 𝐻

𝑞

(9)

Where, ℎ (𝑡) represents the predicted core utilization and 𝑞 represents rate of the task within operating time 𝑀 for a multi-processor core 𝐵 . From (9), it is verified that ℎ (𝑡) is inversely proportional to the frequency of core 𝕗 (𝑡). The predicted change in utilization ∆ℎ (𝑡) for a multi-core architecture 𝐵 can be expressed as ∆ℎ (𝑡) = 𝑙 (𝑡) ∑ 𝑖

𝑞

+ ∆𝑎 (𝑡) ∑ 𝐷

𝑞

(10)

Where, 𝑙 (𝑡) = 𝕗 (𝑡) − 𝕗 (𝑡 − 1) and ∆𝑎 (𝑡) = 𝑎 (𝑡) − 𝑎 (𝑡 − 1). Here, ∆ℎ (𝑡) can be termed as the linear function of 𝑙 (𝑡) and ∆𝑎 (𝑡). Here, the (10) replaces the direct utilization of frequency of core 𝕗 (𝑡) to 𝑙 (𝑡). The (10) shows that ∆ℎ (𝑡) is directly proportional to 𝑖 and 𝐷 . Then, the cost function of cloud computing system can be reduced with the help of controller for a multi-core 𝐵 as 𝑍 (𝑡) = ∑

‖𝑣 (𝑡 + 𝑐 − 1|𝑡) − 𝛽𝕗 (𝑡 + 𝑐 − 1|𝑘)‖ + ‖𝑢 (𝑡|𝑡) − 𝑢 (𝑡 − 1|𝑡)‖

(11)

Where, 𝑅↓, ≤ 𝕗 (𝑡) ≤ 𝑅↑,

(12)

Cache optimization cloud scheduling (COCS) algorithm based on last level caches (K. Vinod Kumar)

190

 𝑎 (𝑡) ≤ 𝑎

ISSN: 2252-8814 (13)

Where, 𝑢 (𝑡) =

∆

( ) ( )

and 𝐸 represents estimated horizon to estimate the behavior of the device

in 𝐸 managing periods. Here, 𝛽𝕗 (𝑡 + 1|𝑡) represents the trajectory with respect to the utilization factor 𝑣 (𝑡 + 𝑐 − 1|𝑡) must change from the present utilization factor 𝑣 (𝑡) to utilization point of sets 𝑉 . The size of cache 𝑎 (𝑡) for a multi-core system 𝐵 is limited by 𝑎 , to satisfy (5). From this above modelling, the optimization of cache memory can be easily achieved and optimization and least square problems are also minimized. The optimized power consumption can be presented using our proposed COCS technique as 𝑒 (𝑡) = 𝑆 𝕗 (𝑡) + 𝑌 𝑎 (𝑡) + 𝐶

(14)

Where, 𝑅↓, ≤ 𝕗 (𝑡) ≤ 𝑅↑,

(15)

𝑎 (𝑡) ≤ 𝑎

(16)

Where, 𝑆 , 𝑌 𝑎𝑛𝑑 𝐶 are the power factors of the cloud computing multi-core processor. The power consumption of cloud computing multi-core processor is represented as the sum of the power consumption of cores and shared caches. The total power consumption is directly depending on the dynamic components of power 𝑆 𝕗 (𝑡) and leakage power 𝐶 . Therefore, the power consumption in cache can be optimized using the proposed 𝐶𝑂𝐶𝑆 technique. In this way, energy consumption of a cloud computing model is minimized and performance of the system can be enhanced. 4.

PERFORMANCE EVALUATION In this modern era, the utilization of cloud computing embedded devices in daily life has tremendously increased due to abundant utilization of various portable gadgets, information systems and digital appliances etc. However, these embedded devices require high amount of power to operate properly, Thus, the performance of this multi-core cloud computing systems and embedded devices must be very high due to satisfy the ample demand of the market. However, in recent time, the performance of various cloud computing centers has degraded due to large amount of energy consumption in these cloud computing centers. Another major problem while using the cloud computing systems is the storage of the system. The amount of data present in these cloud computing systems is extremely high and requires ample amount of storage to handle these large data. Therefore, to maintain a balancing between high amount of energy consumption and storage enhancement of cloud computing systems, we have introduced a Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) Algorithm Based on Last Level Caches to ensure high cache memory Optimization and to enhance the processing speed of I/O subsystem in a cloud computing environment which rely upon Dynamic Voltage and Frequency Scaling (𝐷𝑉𝐹𝑆). The proposed cache Optimization technique helps to minimize the mismanagement of last level caches and to identify the behaviors of cache patterns. Here, we have tested our model on 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 scientific dataset and various sizes of tasks are considered like 30, 50, 100, and 1000 tasks to calculate time taken to accomplish tasks. 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 Workflow is produced with the help of four type of jobs like 30, 50, 100 and 1000. It is used for the Southern California Earthquake center to identify earthquake dangers in a specified place [20]. It requires large amount of storage and CPU resources. The experimental outcomes are presented in terms of time taken to execute task of different size, power consumed in the VMs and average power required to handle tasks against number of tasks considered. Different factors are considered to calculate operating time and energy consumption which is demonstrated in the following section I Table 1. Our proposed 𝐶𝑂𝐶𝑆 model carried out on 64-bit windows 10 OS with 16 GB RAM which contains an 𝐼𝑁𝑇𝐸𝐿 (𝑅) 𝑐𝑜𝑟𝑒 (𝑇𝑀) 𝑖5 − 4460 processor. It consists of 3.20 GHz CPU. This project is simulated using 𝐸𝑐𝑙𝑖𝑝𝑠𝑒𝑊𝑆 Neon.3 editor and code is written in 𝐽𝐴𝑉𝐴. 4.1. Comparative analysis Recently, the modern high-tech cloud computing systems are attempting to match high-end device performance to fulfill the extreme demand of market. However, achieving high-end system performance can head to the extreme energy consumption in these high-tech cloud centers. This is due to this device are mainly battery-oriented and it is very critical to accurately control high amount of energy consumption. If not, then the lifespan of these devices becomes limited due smaller battery life and it may frustrate cloud Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 184 – 194

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

191

subscribers. Therefore, efficient task scheduling and storage capacity enhancement by removing unwanted data is the best way to keep up with the high performance and minimum energy consumption. Therefore, here, we have introduced a Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) Algorithm Based on Last Level Caches to ensure high cache memory Optimization and to enhance the processing speed of I/O subsystem in a cloud computing environment which rely upon Dynamic Voltage and Frequency Scaling (𝐷𝑉𝐹𝑆). The proposed technique helps to maintain balancing between high perform and minimum energy consumption. Therefore, here, we have tested our experiment on 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 scientific dataset and the experimental results are compared with different conventional techniques in terms of time taken to accomplish task, power consumed in the VMs and average power required to handle tasks. These 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 scientific dataset is widely utilized in evaluating the performance of various scheduling techniques and 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 workflow is utilized in our experiment. Here, Table 1 demonstrates the performance comparison of our proposed COCS technique in terms of task completion time, power sum, average power and power consumption occurs in VMs using 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 scientific dataset of different size like 30, 50,100 and 1000. Power consumption using the proposed 𝐶𝑂𝐶𝑆 technique for 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 30 is 128.4845 Watts, 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 50 is 151.5101 Watts𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒100 is 258.672 Watts and 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 1000 is 1696.490 Watts demonstrated in Table 1 which is extremely low in contrast to conventional standard technique. The proposed 𝐶𝑂𝐶𝑆 technique compares Average task completion time using the proposed 𝐶𝑂𝐶𝑆 technique is compared with other conventional techniques utilizing 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 scientific dataset of different size like 30, 50, 100 and 1000 as demonstrated in Table 1. The Average Task Completion Time with the proposed 𝐶𝑂𝐶𝑆 technique for scientific dataset 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 30 is 0.368431 sec, 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 50 is 0.193506 sec, 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 100 is 0.140568 sec and 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 1000 is 0.058476 sec and compared with various state-of-art-techniques. The average task completion time is much smaller than any other state-of-art techniques like EMS-C, SPEA2, MODE, NSPSO, ∈ −𝐹𝑈𝑍𝑍𝑌 PSO, MOHEFT [20] using our proposed 𝐶𝑂𝐶𝑆 technique demonstrated in Table 2. The average Power required is much smaller than any other state-of-art techniques like EMS-C, SPEA2, MODE, NSPSO, ∈ −𝐹𝑈𝑍𝑍𝑌 PSO, MOHEFT [20] using our proposed 𝐶𝑂𝐶𝑆 technique demonstrated in Table 2. The VMs types with description used in our experiment is shown at Table 3. The machine configuration used to compute these results are demonstrated in Table 4. The Power Consumption required is much smaller than any other state-of-art techniques like EMS-C, SPEA2, MODE, NSPSO, ∈ −𝐹𝑈𝑍𝑍𝑌 PSO, MOHEFT [20] using our proposed 𝐶𝑂𝐶𝑆 technique demonstrated in Table 2. The machine configuration used to compute these results are is demonstrated in Table 5. Figure 2 demonstrates the internal architecture of 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 workflow. Table 1. Various parameters comparison for proposed COCS technique vs DVFS using scientific model Cybershake Parameters DVFS

𝐶𝑂𝐶𝑆

Cybershake 30 Cybershake 50 Cybershake 100 Cybershake 1000 Cybershake 30 Cybershake 50 Cybershake 100 Cybershake 1000

VM=20 VM=30 VM=50 VM=30 VM=20 VM=30 VM=50 VM=30

Total Execution Time (s) 6359.41 14448.90 30124.41 74543.57 262.73 283.68 443.21 1328.05

Power Sum (W) 12175922.64 29068552.89 61177338.40 149968122.08 415469.6734 451080.8167 704751.2215 2111745.021

Average Power (W) 19.146320 20.118183 20.308229 20.118184 15.8135 15.9010 15.9010 15.9010

Power Consumption (𝑊ℎ) 3495.42 8518.39 18966.33 236303.28 128.4845 151.5101 258.672 1696.490

Table 2. Runtime ratios of the peer algorithms against the proposed EMS-C on the real-world workflows (i.e., runtime (peer algorithm) =runtime (COCS)) DAGs

Number of nodes

EMS-C

DVFS

SPEA2

𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 30 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 50 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 100 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 1000

30 50 100 1000

23.77 29.32 31.53 22.71

211.98 288.978 301.244 74.54

1.62 1.28 0.92 0.32

Average Execution time (s) ∈ −𝑓𝑢𝑧𝑧𝑦 MODE NSPSO 𝑃𝑆𝑂 1.18 21.25 15.04 1.37 30.17 26.81 1.31 44.75 42.22 0.97 70.72 69.56

MOHFET 10.39 38.16 110.42 --

𝐶𝑂𝐶𝑆 0.368431 0.193506 0.140568 0.058476

Table 3. VMs types with description used in our experiment Type m1. small m1. large m1. xlarge

Memory (GB) 1.7 7.5 15

Core Speed (ECU) 1 4 8

Cores 1 2 4

Cache optimization cloud scheduling (COCS) algorithm based on last level caches (K. Vinod Kumar)

192



ISSN: 2252-8814

Figure 2. Internal architecture of Cybershake workflow Table 4. Runtime ratios of the peer algorithms against the proposed EMS-C on the real-world workflows (i.e., average power (peer algorithm) = average power (COCS)) DAGs 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 30 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 50 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 100 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 1000

Number of nodes 30 50 100 1000

EMS-C 34.0079 48.186 35.666 6.175

DVFS 141.02 156.72 151.92 52.191

SPEA2 2.317 2.103 1.040 0.087

Average power MODE NSPSO 1.688 30.402 2.251 49.583 1.481 50.621 0.263 19.230

∈ −𝑓𝑢𝑧𝑧𝑦 𝑃𝑆𝑂 21.51782 44.06124 47.75910 18.91500

𝐶𝑂𝐶𝑆 0.5271166 0.31802 0.15901 0.015901

Table 5. Runtime ratios of the peer algorithms against the proposed EMS-C on the real-world workflows (i.e., power consumption (peer algorithm) = power consumption 𝐶𝑂𝐶𝑆) DAGs 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 30 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 50 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 100 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 1000

Number of nodes 30 50 100 1000

EMS-C 276.31 459.13 580.21 658.85

DVFS 1145.8 1493.2 2471.4 5568.3

SPEA2 18.83 20.04 16.92 9.283

Power consumption MODE NSPSO 13.71 247.02 21.45 472.44 24.10 823.48 28.14 2051.7

∈ −𝑓𝑢𝑧𝑧𝑦 𝑃𝑆𝑂 174.8320 419.8304 776.9287 2018.056

𝐶𝑂𝐶𝑆 4.2828166 3.030202 2.58672 1.69649

4.2. Graphical representation This section describes about the graphical demonstration of experimental outcomes using proposed 𝐶𝑂𝐶𝑆 technique. Here, Figure 3 demonstrates task completion time Comparison of proposed 𝐶𝑂𝐶𝑆 technique with conventional 𝐷𝑉𝐹𝑆 technique with the help of scientific dataset 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 for different task sizes as 30, 50,100 and 1000. Here, Figure 4 demonstrates Power Sum Comparison of proposed 𝐶𝑂𝐶𝑆 technique with conventional 𝐷𝑉𝐹𝑆 technique with the help of scientific dataset 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 for different task sizes as 30, 50,100 and 1000. Here, Figure 5 demonstrates Average Power Comparison of proposed 𝐶𝑂𝐶𝑆 technique with conventional 𝐷𝑉𝐹𝑆 technique with the help of scientific dataset 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 for different task sizes as 30, 50, 100 and 1000. Similarly, Figure 6 demonstrates Power Consumption Comparison of proposed 𝐶𝑂𝐶𝑆 technique with conventional 𝐷𝑉𝐹𝑆 technique with the help of scientific dataset 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 for different task sizes as 30, 50, 100 and 1000. Furthermore, Figure 7 demonstrates average task completion time Comparison of proposed 𝐶𝑂𝐶𝑆 technique with various state-of-art-techniques using scientific dataset 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 for different task sizes as 30, 50, 100.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 184 – 194

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



193

Figure 3. Task completion time comparison of proposed COCS technique vs DVFS using scientific workload 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒

Figure 4. Power sum comparison of proposed COCS technique vs DVFS using scientific workload 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒

Figure 5. Average power comparison of proposed COCS technique vs DVFS using scientific workload 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒

Figure 6. Power consumption comparison of proposed COCS technique vs DVFS using scientific workload 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒

Figure 7. Average task completion time comparison of proposed COCS technique with various state-of-art-techniques using scientific workload 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 5.

CONCLUSION An efficient task scheduling in each cloud computing VMs and reduction of high energy consumption in these devices is become a primary priority. Various researchers have provided different techniques to decrease energy consumption in multi-core cloud computing processors. However, still it remains an unsolved issue. Therefore, one method to decrease storage and energy consumption in multi-core processors is cache minimization. Therefore, we have adopted Cache Optimization Cloud Scheduling (𝐶𝑂𝐶𝑆) Algorithm Based on Last Level Caches to ensure high cache memory Optimization and to enhance the processing speed of I/O subsystem in a cloud computing environment which rely upon Dynamic Voltage Cache optimization cloud scheduling (COCS) algorithm based on last level caches (K. Vinod Kumar)

194



ISSN: 2252-8814

and Frequency Scaling (𝐷𝑉𝐹𝑆). This technique helps to reduce last level caches and identifies the behavior of different caches in every 𝑉𝑀 from different working nodes and gathers all the VMs to decrease the final cache failures. A detailed modelling is presented to minimize the energy consumption and storage in cloud computing multi-core processors by reducing cache memory. We have tested our experiment on 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 scientific dataset and the experimental results are compared with different conventional techniques in terms of time taken to accomplish task, power consumed in the VMs and average power required to handle tasks. The Average Task Completion Time with the proposed 𝐶𝑂𝐶𝑆 technique for scientific dataset 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 30 is 8.7576 sec, 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 50 is 5.6736 sec, 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 100 is 4.4321 sec and 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 1000 is 1.328 sec. Power consumption using the proposed COCS technique for 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 30 is 128.4845 Watts, 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 50 is 151.5101 Watts 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒100 is 258.672 Watts and 𝐶𝑦𝑏𝑒𝑟𝑠ℎ𝑎𝑘𝑒 1000 is 1696.490 Watts which is very less compare to any other state-of-art-techniques. Experimental results verify superiority of our proposed 𝐶𝑂𝐶𝑆 technique in terms of task completion time, average power required and energy consumption. REFERENCES [1] Wikipedia, Big data, 2014a. [Online]. Available: http://en.wikipedia.org/wiki/Big_data. [2] M. A. Beyer and D. Laney, The Importance of ‘big data’: A Definition. Gartner, Stamford, CT, 2012. [3] P. Mell and T. Grance, "The NIST Definition of Cloud Computing," National Institute of Standards and Technology, 2009.

[4] H. Topcuoglu, S. Hariri, and M.y. Wu, "Performance-effective and low-complexity task scheduling for heterogeneous computing," IEEE Trans. Parallel Distrib. Syst., vol. 13(3), pp. 260-274, 2002.

[5] K. Li, "Energy-efficient task scheduling on multiple heterogeneous computers: Algorithms, analysis, and performance evaluation," IEEE Trans. Sustain. Comput., vol. 1(1), pp. 7–19, 2017.

[6] K. Li, "Power and performance management for parallel computations in clouds and data centers," J. Comput. Syst. Sci., vol. 82(2), pp. 174–190, 2016.

[7] X. Xiao, G. Xie, R. Li, and K. Li, "Minimizing schedule length of energy consumption constrained parallel applications on heterogeneous distributed systems," Proc. 14th IEEE Int. Symp. on Parallel and Distributed Processing with Applications, IEEE Computer Society, pp. 1471-1476, 2016. [8] G. Xie, X. Xiao, R. Li, and K. Li, "Schedule length minimization of parallel applications with energy consumption constraints using heuristics on heterogeneous distributed systems," Concurrency Comput. Parctice Experience, vol. 29(16), 2016. [9] “Enhanced intel speedstep technology for the intel pentium m processor,” March 2014. [Online]. Available: http://downloa10] “Amd powernow!? technology informational white paper,” November 2000. [Online]. Available: http://www.amd-k6.com/wpcontent/ uploads/2012/07/24404a.pdf [10] K. Flautner, D. Flynn, and M. Rives, "A combined hardware software approach for low-power socs: Applying adaptive voltage scaling and intelligent energy management software," 2003. [Online]. Available: intel.com/design/network/papers/30117401.pdf [11] V. Kelefouras, G. Keramidas, and N. Voros, "Cache Partitioning + Loop Tiling: A Methodology for Effective Shared Cache Management," 2017 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), Bochum, pp. 477-482, 2017. [12] K. Saito, R. Kobayashi, and H. Shimada, "Reduction of cache energy by switching between L1 high speed and low speed cache under application of DVFS," 2016 International Conference on Advanced Informatics: Concepts, Theory and Application (ICAICTA), pp. 1-6, 2016. [13] M. Mavropoulos, G. Keramidas, and D. Nikolos, "A defect-aware reconfigurable cache architecture for low-Vccmin DVFS-enabled systems," 2015 Design, Automation & Test in Europe Conference & Exhibition (DATE), pp. 417-422, 2015. [14] M. Kusuma, Widyawan, and R. Ferdiana, "Performance comparison of caching strategy on wordpress multisite," 2017 3rd International Conference on Science and Technology - Computer (ICST), pp. 176-181, 2017. [15] Q. Wang and X. Chu, “GPGPU performance estimation with core and memory frequency scaling,” arXiv preprint arXiv:1701.05308, 2017. [16] S. Mittal, “A Survey of Cache Bypassing Techniques,” MDPI J. Low Power Electron. Appl., vol. 6(5), 2016. [17] Joonho Kong and Kwangho Leeb, "A DVFS-aware cache bypassing technique for multiple clock domain mobile SoCs," IEICE Electronics Express, vol. 14(11), pp. 1-12, 2017. [18] Y. H. Chen, Y. L. Tang, Y. Y. Liu, A. C. H. Wu, and T. Hwang, "A Novel Cache-Utilization-Based Dynamic Voltage-Frequency Scaling Mechanism for Reliability Enhancements," IEEE Transactions on Very Large-Scale Integration (VLSI) Systems, vol. 25(3), pp. 820-832, 2017. [19] Z. Zhu, G. Zhang, M. Li, and X. Liu, "Evolutionary Multi-Objective Workflow Scheduling in Cloud," IEEE Transactions on Parallel and Distributed Systems, vol. 27(5), pp. 1344-1357, 2016. [20] E. Deelman, D. Gannon, M. Shields, and I. Taylor, "Workflows and e-science: An overview of workflow system features and capabilities," Future Generat. Comput. Syst., vol. 25(5), pp. 528-540, 2009.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 184 – 194

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 195~203 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp195-203



195

On approach for homogeneity increasing of films grown from the gas phase with account natural convection and changes in the rate of chemical interaction between materials E. L. Pankratov1,2 1

Mathematical and Natural Science Department, Nizhny Novgorod State University, Russia 2Higher Mathematics Department, Nizhny Novgorod State Technical University, Russia

Article Info

ABSTRACT

Article history:

We analyzed growth of films by gas phase epitaxy. Based on the analysis we formulate conditions to increase of homogeneity of properties of obtained films. We also present an analytical approach for analysis processes framework gas phase epitaxy with account natural convection and the possibility of changing the rate of chemical interaction between reagents.

Received Apr 6, 2019 Revised Aug 24, 2019 Accepted Aug 31, 2019 Keywords: Accounting of chemical reaction Accounting of natural convection Analytical approach for prognosis Gas phase epitaxy Prognosis of heat transport Prognosis of mass transport

Corresponding Author: E. L. Pankratov, Mathematical and Natural Science Department, Nizhny Novgorod State University, 23 Gagarin avenue, Nizhny Novgorod, 603950, Russia. Email: elp2004@mail.ru

INTRODUCTION At present large number of electronics devices based on heterostructures. Heterostructures could be grown by using different well-known approaches [1-12]. At the same time at large number of experimental studies on growth of heterostructures [1-12] we find small quantity of theoretical works with prognosis of processes [12]. Main aim of this paper is to analysis changing of properties of epitaxial layers with a changing of the values of different parameters of the epitaxy process with account chemical interaction natural convection. We consider a vertical reactor for gas phase epitaxy (see Figure 1). The reactor includes into itself (i) external casing; (ii) keeper of substrate with a substrate; (iii) spiral around the shell in the region. The spiral generats an induction heating to activate chemical reactions. The heating activate chemical reactions between materials during epitaxy. A mixture of reagents in gas phase enter to inlet of the reaction chamber in atmosphere of a gas-carrier. Our aim framework this paper is analysis of changing growth process with changing of physical parameters and parameters of technological process. An accompany aim of our paper is accounting of natural convection and chemical reaction. The third aim of our paper is development of analytical approach for prognosis of growth process to increase adequateness of the prognosis. It should be noted, that analytical approach for modeling are usually more demonstrative in comparison with numerical one and leads to decrease calculation time. Journal homepage: http://iaescore.com/online/index.php/IJAAS

196



ISSN: 2252-8814

+1

1

-1

-1

Figure 1 (a). Structure of the considered reactor for epitaxy from gas phase with sloping keeper of substrate

Figure 1 (b). View of keeper of substrate from side with approximation of the keeper by sloping lines with angle of sloping 1

METHOD OF SOLUTION We solve our aims by determination and analysis of spatio-temporal distribution of temperature and concentrations of reagents in the considered reactor. The spatio-temporal variation of temperature was obtained by induction heating due to electric current in spiral around keeper of substrate with a substrate to activate of chemical reactions gas-reagents before substrate during formation of required epitaxial layer on the substrate. We determined the considered temperature by solution of the following boundary value problem [13]. ( , , , )

𝑐

= 𝑑𝑖𝑣 𝜆 ⋅ 𝑔𝑟𝑎𝑑[𝑇(𝑟, 𝜙, 𝑧, 𝑡)] − 𝑣⃗(𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄⃗ (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝑐(𝑇) ⋅ 𝑇(𝑟, 𝜙, 𝑧, 𝑡) ⋅

𝐶 (𝑟, 𝜙, 𝑧, 𝑡) + 𝑝(𝑟, 𝜙, 𝑧, 𝑡),

(1)

Here is the flow velocity of the gases-reagents; c is the heat capacity of the system; T (r,,z,t) is the distribution of temperature in space and time; p (r,,z,t) is the heat power density in the considered reactor; r, , z and t are the current cylindrical coordinates and time; C(r,,z,t) is the distribution of the concentration of the mixture of reagents in space and time (we assume, that two reagents with concentrations of inlet in reactor with concentrations C1(r,,z,t) and C2 (r,,z,t)); C3(r,,z,t) is the material of the epitaxial layer in the gas phase;  is the thermal conductivity coefficient, value of this coefficient has been determined by following relation:   v l cv  3 , here v  2k T m describes modulus of the mean-square gas molecules velocity; l describes mean free path of gas molecules between collisions, cv is the heat capacity of the gas at constant volume,  is the gas density. To solve above equations we shall to take into account moving and quantity of reagents. In this situation we shall to solve equation of Navier-Stokes with the second equation of Fick. We assume that the radius of the substrate holder R larger, than diffusion and boundary layers thickness. We also assume that the flow of gas is laminar. In this case the above equations could be written in the form ⃗

+ (𝑣⃗ ⋅ 𝛻)𝑣⃗ = −𝛻

+ 𝜈𝛥𝑣⃗,

(2)

( , , , )

= 𝑑𝑖𝑣 𝐷 ⋅ 𝑔𝑟𝑎𝑑[𝐶 (𝑟, 𝜙, 𝑧, 𝑡)] − 𝑣⃗(𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄⃗ ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑘 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡), (3a) ( , , , )

= 𝑑𝑖𝑣 𝐷 ⋅ 𝑔𝑟𝑎𝑑[𝐶 (𝑟, 𝜙, 𝑧, 𝑡)] − 𝑣⃗(𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄⃗ ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑘 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡), (3b) ( , , , )

= 𝑑𝑖𝑣 𝐷 ⋅ 𝑔𝑟𝑎𝑑[𝐶 (𝑟, 𝜙, 𝑧, 𝑡)] − 𝑣⃗(𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄⃗ ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) + 𝑘 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡), (3c) Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 195 – 203

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

197

Here Di are the coefficients of diffusion of reagents of the considered reagents. As an example of reagents, we consider trimethylgallium (CH3)3Ga and arsenic hydride AsH3. As result of reaction we consider GaAs. As gas-carrier we consider hydrogen. P describes pressure of gas in the reactor;  describes kinematic viscosity of reagents. We assume, that all molecules will deposit on substance. Based on this assumption we used the following boundary and initial conditions. We also assume, that reactor is cylindrical and d2<<d1. In this situation we can write C1 (r,,-L,t) = C2 (r,,-L,t) = C0, C3 (r,,-L,t) = 0; Ci (r,-1,z,t) = Ci (r,1,z,t) = Ci (r,-1,z,t) = Ci (r, +1,z,t); C1 (r,,z,0) = C2 (r,,z,0) = C0 (z+L), C3 (r,,z,0) =0; Ci (0,,z,t)  ; 𝐶 (𝑟, 𝜙, 𝑧, 𝑡)| = 0; T (r,-1,z,t) = T (r,1,z,t) = T (r,-1,z,t) = T (r, +1,z,t); T (r,,z,0) = Tr; ( , , , ) ( , , , ) ( , , , ) ( , , , ) −𝜆 = 𝜎𝑇 (𝑅, 𝜙, 𝑧, 𝑡); = ; = 0; ( , , , )

( , , , )

= 0; −𝜆

= 𝜎𝑇 (𝑟, 𝜙, −𝐿, 𝑡);

( , , , )

= 0; vr (r,-1,z,t) = vr (r,1,z,t)

= vr (r,-1,z,t) = vr (r,+1,z,t); T (0,,z,t)  ; v (r,-1,z,t) = v (r,1, z,t) = v (r,-1,z,t) = v (r,+1,z,t); vz (r,-1,z,t) = vz (r,1,z,t) = vz (r,-1,z,t) = vz (r,+1,z,t); vr (r,,-L,t) = 0; vr(r,,L,t) = 0; vr (0,,z,t)  ; vz(r,,-L,t) = V0; vz(rd2/2,,z[-d2/2,d2/2],0) =  zcos ()tg (1); v (r,,L,t) = 0; v(0,,z,t)  ; vz(r,,0,t) = 0; vz(r,,L,t) =V0, vz(r,,L,t) = V0, vz(0,,z,t)  ; vr(r,,z,0) = 0; v(r,,z,0) =0; v(r,,0,t) =  r, (4) Here  =5,6710-8 Wm-2K-4, Tr describes room temperature,  describes frequency of rotation of substrate. Using cylindrical system of coordinate leads to the following form of equations for projections of reagents velocity =𝜈

𝑟

=𝜈

𝑟

=𝜈

𝑟

( , , , )

−𝑣

( , , , )

−

−𝑣

−

−𝑣

−

−𝑣

(5a)

−

−𝑣 −𝑣

(5b)

− −

(5c)

Now we will calculate solution of (5) by method of averaging functional corrections [14-19]. Equations for first-order approximations vr1, v1, vz1 of the considered components takes the form ,

=−

(6)

Now we integrate both sides of the above equations. The result of integration could be written as 𝑣

=−

∫

𝑑𝜏, 𝑣

=−

∫

𝑑𝜏, 𝑣

=−

∫

𝑑𝜏.

(7)

Equations of the second-order approximations of the velocity projections vr1, v1, vz1 could be written in the form by using standard procedure of method of averaging functional corrections =𝜈 (𝛼

(𝛼

𝑟

+𝑣 )

−

− (𝛼

+𝑣 )

−

− (8a)

−

− (𝛼

+𝑣 )

−

+𝑣 ) =𝜈

+ ,

+𝑣 ) =𝜈

(𝛼

𝑟

(8b) +

−

− (𝛼

+𝑣 )

−

− (8c)

On approach for homogeneity increasing of films grown from the gas phase with … (E. L. Pankratov)

198



ISSN: 2252-8814

Integration of the above equations on time t leads to the following result 𝑣

= 𝜈∫

𝑟

= 𝜈∫

𝑟

∫ 𝑣

𝑑𝜏 − ∫ (𝛼

∫ 𝑣

𝑑𝜏 − ∫ (𝛼

+𝑣 )

𝑟

+𝑣 )

𝑑𝜏 − (8d)

𝑑𝜏 −

𝑑𝜏 − ∫ (𝛼

∫

+𝑣 )

𝑑𝜏 −

𝑑𝜏,

𝑑𝜏 − ∫ (𝛼

∫

𝑑𝜏,

+𝑣 )

= 𝑉 +𝜈∫

∫

𝑑𝜏 −

(8e)

𝑑𝜏 −

∫

𝑑𝜏 − ∫ (𝛼

+𝑣 )

𝑑𝜏 −

𝑑𝜏.

+𝑣 )

(8f)

Average values 2r, 2, 2z have been calculated by the following standard relations 𝛼

𝑣

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡, 𝛼

− 𝑣 )𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 ,

∫ (𝑣

∫ ∫ 𝑟∫ =

∫ ∫ 𝑟∫

∫ (𝑣

𝛼

∫ ∫ 𝑟∫

− 𝑣 )𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 ,

∫

𝑣

−

(9)

Here  is the continuance of technological process. Using of the first- and the second-order approximations of the velocity projections in the above relation leads to necessity of solution of the algebraic system of equations, which is presented bellow 𝐴 𝛼 𝐴 𝛼 𝐴 𝛼

+𝐵 𝛼 +𝐵 𝛼 +𝐵 𝛼

+𝐶 𝛼 +𝐶 𝛼 +𝐶 𝛼

=𝐷 =𝐷 =𝐷

(10)

where, 𝐴 = 1 + ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫

∫

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 , 𝐵 = ∫ (𝛩 − 𝑡) ∫ ∫

𝐶 = 𝛩 𝑅 𝑉 ,𝐷 = 𝜈 × ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫ ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫

𝐷 = 𝜈 ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫ 𝑡) ∫ 𝑟 ∫

∫ 𝑣

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 , 𝐵 = 1 + ∫ (𝛩 − 𝑡) ∫ ∫

∫ ∫

𝑟

𝐵 = ∫ (𝛩 − 𝑡) ∫ 𝑟 × ∫

∫

𝐷 = 𝜈 ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫

∫

∫ 𝑣

∫

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 − 𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 ,

∫ 𝑣 ∫

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 ,

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 − ∫ (𝛩 −

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 − 𝛩 𝑅 𝑉 − ∫ (𝛩 − 𝑡) ∫ ∫

𝐶 = 1 + 𝛩 𝑅 𝑉 , 𝐴 = ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫

∫ (𝛩 − 𝑡) ∫ 𝑟 ∫

𝑟

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 − 𝛩 𝑅 𝑉 − ∫ (𝛩 − 𝑡) ∫ ∫

∫ 𝑣

𝐴 = ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫

∫

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 , 𝐶 =

∫

∫ 𝑣

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 ,

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 , 𝑟

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 − ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 − 𝛩 𝑅 𝑉 − ∫ 𝑣

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 .

Solution of the above equations could be obtained by standard approaches [20] and written in the form 𝛼

= 𝛥 ⁄𝛥 , 𝛼

= 𝛥 ⁄𝛥, 𝛼

= 𝛥 ⁄ 𝛥,

where 𝛥 = 𝐴 (𝐵 𝐶 − 𝐵 𝐶 ) − 𝐵 (𝐴 𝐶 − 𝐴 𝐶 ) + 𝐶 (𝐴 𝐵 − 𝐴 𝐵 ), 𝛥 = 𝐷 (𝐵 𝐶 − 𝐵 𝐶 ) − 𝐵 (𝐷 𝐶 − 𝐷 𝐶 ) + 𝐶 (𝐷 𝐵 −𝐷 𝐵 ), 𝛥 = 𝐷 (𝐵 𝐶 − 𝐵 𝐶 ) − 𝐵 (𝐷 𝐶 − 𝐷 𝐶 ) + 𝐶 (𝐷 𝐵 − 𝐷 𝐵 ), 𝛥 = 𝐴 (𝐵 𝐷 − 𝐵 𝐷 ) − 𝐵 (𝐴 𝐷 − 𝐴 𝐷 ) + 𝐷 (𝐴 𝐵 − 𝐴 𝐵 ). Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 195 – 203

(11)

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

199

In this section, we calculate the required projections of the considered velocity by the method of method of averaging functional corrections. Usually, the such approximation is enough to obtain main results of analysis [15-19]. Now we will analyze distributions of temperature and concentrations of components of gas-reagents in space and time. In this section we will consider (1) and (3) by using cylindrical system coordinate 𝑐

( , , , )

=𝜆

( , , , )

+𝜆

( , , , )

+𝜆

𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)] ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝑇(𝑟, 𝜙, 𝑧, 𝑡)} −

( , , , )

𝑟𝐷

𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)]} − +

𝐷

𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)]} +

𝐷

𝑘 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) −

𝑟𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡) −

−

( , , , )

𝐷

(13a)

{𝑟𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) −

−

𝑟𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡) −

−

( , , , )

(12)

{𝑟𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) −

−

{𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)]},

( , , , )

𝑟𝐷

( , , , )

𝐷

( , , , )

𝑘 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) − ( , , , )

𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅

{𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)]},

( , , , )

𝑟𝐷

𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)]} +

( , , , )

𝑘 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) − ( , , , )

{[𝑣 (𝑟, 𝜙, 𝑧, 𝑡) −

{[𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)] ⋅ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝑇(𝑟, 𝜙, 𝑧, 𝑡)},

𝑇(𝑟, 𝜙, 𝑧, 𝑡) − 𝑐 ⋅ ( , , , )

+ 𝑝(𝑟, 𝜙, 𝑧, 𝑡) − 𝑐 ⋅

( , , , )

𝐷

(13b)

{𝑟𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) −

−

𝑟𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ 𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡) −

−

{𝐶 (𝑟, 𝜙, 𝑧, 𝑡) ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)]}.

(13c)

To calculate distributions of temperature and concentration of reagents in space and time, we use the same method of averaging functional corrections as for solution of (2). We used the same algorithm of method of averaging functional corrections to solve (12) and (13). The result of calculation of the first- order approximations T1(r,,z,t) and Ci1(r,,z,t) of considered distributions could be written as 𝑇 (𝑟, 𝜙, 𝑧, 𝑡) = 𝑇 + ∫ ∫

( , , , )

̄ ( , , , )

𝐶 (𝑟, 𝜙, 𝑧, 𝑡) = − ∫

( , , , )

∫

𝑑𝜏 − 𝛼 𝛼

∫

( , , , )

𝑑𝜏 − 𝛼

{ [

̄ ( , , , )

𝑑𝜏 − 𝛼 𝛼

𝑑𝜏 − 𝛼

{ [

̄ ( , , , )

𝐶 (𝑟, 𝜙, 𝑧, 𝑡) = − ( , , , )

{ [

̄ ( , , , )

𝐶 (𝑟, 𝜙, 𝑧, 𝑡) = − ( , , , )

( , , , )

𝑑𝜏 − 𝛼

[

( , , , )

̄ ( , , , )]

[

( , , , )

̄ ( , , , )]

̄ ( , , , )]}

∫

[

̄ ( , , , )]}

∫

[

∫

̄ ( , , , )]

𝑑𝜏 + 𝛼

( , , , )

𝑑𝜏,

(14)

∫ 𝑘 (𝑟, 𝜙, 𝑧, 𝜏)𝑑𝜏 + 𝐶 −

̄ ( , , , )]

𝑑𝜏 + 𝐶 − 𝛼

( , , , )

̄ ( , , , )]} [

𝑑𝜏 − 𝛼

( , , , )

𝑑𝜏 −

𝑑𝜏,

(15a)

∫ 𝑘 (𝑟, 𝜙, 𝑧, 𝜏)𝑑𝜏 − 𝑑𝜏 ,

(15b)

∫ 𝑘 (𝑟, 𝜙, 𝑧, 𝜏)𝑑𝜏 −

̄ ( , , , )]

𝑑𝜏 .

(15c)

where 1T and i1C are the average values of the considered approximations. Now let us calculate these average values by using the following relations 𝛼

∫ ∫ 𝑟∫

∫ 𝑇 (𝑟, 𝜙, 𝑧, 𝜏)𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡,

𝛼

∫ ∫ 𝑟∫

∫ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡

(16)

On approach for homogeneity increasing of films grown from the gas phase with … (E. L. Pankratov)



200

ISSN: 2252-8814

We calculate average values 1T and i1C by using the following standard relations 𝛼

∫ (𝛩 − 𝑡) ∫

=𝐶 𝐿⋅ 1+

𝛼

∫ (𝛩 − 𝑡) ∫ 𝑟 ∫

= 𝑇 +

( , , , )

∫

𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡

∫ [𝑣 (𝑅, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑅, 𝜙, 𝑧, 𝜏)]𝑑𝑧𝑑𝜙𝑑𝑡 +

𝑡) ∫

𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)]𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡

∫ [𝑣 (𝑅, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑅, 𝜙, 𝑧, 𝑡)]𝑑𝑧𝑑𝜙𝑑𝑡 +

× (𝛩 − 𝑡)𝑑𝑡

∫ (𝛩 −

−

∫ (𝛩 − 𝑡) ∫ ∫

∫ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) −

We calculate approximations of temperature and concentrations of gases with the second-order by method of averaging of function corrections [14-19]. The result of calculation could be written as ( , , , )

𝑐 ⋅ 𝑇 (𝑟, 𝜙, 𝑧, 𝑡) = 𝜆 ∫ [𝛼

+ 𝑇 (𝑟, 𝜙, 𝑧, 𝜏)]}𝑑𝜏 + 𝜆

𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏) ⋅ [𝛼 ( , , , )

+𝜆 ∫

∫ {[𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ [𝛼

𝑑𝜏 − 𝑐 ⋅ ( , , , )

∫

∫ [𝛼

𝑑𝜏 −

𝑑𝜏 − 𝑐 ⋅

( , , , )

𝐶 (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏 − ∫ [𝛼

−

(16)

∫ {[𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ [𝛼

𝑑𝜏 +

( , , , )

∫ 𝐷

𝑟 ∫ [𝛼

−

∫ 𝑟𝐷

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ [𝛼 ( , , , )

𝑑𝜏 + 𝐶 𝛿(𝑧 + 𝐿) − −𝛼

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ 𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏) 𝑑𝜏 − ( , , , )

𝑑𝜏 + 𝐶 𝛿(𝑧 + 𝐿) − 𝛼

( , , , )

∫ 𝑟𝐷

∫ 𝑘 (𝑟, 𝜙, 𝑧, 𝜏)[𝛼

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏 −

∫ [𝛼

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ 𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏) 𝑑𝜏 − ( , , , )

∫ 𝐷

𝑑𝜏 + 𝛼

∫ 𝑘 (𝑟, 𝜙, 𝑧, 𝜏)[𝛼

( , , , )

∫ 𝑟𝐷

∫ [𝛼

(17a)

( , , , )

∫ 𝐷

𝑑𝜏 +

(17b) ∫ [𝛼

∫ [𝛼

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅

𝑑𝜏 +

( , , , )

∫ 𝐷 𝑟 ∫ [𝛼

𝑑𝜏 +

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ 𝑣 (𝑟, 𝜙, 𝑧, 𝜏) −

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏 .

∫ [𝛼

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅

𝑑𝜏 +

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏 −

𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏 − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏) 𝑑𝜏 −

∫ 𝑘 (𝑟, 𝜙, 𝑧, 𝜏)[𝛼

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏 −

𝑟 ∫ [𝛼

[𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏,𝐶 (𝑟, 𝜙, 𝑧, 𝑡) =

𝑑𝜏 +

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏 −

[𝑣 (𝑟, 𝜙, 𝑧, 𝜏) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜏,𝐶 (𝑟, 𝜙, 𝑧, 𝑡) = ∫ 𝐷

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅ 𝑣 (𝑟, 𝜙, 𝑧, 𝜏) −

+ 𝑇 (𝑟, 𝜙, 𝑧, 𝜏)] 𝑑𝜏 + 𝑇 +

𝑇 (𝑟, 𝜙, 𝑧, 𝜏)]}𝑑𝜏 + ∫ 𝑝(𝑟, 𝜙, 𝑧, 𝜏)𝑑𝜏 ,𝐶 (𝑟, 𝜙, 𝑧, 𝑡) = ∫ 𝐷

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)] ⋅

(17c)

Average values of the above approximations were determined by standard relations 𝛼

∫ ∫ 𝑟∫

∫ (𝑇 − 𝑇 )𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 ,

𝛼

∫ ∫ 𝑟∫

∫ (𝐶 − 𝐶 )𝑑𝑧𝑑𝜙𝑑𝑟𝑑𝑡 .

(18)

We calculate average values 2T and i2C by using the following standard relations 𝛼

∫ (𝛩 − 𝑡) ∫

∫ 𝑇 (𝑅, 𝜙, 𝑧, 𝑡)𝑑𝑧𝑑𝜙𝑑𝑡 −

𝑡) ∫

∫ 𝑇 (𝑅, 𝜙, 𝑧, 𝑡)𝑑𝑧𝑑𝜙𝑑𝑡 +

𝑡) ∫

∫ {𝑇 (𝑅, 𝜙, 𝑧, 𝑡)[𝛼

× ∫ (𝛩 − 𝑡) ∫

∫ (𝛩 − ∫ 𝑇 (0, 𝜙, 𝑧, 𝑡)𝑑𝑧𝑑𝜙𝑑𝑡 −

∫ (𝛩 −

+ 𝐶 (𝑅, 𝜙, 𝑧, 𝑡)] − 𝛼 𝛼 } × [𝑣 (𝑅, 𝜙, 𝑧, 𝑡) −

𝑣̄ (𝑅, 𝜙, 𝑧, 𝑡)]𝑑𝑧𝑑𝜙𝑑𝑡 − ∫ (𝛩 − 𝑡) ∫ ∫

∫ {[𝛼

𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)] × 𝑇 (𝑟, 𝜙, 𝑧, 𝑡)𝑑𝑧𝑑𝜙𝑟𝑑𝑟𝑑𝑡

−

𝐶 ) − 𝛼 𝛼 ]𝑑𝜙𝑑𝑟𝑑𝑡 ∫ (𝛩 − 𝑡) × ∫

+ 𝐶 (𝑟, 𝜙, 𝑧, 𝑡)] − 𝛼 𝛼 } ⋅ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − ∫ (𝛩 − 𝑡) ∫ 𝑟 ∫ [𝑇 (𝑟, 𝜙, 𝐿, 𝑡)(𝛼

∫ [𝑣 (𝑅, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑅, 𝜙, 𝑧, 𝑡)] ∑ [𝛼

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 195 – 203

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

𝐶 (𝑅, 𝜙, 𝑧, 𝑡)] 𝑑𝑧𝑑𝜙𝑑𝑡

+1−∫ ∫ 𝑟∫

𝐶 (𝑅, 𝜙, 𝑧, 𝑡)] 𝑑𝑧𝑑𝜙𝑑𝑟(𝛩 − 𝑡)𝑑𝑡

∫ [𝑣 (𝑟, 𝜙, 𝑧, 𝑡) − 𝑣̄ (𝑟, 𝜙, 𝑧, 𝑡)] ⋅ ∑ [𝛼 ,𝛼

+𝐶 )

+ 2𝛩(∑ 𝛼

201

∫ (𝛩 −

𝑡) ∫ 𝑟 ∫ [𝐶 (𝑟, 𝜙, 𝑧, 𝜏) − 𝐶 (𝑟, 𝜙, 𝑧, 𝜏)]𝑑𝜙𝑑𝑟𝑑𝑡 . 3.

RESULTS AND DISCUSSION Now we analyzed redistribution of reagents and temperature in the considered reactor during growth of films to formulate conditions for improvement of properties of films. Figure 2 illustrates dependence of concentrations of reagents on substrate’s rotation frequency. Curve 1 shows such dependence at atmospheric pressure without accounting natural convection. Curve 2 shows such dependence of concentrations of reagents on substrate’s rotation frequency at decreased in 10 times pressure. Curve 3 describes the dependence of the concentration of reagents on substrate’s rotation frequency at atmospheric pressure with account natural convection. The figure shows, that increasing of rotation frequency leads to increasing of homogeneity of epitaxial layer. Figure 3 illustrates dependence of concentration of reagents on diffusion coefficient D in gas-carrier. Curve 1 shows dependence of the concentration of reagents at atmospheric pressure without accounting natural convection. Curve 2 shows dependence of the concentration of reagents at decreased in 10 times pressure without accounting natural convection. Curve 3 shows dependence of the concentration of reagents at atmospheric pressure with accounting natural convection. This figure describes monotonous decreasing of considered concentration. Figure 4 illustrates dependence of concentration of kinematic viscosity of gas-carrier without accounting natural convection. Curves, analogous corves 2 and 3 on previous figures, have small differences with the presented curve. Figure 5 illustrates dependence of concentration of reagents on velocity of mixture of reagents and gas-carrier at the entrance to the reaction zone V0. Curve 1 shows dependence of the concentration of reagents at atmospheric pressure without accounting natural convection. Curve 2 shows dependence of the concentration of reagents at decreased in 10 times pressure without accounting natural convection. Curve 3 shows dependence of the concentration of reagents at atmospheric pressure with accounting natural convection. This figure describes monotonous decreasing of considered concentration.

1.0

3 1

0.5

C /C 0

0.5 3 1

0.0

0.0 0

, rad/s

Figure 2. The dependence of the concentration of the mixture of gases on the frequency of rotation of the substrate

D, cm /s

Figure 3. The dependence of the concentration of the mixture of gases from its diffusion coefficient

On approach for homogeneity increasing of films grown from the gas phase with … (E. L. Pankratov)

202



ISSN: 2252-8814

Figure 4. The dependence of the concentration of the mixture of gases from the kinematic viscosity 1.00

C/C0

0.75 1

0.50

0.25

0.00 0

V0 , cm/s

Figure 5. The dependence of the concentration of the mixture of gases from its input velocity It is attracted an interest choosing of power of induction heating of reaction zone. This choosing should compensate losses of heat due to convective heat transfer [21]. In this situation relaxation time of temperature could be estimate by using recently introduced criterion [22]. This time is equal to  (6 -1) R2/240, where 0 is the average value of the thermal conductivity coefficient. In this case, the power required to compensate for the cooling of the region of the formation of the epitaxial layer can be estimated from the following relation ∫ 𝑟 ⋅ 𝑝(𝑟, 𝜙, 𝑧, 𝑡)𝑑𝑟 ≈ 𝜎 ⋅ 𝑇 (𝑅, 𝜙, 𝑧, 𝑡) + 𝛩 ⋅ 𝑣 (𝑅, 𝜙, 𝑧, 𝑡)⁄4𝜋𝐿𝑅 [16]. 4.

CONCLUSION In this paper based on considered analytical approach for analysis of mass and heat transfer during film growth in reactors for gas phase epitaxy we analyzed this growth with account natural convection and chemical interaction between reagents. In this situation we analytically analyzed growth of films from gas phase in more common case in comparison with cited references to increase predictability of this technological process (it should be noted, that analytical approach for modeling are usually more demonstrative in comparison with numerical one and leads to decrease calculation time). As a result of this analysis, the obtain conditions to increase homogeneity of the grown epitaxial layers with changing of values of parameters of growth process with higher exactness.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 195 – 203

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



203

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22]

I. P. Stepanenko, Basis of microelectronics. Moscow: Soviet radio, 1980. V. G. Gusev and Yu M. Gusev, Electronics. Moscow: Higher School, 1991. V. I. Lachin and N. S. Savelov, Electronics. Rostov-on-Don: Phenics, 2001. A. A. Vorob’ev, et al., “The use of magnesium to dope gallium nitride obtained by molecular-beam epitaxy from activated nitrogen,” Semiconductors, vol. 37(7), pp. 838-842, 2003. L. M. Sorokin, et al., “Electron-microscopic investigation of a SiC/Si (111) structure obtained by solid phase epitaxy,” Technical Physics Letters, vol. 34(11), pp. 992-994, 2008. V. V. Lundin, et al., “Effect of carrier gas and doping profile on the surface morphology of MOVPE grown heavily doped GaN: Mg layers,” Semiconductors, vol. 43(7), pp. 963-967, 2009. Y. E. Bravo-García, et al., “Growth and characterization of InAsSb layers on GaSb substrates by liquid phase epitaxy,” Mat. Sci. Sem. Proc., vol. 40, pp. 253-256, 2015. Y. Li, L. E. Antonuk, et al., “Effects of x-ray irradiation on polycrystalline silicon, thin-film transistors,” J. Appl. Phys., vol. 99(6), pp. 064501, 2006. A. Chakraborty, et al., “Nonpolar -plane p-type GaN and p-n-junction diodes,” J. Appl. Phys., vol. 96(8), pp. 4494, 2004. H. Taguchia, et al., “Evaluation of crystallinity of GaN epitaxial layer after wafer dicing,” Mat. Sci. Sem. Proc., vol. 41, pp. 89-91, 2016. M. Mitsuhara, et al., “Beryllium doping of InP during metalorganic molecular beam epitaxy using bismethylcyclopentadienyl-beryllium,” J. Cryst. Growth, vol. 183(1-2), pp. 38-42, 1998. R. A. Talalaev, et al., “On low temperature kinetic effects in metal–organic vapor phase epitaxy of III–V compounds,” J. Cryst. Growth, vol. 230(1-2), pp. 232-238, 2001. H. S. Carslaw and J. C. Jaeger, Conduction of heat in solids. Oxford: At the Clarendon Press, 1964. Yu. D. Sokolov, “About the definition of dynamic forces in the mine lifting,” Applied Mechanics, vol. 1(1), pp. 23-35, 1955. E. L. Pankratov and E. A. Bulaeva, “On optimization of technological process to decrease dimensions of transistors with several sources,” Micro and Nanosystems, vol. 8(1), pp. 52-64, 2016. E. L. Pankratov and E. A. Bulaeva, “An analytical approach for analysis and optimization of formation of fieldeffect heterotransistors,” Multidiscipline Modeling in Materials and Structures, vol. 12(4), pp. 578-604, 2016. E. L. Pankratov, “On optimization of manufacturing of power amplifier circuit based on bipolar heterostructures to increase density of their elements. Influence of miss-match induced stress,” Advanced science, engineering and medicine, vol. 9(10), pp. 849-863, 2017. E. L. Pankratov, “On decreasing of dimensions of transistors with two sources by optimization of technological process,” Journal of computational and theoretical nanoscience, vol. 14(10), pp. 4947-4954, 2017. E. L. Pankratov and E. A. Bulaeva, “Doping of materials during manufacture p-n-junctions and bipolar transistors. Analytical approaches to model technological approaches and ways of optimization of distributions of dopants,” Reviews in Theoretical Science, vol. 1(1), pp. 58-82, 2013. G. Korn and T. Korn, Mathematical Handbook for scientists and engineers. Definitions, theorems and formulas for reference and review. New York, Second edition. McGraw-Hill Book Company, 1968. E. L. Pankratov and E. A. Bulaeva, “On prognosis of epitaxy from gas phase process for improvement of properties of films,” 3D research, vol. 6(4), paper 40, 2015. E. L. Pankratov and E. A. Bulaeva, “Optimal criteria to estimate temporal characteristics of diffusion process in a media with inhomogenous and nonstationary parameters. Analysis of influence of variation of diffusion coefficient on values of time characteristics,” Reviews in Theoretical Science, vol. 1(3), pp. 305-316, 2013.

On approach for homogeneity increasing of films grown from the gas phase with … (E. L. Pankratov)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 204~207 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp204-207



204

Improved speed response of DC motor via intelligent techniques Hassan Farhan Rashag Technical Institute of Babylon, Al-Furat Al-Awsat Technical University, Iraq

Article Info

ABSTRACT

Article history:

The classical Proportional – Integral (PI) control for Direct Current (DC) motor causes slow response of actual speed with high overshoot and undershoot which leads to sluggishness of the system. To minimize the problem of PI controller, intelligent technique based on hybrid neural network sliding mode control NN-SMC is suggested. The benefits of SMC are that it is simple, and tough to parameter deviations as compared with other controllers. In this paper, the neural network NN is used to minimize the error between reference speed and actual speed. In addition, the SMC aim is to control and optimize the voltage that is supplies the DC motor which guarantees the robust performance of the speed controller under disturbances. The proposed method for the speed control is first calculated and executed to DC motor by using MATLAB SIMULINK. The results of the suggested NNSMC are compared with the traditional PI controller. The results obviously show the supremacy of NN-SMC over PI controller.

Received Apr 29, 2019 Revised Jun 20, 2019 Accepted Jul 5, 2019 Keywords: DC motor Neural network (NN) Sliding mode control (SMC) PI controller

Corresponding Author: Hassan Farhan Rashag, Technical Institute of Babylon, Al-Furat Al-Awsat Technical University, Kufa, Irak. Email: Hassan_rashag@yahoo.com

INTRODUCTION DC motor is commonly used in manufacturing uses because of its dependability. Many researchers suggested methods to control the speed of DC motors. PI controller is used for adjusting the speed of DC motor and implemented in Field-Programmable-Gate-Array (FPGA) [1]. Fuzzy controller was proposed to control the speed of DC motors based on feedback [2]. The real systems were controlled by SMC because of its robustness and simplicity [3]. Thus, these SMC with fuzzy logic controller was suggested to control the DC servo motors dynamic of sensors [4, 5]. New approaches of two SMC were employed for angular velocity of DC motor [6]. Speed of servo motor was enhanced based on discrete SMC [7]. Furthermore, high contribution for control of real systems was recorded based on fractional order SMC [8]. In the nonlinear systems, for multi inputs and multi outputs, the SMC performance was modified via the fractional order SMC [9]. In addition, the fractional-order sliding surface is used to decrease the distortion of speed [10]. The effectiveness of fractional order SMC for speed of a PMSM [11] and position of a DC motor is also confirmed [12]. The main purpose of PI controllers is to control the speed. However, due to parameter deviations of nonlinear system, PI controller is unable to adjust the speed. SMC is designed for control the speed of permanent magnet DC motor based on fractional order to minimize the load disturbance and parameters variations [13, 14]. In the current paper, new approach for optimize the speed of a DC motor is suggested based on hybrid intelligent NN-SMC. In addition, the aim of NN is to decrease the error between the desired speed and actual speed through back propagations BP. Also, the SMC is used to adjustable the voltage which supplies the DC motor. from simulation results, it can be noted that the intelligent technique based on NN-SMC is robustness for nonlinear system.

Journal homepage: http://iaescore.com/online/index.php/IJAAS

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

205

MATERIAL AND PROPOSED METHOD This proposed method is built by using Matlab Simulink with toolbox. The classical control of DC motor based on PI controller is suffered from overshoot, undershoot, and slow response of speed which degrades the system performance. In order to overcome this problem, hybrid technique based on neural network with sliding mode control is suggested. Here, the reference speed and actual speed are the inputs to NN and the error is the output. This error with change of error will apply to SMC to generate the voltage that supply the DC motor. The simulink of NN with SMC is shown in Figure 1.

Figure 1. Simulink of NN with SMC 3.

THE SIMULATION RESULTS AND DISCUSSION From Figure 2 and Figure 3, the speed of DC motor through NN-SMC is good tracking with reference speed and fast response. In contrast, the speed based on classical PI controller has overshoot and under shoot with slow response.

Figure 2. Comparison of speed

Figure 3. Response of actual speed with reference speed

Improved speed response of DC motor via intelligent techniques (Hassan Farhan Rashag)

206



ISSN: 2252-8814

Figure 4 shows the robustness of voltage based NN-SMC with variation of reference speed but the voltage of classical PI controller is unstable. From Figure 5 when the load torque is applied on DC motor, the performance of PI speed response is deteriorating and the effect is obvious while the effect of torque on proposed NN-SMC speed is trivial.

Figure 4. Voltage comparison

Figure 5. Speed response with variation of load torque

The error between the actual speed and reference speed of proposed NN-SMC is almost approach to zero as shown in Figure 6. Furthermore, the error of classical PI controller is fluctuating between 40rad/sec to 30rad/sec which caused instability of system.

Figure 6. Comparison the error of speed Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 204 – 207

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



207

CONCLUSION This proposed method exhibits an enhanced control for speed of DC motor using intelligent technique NN-SMC. The PI controller can be used for regulating the speed but it needs long time to reach the steady state due to overshoot and undershoot. Therefore, the PI controller is replaced by the intelligent NNSMC to modify the system. The NN-SMC gave better performance under load disturbances as compared with PI controller. Finally, the system with NN-SMC is more accuracy, fast response of speed, and low effect under parameters variation condition. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

C. F. Hsu and B. K. Lee, “FPGA-based adaptive PID control of a DC motor driver via sliding-mode approach,” Expert Syst. Appl., vol. 38, pp. 11866-11872, Sept. 2011. G. G. Rigatos, “Adaptive fuzzy control of DC motors using state and output feedback,” Electr. Pow. Syst. Res., vol. 79, pp. 1579-1592, Nov. 2009. A. Pisano and E. Usai, “Sliding mode control: A survey with applications in math,” Math. Comput. Simulat., vol. 81, pp. 954-979, Jan. 2011. J. X. Xu, T. H. Lee, and Y. J. Pan, “On the sliding mode control for DC servo mechanisms in the presence of unmodeled dynamics,” Mechatronics, vol. 13, pp. 755-770, Sept. 2003. J M. J. Knight and R. Sutton, “Fuzzy model based sliding-mode control of a d.c. servomechanism,” P. I. Mech. Eng. I-J. Sys., vol. 218, pp. 211- 225, May 2004. A. Susperregui, G. Tapia, and A. Tapia, “Application of two alternative sliding-mode control approaches to DC servomotor position tracking,” IET Electr. Power Appl., vol. 1, pp. 611-621, July 2007. C. Milosavljevic, B. Perunicic-Derazenovic, and B. Veselic, “Discretetime velocity servo system design using sliding mode control approach with disturbance compensation,” IET Trans. Industr. Inform., vol. 9, pp. 920-927, May 2013. I. Podlubny, Fractional Differential Equations. San Diego, CA, USA: Academic Press, 1999. M. O. Efe and C. Kasnakoglu, “A fractional adaptation law for sliding mode control,” Int. J. Adapt. Control Signal Process., vol. 22, pp. 968- 986, Dec. 2008. M. L. Corradini, R. Giambo, and S. Pettinari, “On the adoption of a fractional-order sliding surface for the robust control of integer-order LTI plants,” Automatica, vol. 51, pp. 364-371, Jan. 2015. B. Zhang, Y. Pi, and Y. Luo, “Fractional order sliding-mode control based on parameters auto-tuning for velocity control of permanent magnet synchronous motor,” ISA Trans., vol. 51, pp. 649-656, Sept. 2012. W. Khan, “A comparison between integer and fractional order algebraic approaches for on-line parameter estimation and position control of DC motor,” 11 th International Bhurban Conference on Applied Sciences and Technology, Islamabad, pp. 103-107, 2014. Vadim Utkin, Jürgen Guldner, and Jingxin Shi, Sliding Mode Control in Electro-Mechanical Systems. CRC Press Boca Raton, 2009. Saeed Heidarpoor, Hamed Khodadadi, and Mohammad Tabatabaei, “Speed control of a DC motor using a fractional order sliding mode controller” IEEE International Conference on Environment and Electrical Engineering and 2017 IEEE Industrial and Commercial Power Systems Europe (EEEIC / I&CPS Europe), 2017.

Improved speed response of DC motor via intelligent techniques (Hassan Farhan Rashag)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 208~216 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp208-216



208

Theory and development of magnetic flux leakage sensor for flaws detection: A review Nor Afandi Sharif1, Rizauddin Ramli2, Abdullah Zawawi Mohamed3, Mohd Zaki Nuawi4 1,2,4

Centre for Materials Engineering and Smart Manufacturing, Universiti Kebangsaan Malaysia, Malaysia 1 Department of Industrial Electronic, German Malaysian Institute, Malaysia 3 PETRONAS Research Sdn Bhd, Malaysia

Article Info

ABSTRACT

Article history:

This paper presents a review of state-of-art in the Magnetic Flux Leakage (MFL) sensor technology, which plays an important role in Nondestructive Testing (NDT) to detect crack and corrosion in ferromagnetic material. The demand of more reliable MFL tools and signal acquisition increase as it has a direct impact on structure integrity and can lead to be major catastrophic upon questionable signal analysis. This is because the size, cost, efficiency, and reliability of the extensive MFL system for NDT applications primarily depend on signal acquisition as a qualitative measure in producing a trustworthy analysis. Therefore, the selection of appropriate tools and methodology plays a major role in determining the comprehensive performance of the system. This paper also reviews an Artificial Neural Network (ANN) and Finite Element Method (FEM) in developing an optimum permeability standard on the test piece.

Received Apr 29, 2019 Revised Jul 1, 2019 Accepted Jul 16, 2019 Keywords: Artificial neural network Fenite element method Magnetic flux leakage Non-destructive test

Corresponding Author: Rizauddin Ramli, Centre for Materials Engineering and Smart Manufacturing, Universiti Kebangsaan Malaysia, 43600 UKM Bangi, Selangor, Malaysia. Email: rizauddin@ukm.edu.my

INTRODUCTION The Nondestructive testing (NDT) technique is a popular method used by many industries, especially in oil and gas (O&G) and as well as in refinery plants [1, 2]. There are many varieties of NDT technique used in industries, which largely depends upon the methods of their applications [3, 4]. Therefore, determining the quality and integrity of the material is the main purpose on NDT without affecting the capability to perform their intentional functions. In order to select the suitable NDT technique, several basic factors need to be considered such as the product diameter, length, wall thickness, fabrication methods, types, location of potential discontinuities and specification requirements. Furthermore, extraneous variables such as a scratch and oxidation that might cause a rejectable indication, even though the product is acceptable is also an important aspect [5-7]. The emerging technology in NDT has triggered a challenge for the new researchers on meeting the demand of more rapid and accurate data requisition, which saw them progressively developing a modern method to increase its reliability. Several NDT processes such as ultrasonic, radiographic, Magnetic Flux Leakage (MFL) and Eddy currents are the NDT method uses to investigate large structure, piping and tank [8]. Currently, petroleum and chemical industries are the main sectors which employ a tank bottom as storage before the secondary process can be established. Most of the storage tanks are built on top of the ground and exposed to the external environment such as humidity and physical damages [9]. As a result, it will lead to the inter-granular attack and causes severe corrosion and deformation on the tank structure. A fatal accident such as an explosion and major leaking will occur because of a small defect at the tank Journal homepage: http://iaescore.com/online/index.php/IJAAS

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

209

bottom. Usually, the storage tanks in petroleum refineries and heavy industries stores hazardous and flammable chemical. Hence, a minor defect may result a catastrophic accident in term of casualties, production interruption and property loss [10]. Among all the NDT methods, the MFL inspection is one of the most reliable methods in oil and gas industries in producing a credible and prompt result and analysis [11, 12]. It has been used as early as 1868 by Institute of Naval Architecture in England in inspecting a cannon tube with compass. The theories behind the MFL area inspection are similar to the principle of Magnetic Particle Inspection (MPI) except for the sensor used between these two methods. The assessment of MFL technique is depended on evaluating the surface of the specimen which magnetised near to the saturation point before the specimen condition can be viewed based on measuring the magnetic leakage field [13]. The MFL sensor detects volumetric changes of the leakage field on the corrosion spot [14]. According to Mix [15], the permeability of magnetized pieces changed drastically, and leakage flux will emanate from the discontinuity. Hence, by measuring the intensity of the flux leakage, severity and condition of the defect can be determined. Pipe, rod, storage tank plates are the types of ferromagnetic parts that have been widely tested by MFL. In general, there are two types of defects existing in the service of the storage tank which are the corrosion and groove. Therefore, it is important to classify the pattern of the defect in order to evaluate the safety of the components [16]. 2.

METHODOLOGY FOR LITERATURE SEARCH This paper addressed a review of previous research related to the theory and development of MFL sensor in plate and pipe detection by NDT method. However, issue related to different kinds of sensors in MFL signal is not considered in this paper. All the papers in the research and commercialization stages were collected from SCOPUS, the SCI web and company websites, to which the searches were limited to the period between the years of 1999 to 2016. 3.

TYPES OF MFL SENSORS There are many types of MFL sensors in magnetic signal detection such as Hall sensor, superconducting quantum interference device (SQUID) and giant magnet resistance (GMR) [17-19]. The mechanism of Hall Effect sensor is by detecting the change of magnetic field and converts into an electrical signal by processing a raw data into voltage [20]. By using Hall component’s sensor, defect information can be acquired by catching the faulty of leakage and process the electrical signal transformed from the magnetic field. Magnetic field distribute homogenously by passing the ferromagnetic material in the absence of defect. If the sensor detects the presence of the imperfection on the surface of the material, the magnetic field passing this area will be distorted. As a result, the resistance around the defect will increase and expose the magnetic leakage at the defective area [21]. There also several sensors in MFL testing that have been introduced to the Non-Destructive Evaluation (NDE) such as fluxgate [22-24], Giant Magneto Impedance (GMI) [25, 26] and Stress Impedance (SI) [27, 28]. Table 1 shows the comparison of MFL types of sensors. Table 1. Magnetic sensor comparison [29] Sensor Hall sensor GMR sensor Fluxgate SI sensor

Head length 10-100µm 10-100µm 10-20mm 1-2mm

Resolution 0.5Oe/±1kOe 0.01Oe/±20Oe 1µOe/±3Oe 0.1Gal/30Gal

Response speed 1MHz 1MHz 5kHz 10kHz

Power consumption 10mW 10mW 1W 5mW

3.1. Hall sensor Hall sensor is constructed according to the Hall-effect principle by embedding a thin electric conductor in sensing a magnetic field fluctuation in a ferromagnetic material [30, 31]. An electrical current will flow through the strip when the magnetic field is perpendicular to the thin strip of conducting material [32, 33]. Hall sensor has been proved of having a high sensitivity in detecting defect in ferro magnetic material using direct current (DC) as an excitation [34]. The hall voltage can be represented as 𝑉 =

(1)

Where Vh, I, B, ne and b represent the Hall voltage, impose current, magnetic induction intensity, Hall element sensitivity and the thickness of the hall element, respectively [35]. By substituting the Hall element sensitivity, Kh = (neb)-1 in to (1), the equation can be shown as follows Theory and development of magnetic flux leakage sensor for flaws detection: A review (Nor Afandi Sharif)

210



ISSN: 2252-8814

𝐵=

(2)

The (2) shows the linear relationship between the magnetic flux density and the hall voltage. The output voltage of the sensor is directly proportional to the flex density. The sensitivity of the Hall sensor’s base silicon is 1mV/mT for a 1mA current and Hall base sensor Indium arsenide (InAs) typically, 2 mV/mT. The sensitivity of the Hall sensor can be increased by 5 mV/mT with thin film of Indium Antimonide (InSb) [36]. The frequency of sensitivity for the Hall sensor is ranged from near DC up to 100 kHz [37] with high resolution contribute to higher sensitivity with minimal power consumption. An early work by Clauzon [38] integrating Hall probe with Eddy current in characterizing a defect under different depth and compared with Finite Element Analysis (FEA). In the observation of the study, they found the correlation between signal characterization and the flaw. Chen [39] investigated a defect classification by integrating Pulse Eddy Current (PEC) which has a pulse excitation providing big frequency information and hall-effect device. The result shows more accurately a 3-dimensional (3D) defect classification compared to the conventional method. A more recent study by Le [40] validated an integration of Dipole Model Method (MDM) using 1024 units of Hall sensor in arrays using alternating current (AC) in estimating the shape and volume of the crack. The time in estimating the crack is proved to be faster and reliable compared to the conventional method by eliminating off-line analysis method. 3.2. Superconducting Quantum Interference Device (SQUID) SQUID sensor known to have an outstanding sensitivity in detecting signal frequency between near DC and low MHz range [41-43]. An experiment done by Faley et al. [44] analysed a spectral density of the SQUID signal which shows a significant decrease of noise value especially in a strong magnetic field as illustrated in Figure 1. The sensor is very useful in detecting a high conductivity material with a deeper defect allocation due to its sensitivity under a low frequency. H. Krause et al. [45] demonstrate a defect scanning using four High Temperature Superconductor (HTS) Direct Current (DC) SQUID that can be operated in a strong magnetic field. The experiment concluded that SQUID has a very excellent performance in detecting a very deep fault. In contrary, in order to achieve a critical temperature, a superconductor in a SQUID has to be cooled off constantly to gain it zero resistance [46]. The main element in the SQUID sensor are the superconducting loop and Radio Frequency (RF) SQUIDs [47] or two DC SQUIDs also known as Josephson junction [48]. The circuit diagram of a dcSQUID sensor comprises of flux transformer and read out electronic [49]. SQUID sensor has a better signal to noise ratio in comparison with a conventional method up to three orders of magnitude for crack exceeding 13 mm of thickness [50].

Figure 1. Spectral density of the SQUID sensor using magnetic field up to 13 mT [44] 3.3. Magnetoresistance (MR) sensor The principle magneto resistive sensor is by detecting a change in inductance caused by Eddy current NDE in magnetic field of the specimen [51, 52]. The magnetic field will be distributed to the defect properties such as the hole or crack on the subject. The magnetic field is also known for the good capability in testing an ultra-high density magnetic recording in a field of NDE. Tsukada et. al. used MFL technique in detecting a defect in spot welds by MR sensor to investigate the interrelation between the strength and the magnetic measurement of the spot weld [53]. The development of the experiment used two induction coils on both end of the yoke and MR sensor in the middle of the specimen. The output voltage is channel into lock in amplifier before the output data can be obtained. Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 208 – 216

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



211

Atzlesberger [54] applied four giant magnetoresistance (GMR) sensors in the experiment to examine a blind hole in ferro magnetic material. However, the sensor likely more sensitive than Anisotropic Magneto Resistor (AMR) due to the design known to have a sensitive axis and nonsensitive axis orthogonally [55]. In the experiment, Helmholtz coil [56] is used by positioning the sensor in the centre of the coil in order to maintain a homogenous magnetic field before the defect properties can be acquired. In order to measure the magnetic flux density on the x-axis between the Helmholtz coil, Biot-Savart´s law [57, 58] is used to determine the potential of uncertainty in normal flux distribution. An experiment by Basu et al. [57] by demonstrated an empirical comparison of electromagnetic from the FEA and field computation using the Biot-Savart’s law. The field computation is important to ensure the homogeneous condition for the sensor before the measured subject can be acquired. 4.

PRINCIPLE AND INFLUENCE PARAMETER IN MFL DETECTION The early work by Sauderson et al. [59] proposed the MFL inspection for above ground storage tanks (ASTs) because of its advantage of being able to cover a large area speedily. At that time, the standard procedure in NDT is by the ultrasonic method which was more tedious and took long time. Qi [60] then explained the advantages of MFL technique for the signal analysis which revealed the correlation between the width of the amplitude of the signal and the defect. It indicates that the MFL signal amplitude varies depend to the width and length of the defect. Furthermore, his finding shows complex relations that interleaves between the variation in of three defect geometries, i.e., width, length and depth of the defects. An analytical function describing the relationship between width, length and signal was carried out by Saha et al. [61] in order to ascertain their depth. The study estimated the depth of the MFL defect in conjunction with the length and width function. Corrosion that was considered as general tended to be oversized in error by less than 5 %. However, there are still some ambiguities that can occur. The flux density of a plate in a tank wall can be magnetized using MFL devices [23] The dynamic of an MFL is highly affected by the existence of the defect and detected by hall sensor [62, 63]. According to Wang et al. [64], the detection of the signal in MFL does not need a pre-processing. It is because the on-line detection can easily carried out today by implementing high degree of automation. There are several types of defects that can be detected by MFL, for example corrosion pitting, external surface and surface defect [65]. Both circumference and axial direction can be detected by the MFL and it is the most widely used method on ASTs for locating defect on the tank floor, although it is sensitive by other factors [66, 67]. The life of the tank can be increased by repairing the defect includes replacing the entire tank floor [68]. Based on the prevailing damage, individual damaged also can be repaired by welding plates. Product containing impurities in the tank could be the cause of the defect of the tank floor and the reaction between soil and environment is the cause of a defect on the bottom to the tank [69]. Mandache and Clapham, [70] stated that the direct (forward) approach on the main geometry identification which comprises three steps, which are establishing MFL runs, inspection and analysis of the result. The MFL sensor detects volumetric changes within the leakage field at the corrosion spot [71, 72]. Hence, by measuring the intensity of the flux leakage, the severity of the defect can be determined. The defect of the test piece is detected by measuring the magnetic leakage which making a detour under the defect when magnetic flux passes through the detected region. Figure 2 shows two conditions of MFL sensor in detecting up presence of defect on the test piece. It shows the correlation of the width, amplitude and depth to the defect signal. Both width and amplitude of the signal are proportional to the defect length (axial dimension) but the depth of the defect also affecting the amplitude of the signal [73]. 5.

DEFECT GEOMETRIES IN MFL SIGNAL Zhang [74] explained the defect geometry parameters such as depth, width and length with their relations between MFL signal features. The vertical component of the defect is unable to be detected by broad and shallow defects and the same result for the defect parallel to the magnetic field. In case of tubes, rods and bars, the longitudinal defect is more easily to be detected with the circular magnetic fields [75]. It is because the magnetic field leaks out from the material if there is any local gradient and geometrical discontinuity in magnetic permeability when the magnetic field applied to a ferromagnetic material. The three geometric components which is radial, axial and circumferential can be regarded as vectors of the MFL signal [76]. In order to achieve a higher defect sizing capability in pipeline inspection gauges (PIGs), tools for pipeline inspection have also evolved to be able to measure three dimensionsional defects. Typically, only defect geometry that is parallel to the pipeline axis namely axial is being measured [77]. The related work for three dimensionsial defects was done by Xiao-Chun et al. [78] whose optimized the MFL inspection tools for ASTs using finite-element modelling. Theory and development of magnetic flux leakage sensor for flaws detection: A review (Nor Afandi Sharif)

212



ISSN: 2252-8814

Figure 2. Principle on the MFL signal detection [73] Feng [79] illustrated three components in the natural cylindrical coordinate named as radial, axial and transverse. All MFL tools are commonly used to measure the axial field components by detecting the disruptions of magnetic field and the volume of the defects. Both traverse and radial field similarly occur when a defect is present and tends to characterize the profile of the feature. Zhang [80] studied the defect allocation on the circumferential component in relative to the sensor spacing, magnetization level and velocity. They study found that crack at the surface of the pipe occurs due to the great difference of pressure, in which the classification wa extremely hard to be seen by the naked eye because the crack is long and very narrow. Garcia [81] further demonstrated the magnitude behaviours between a radial and axial signal. Their research stated that the MFL axial signal's magnitude would be very high compared to the radial signal, and this will happen because the established MFL axial component's direction will parallel the external magneticfield direction. 6.

FEM BACKGROUND IN MFL The Finite Element Method (FEM) based simulation is a technique of investigating the behaviours of the affected field in the microscopic level [82]. It began in late 1980’s in service inspection of buried pipelines [72, 83]. In order to achieve an accurate result of field distribution, a full 3-D simulation is then introduced through commercial software. The material property in FEM can be defined as a forward problem and the reconstruction of the crack shape can be defined as the inverse problem in objective to evaluate a parameter in material e.g.; size of the crack, length and depth in the particular field distribution [84, 85]. According to Tupsie et. al. [86], FEM is a general numerical method used for computer simulation. The advantage of FEM compared to others numerical method is the ability to handle circular geometric problem, non-linear and time dependent. It also the most suitable method in solving the issues of magnetic field effect around the transmission line caused by circular cross- section of voltage conductors. There are several software that can characterizes defect by using raw data, including ANSYS, MagNet, JSOL, COMSOL, Multiphysics, OPERA, MAXWELL, FLUX and others [87-89]. These software has their own ability in computing electrostatics and magnetostatics elements. Harmonic and transient problems involving Eddy current can be well for most codes. There are some remarkable tools in implementation of advance function such as a robust solution or hysteresis loop of a moving conductor induction. The functionality of different tools however evolves rapidly and tend to converge [90]. Ji et al. [91] explained the proved of 2-D FEM is an effective method used to study MFL signal under the different material, different defect shape, magnetizing situation and so forth. However, in 2-D FEM defects are furthermore treated as a 2-D profile rather than actual 3-D geometry, and the resulting MFL signal is the single channel whereas the actual signals are multi-channel. The applications of 3D FEM are to analyze and generalize a potential formulation to the magnetic field in MFL and it also accurately modeled and detailed comparison is done for model with defect and without defect [92]. Through FEM, the characteristic of magnetic field intensity and distribution field can be examined. In addition, the FEM can analyse and quantify the distribution of MFL and alteration of the intensity caused by addition of multiple magnetic circuits in order to identify and analyse the generated defect [92]. Similarly to research done by Zakaria [93], Finite Element Method Magnetic (FEMM) is used to model different type of crack. In order to simulate the output in a small section of the pipeline, new properties have been entered into the software in introducing several cracks. A small displacement in the field carries an actual condition of the pipeline as a result from a disturbance of the magnetic field. The simulation presented a SUS 304 steel pipe material and a coil that surround the pipe. Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 208 – 216

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



213

APPLICATION ON NEURAL NETWORK ON MFL DATA PROCESSING In data mining and clustering, Neural Network (NN) is one of the important tools used as an attempt to build a system that has an ability as a human brain and be capable to learn [94]. Currently, neural network was to use in the field of NDT because its ability to provide a solution in data analysis [95]. Ayad [96] proposed the inversion problem by using the neural Networks for the approximation the mapping from the signal to the defect space. The study points out a crucial problem in signal inversion where the defect profile is recovered from the measured signal by using FEM as an initiator. In tank floor corrosion defect on ASTs, NN also being used to improve accuracy of the defect by establishing Back Propagation Neural Network (BPNN) in order to have a reliable quantitative recognition of width and depth of the defect. Defect in the ASTs can be obtained in three ways, and a lot of MFL testing signal samples are needed in BPNN network [97]. There are three ways of acquiring an MFL testing signal sample of corrosion defect in ASTs in BPNN which need a lot sample: a. The MFL testing corrosion signal must be extract in the tank floor as a sample testing. Sample taken in the process must be authentic and must maintain close to actual situation of MFL testing corrosion of the tank floor. b. Artificial neural network is use as precast corrosion defect in the tank floor, a collection of MFL testing signal is recorded in the corrosion database, external factors such as experimental condition and human factor is largely affected. c. Establishing a finite element as a simulation process to differentiate with the MFL signal collected from the sample of corrosion defect in the tank floor. The interference of numerous detection signals can be avoided and it can solve the problem that the sample number can’t meet. Chen [98] discussed on iterative neural network application on the defect inversion from an MFL signal. There are two kinds of method in solving inverse MFL problems, including model-based iterative method and non-model-based direct method. A non-model based direct method is used to established the relationship between corresponding defect and MFL signal through NN or other tools [99]. Even though the method is advantageous to make a rapid inversion, the prediction of defect profile is not accurate due to lacking of continuous depended of measured MFL signals on defect parameters in non-unique condition. The model-based method on the contrary, use forward model to solve the well-behaved forward problem iteratively in a feedback loop in predicting the MFL signals by begin the algorithm with an initial estimation of the defect and solves the forward problem [100]. The measured and predicted error can be minimised iteratively by updating the defect profile based on gradient-based or optimisation methods [101] 8.

CONCLUSION This paper reviews the concept and developments of MFL inspection with regard to principle and influence in detection, defect geometries, FEM and the application of Neural Network in data processing. Several issues need to be improved on such as a noise, defect profiling, development cost, forward model and inverse model. There are a lot of researches has been conducted on MFL signal analysis. However, there are only a few researches emended comprehensive statistical analysis as an inverse model. There is still room for the improvement and optimization of all these issues. ACKNOWLEDGEMENT This work was supported by the The National University of Malaysia (UKM) under research INDUSTRI-2013-010. REFERENCES [1] N. B. Cameron, "Recommended practice for magnetic flux leakage inspection of atmospheric storage tank floors," 2006. [2] S. Ali and E. Saeedreza, "Intellingent MFL defect detection alghorithm equipped by linear descriminate analysis," IJCNC Int. J. Comput. Networks Commun, vol. 5, pp. 1-21, 2009. [3] J. Dong, Q. Wu, W. Jiang, and Q. Xu, "The review of new NDT methods of metal material fatigue monitoring," Int. J. Hybrid Inf. Technol., vol. 8, no. 8, pp. 225-232, 2015. [4] J. Helal, M. Sofi, and P. Mendis, "Non-destructive testing of concrete: A review of methods," Electron. J. Struct. Eng., vol. 14, no. 1, pp. 97-105, 2015. [5] M. Göktepe, "Investigation of bx and by components of the magnetic flux leakage in ferromagnetic laminated sample," Adv. Mater. Sci. Eng., vol. 2013, 2013. [6] M. Kollar, V. Setnicka, and R. Zubal, "Scanning and data acquisition tools for MFL testing," 2016.

Theory and development of magnetic flux leakage sensor for flaws detection: A review (Nor Afandi Sharif)

214



ISSN: 2252-8814

[7] S. U. N. Liying, et. al., "Comparison of magnetic flux leakage (MFL) and acoustic emission (AE) techniques in corrosion inspection for pressure pipelines," Proc. 31st Chinese Conf., pp. 5375-5378, 2012. [8] C. F. J. Saavedra and S. R. Prada, "Sensitivity analysis of a magnetic circuit for nondestructive testing by the magnetic flux leakage technique," 2014. [9] H. M. Kim et. al., "A study on the measurement of axial cracks in the magnetic flux leakage NDT system," IEEE Int. Conf. Autom. Sci. Eng., pp. 624-629, 2012. [10] J. I. Chang and C. C. Lin, "A study of storage tank accidents," J. Loss Prev. Process Ind., vol. 19, no. 1, pp. 51-59, 2006. [11] M. R. Kandroodi, B. N. Araabi, and M. N. Ahmadabadi, "Detection of natural gas pipeline defects using magnetic flux leakage measurements," 2013. [12] M. M. Salama, B. J. Nestleroth, and M. A. Maes, "Characterization of the accuracy of the MFL pipeline inspection tools," Proc. ASME 2012 31st Int. Conf. Ocean. Offshore Arct. Eng., pp. 1-5, 2016. [13] M. M. Tehranchi, M. Ranjbaran, and H. Eftekhari, "Double core giant magneto-impedance sensors for the inspection of magnetic flux leakage from metal surface cracks," Sensors Actuators, A Phys., vol. 170, no. 1-2, pp. 55-61, 2011. [14] F. Niese, A. Yashan, and H. Willems, "Wall thickness measurement sensor for pipeline inspection using EMAT technology in combination with pulsed eddy current and MFL," Non-destructive testing, Aust., vol. 3, no. 45, pp. 84-87, 2008. [15] P. E. Mix, Introduction to nondestructive testing: A training guide, Second Ed. Canada: A John Wiley & Sons, Inc., Publication, 2005. [16] X. Wu, J. Xu, and Y. Wang, "Signal processing methods for magnetic flux leakage testing based on biospectrum," Proceeding Eur. NDT Conf., 2010. [17] H. Krause and M. Kreutzbruck, "Recent developments in SQUID NDE," Phys. C, vol. 368, pp. 70-79, 2002. [18] A. Sophian, G. Y. Tian, and S. Zairi, "Pulsed magnetic flux leakage techniques for crack detection and characterisation," Sensors and Actuators, vol. 125, pp. 186-191, 2006. [19] D. Perin et. al., "Mass effect on magnetic flux behavior to cite this article mass effect on magnetic flux behavior," 2017. [20] Y. Ma, R. He, and J. Chen, "A method for improving SNR of drill pipe leakage flux testing signals by means of magnetic concentrating effect," IEEE Trans. Magn., vol. 51, no. 9, 2015. [21] M. Yilai, L. Li, J. Kaiwen, and Z. Xulin, "The application of magnetic shielding effect in drill pipe magnetic leakage flux testing," Fifth Conf. Meas. Technol. Mechatronics Autom., no. 1, pp. 1135-1138, 2013. [22] T. Izgi, M. Goktepe, N. Bayri, V. S. Kolat, and S. Atalay, "Crack detection using fluxgate magnetic field sensor," Acta Phys. Pol. A, vol. 125, no. 2, pp. 211-213, 2014. [23] H. Can et. al., "Optimizing the sensing performance of a single-rod fluxgate magnetometer using thin magnetic wires," Meas. Sci. Technol., vol. 26, no. 11, 2015. [24] Z. Zhaoming, S. Taihe, Y. Runqiao, and W. Fu, "A new approach to in-service coiled tubing improved inspection based micro-magnetic technology," Soc. Pet. Eng. - Coiled Tubing Well Interv. Conf. Exhib. 2015, pp. 1-13, 2015. [25] F. Vacher, F. Alves, and C. Gilles-Pascaud, "Eddy current nondestructive testing with giant magneto-impedance sensor," NDT E Int., vol. 40, no. 6, pp. 439-442, 2007. [26] W. Dehui, S. Lingxin, W. Xiaohong, and L. Zhitian, "A novel non-destructive testing method by measuring the change rate of magnetic flux leakage," J. Nondestruct. Eval., vol. 36, no. 2, pp. 24, 2017. [27] K. Mohri, T. Uchiyama, L. P. Shen, C. M. Cai, and L. V. Panina, "Sensitive micro magnetic sensor family utilizing magneto-impedance (MI) and stress-impedance (SI) effects for intelligent measurements and controls," Sensors Actuators A Phys., vol. 91, no. 1-2, pp. 85-90, 2001. [28] N. Bayri and S. Atalay, "Giant stress-impedance effect in Fe71Cr7Si 9B13 amorphous wires," J. Alloys Compd., vol. 381, no. 1-2, pp. 245-249, 2004. [29] K. Mohri, T. Uchiyama, L. P. Shen, C. M. Cai, and L. V. Panina, "Amorphous wire and CMOS IC-based sensitive micro-magnetic sensors (MI sensor and SI sensor) for intelligent measurements and controls," J. Magn. Magn. Mater., vol. 249, no. 1-2, pp. 351-356, 2002. [30] H. M. Kim and G. S. Park, "A study on the estimation of the shapes of axially oriented cracks in CMFL Type NDT system," IEEE Trans. Magn., vol. 50, no. 2, pp. 18-21, 2014. [31] L. Ben Gur, E. Tirosh, A. Segal, G. Markovich, and A. Gerber, "Extraordinary hall-effect in colloidal magnetic nanoparticle films," J. Magn. Magn. Mater., vol. 426, pp. 178-182, 2017. [32] R. Clark, "Rail flaw detection: Overview and needs for future developments," NDT E Int., vol. 37, no. 2, pp. 111-118, 2004. [33] Parker Compumotor Division Hannifin Corporation, "Hall effect sensors," Compumotor - Servo Drive User Guide, pp. 107-118, 2000. [34] K. Kosmas, C. Sargentis, D. Tsamakis, and E. Hristoforou, "Non-destructive evaluation of magnetic metallic materials using hall sensors," J. Mater. Process. Technol., vol. 161, no. 1-2, pp. 359-362, 2005. [35] Y. Lijian, L. Gang, G. Songwei, and A. Coil, "Sensor development and application on the oil-gas pipeline magnetic flux leakage detection," Ninth Int. Conf. Electron. Meas. Instruments, pp. 876-878, 2009. [36] P. Ripka and M. Janosek, "Advances in magnetic field sensors," IEEE Sens. J., vol. 10, no. 6, pp. 1108-1116, 2010. [37] D. G. Park, et. al., "Evaluation of pulsed eddy current response and detection of the thickness variation in the stainless steel," IEEE Trans. Magn., vol. 45, no. 10, pp. 3893-3896, 2009.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 208 – 216

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



215

[38] T. Clauzon, F. Thollon, and A. Nicolas, "Flaws characterization with pulsed eddy currents N.D.T.," IEEE Trans. Magn., vol. 35, no. 3, pp. 1873-1876, 1999. [39] T. Chen, G. Y. Tian, A. Sophian, and P. W. Que, "Feature extraction and selection for defect classification of pulsed eddy current NDT," NDT E Int., vol. 41, no. 6, pp. 467-476, 2008. [40] M. Le, J. Lee, J. Jun, J. Kim, S. Moh, and K. Shin, "Hall sensor array based validation of estimation of crack size in metals using magnetic dipole models," NDT E Int., vol. 53, pp. 18-25, 2013. [41] A. I. Braginski and H. Krause, "Nondestructive evaluation using high-temperature SQUIDs," Phys. C, pp. 179-183, 2000. [42] M. v. Kreutzbruck et. al., "Defect detection and classification using a SQUID based multiple frequency," lEEE Trans. Appl. Supercond., vol. 11, no. 1, pp. 1032-1037, 2001. [43] J. R. Maze et. al., "Nanoscale magnetic sensing with an individual electronic spin in diamond," Nature, vol. 455, no. 7213, pp. 644-647, 2008. [44] F. M. I., P. U., and U. K., "Operation of HTS dc-SQUID sensors in high magnetic fields," IEEE Trans. Appl. Supercond., vol. 9, no. 2, pp. 3386-3391, 1999. [45] H. Krause, W. Wolf, W. Glaas, E. Zimmermann, and M. I. Faley, "SQUID array for magnetic inspection of prestressed concrete bridges," Phys. C., vol. 368, pp. 91-95, 2002. [46] A. Lascialfari, T. Mishonov, A. Rigamonti, P. Tedesco, and A. Varlamov, "Superconducting diamagnetic fluctuations in MgB 2," Phys. Rev. B, vol. 65, no. 0, pp. 4-7, 2002. [47] Y. Zhang, J. Schubert, N. Wolters, M. Banzet, W. Zander, and H. J. Krause, "Substrate resonator for HTS rf SQUID operation," Phys. C Supercond. its Appl., vol. 372-376, no. 1, pp. 282-286, 2002. [48] Y. Makhlin, G. Schön, and A. Shnirman, "Quantum-state engineering with josephson-junction devices," Rev. Mod. Phys., vol. 73, p. 357, 2001. [49] J. Vrba and S. E. Robinson, "SQUID sensor array configurations for magnetoencephalography applications," Supercond. Sci. Technol., vol. 15, no. 9, 2002. [50] P. Gao, C. Wang, Y. Li, and Z. Cong, "Electromagnetic and eddy current NDT in weld inspection: A review," Insight Non-Destructive Test. Cond. Monit., vol. 57, no. 6, pp. 337-345, 2015. [51] K. Tsukada, T. Kiwa, T. Kawata, and Y. Ishihara, "Low-frequency eddy current imaging using mr sensor detecting tangential magnetic field components for nondestructive evaluation," IEEE Trans. Magn., vol. 42, no. 10, pp. 3315-3317, 2006. [52] C. S. Angani, H. G. Ramos, A. L. Ribeiro, and T. J. Rocha, "Evaluation of transient eddy current oscillations response for thickness measurement of stainless-steel plate," Meas. J. Int. Meas. Confed., vol. 90, pp. 59-63, 2016. [53] K. Tsukada, M. Yoshioka, T. Kiwa, and Y. Hirano, "A magnetic flux leakage method using a magnetoresistive sensor for nondestructive evaluation of spot welds," NDT E Int., vol. 44, no. 1, pp. 101-105, 2011. [54] J. Atzlesberger and B. Zagar, "Magnetic flux leakage measurement setup for defect detection," Procedia Eng., vol. 5, pp. 1401-1404, 2010. [55] Y. Kataoka and O. Shinoura, "65.2: Leakage flux testing using a GMR line sensor and a rotating magnetic field," IEEE, pp. 800-803, 2002. [56] J. Lee, J. Hwang, J. Jun, and S. Choi, "Nondestructive testing and crack evaluation of ferromagnetic material by using the linearly integrated hall sensor array," J. Mech. Sci. Technol., vol. 22, no. 12, pp. 2310-2317, 2008. [57] S. Basu, S. S. Pany, P. Bannerjee, and S. Mitra, "Pulsed magnetic field measurement outside finite length solenoid: Experimental results & mathematical verification," J. Electromagn. Anal. Appl., pp. 371-378, 2013. [58] C. Udri and S. Udri, "Biot-savart-laplace dynamical systems," Balk. J. Geom. Its Appl., vol. 1, no. 2, pp. 125-136, 1996. [59] D. H. Sauderson, "The MFE tank floor scanner - a case history," IEE Colloq. non-destructive Eval., vol. 1, 1988. [60] J. Qi, "Experimental study of interference factors and simulation on oil-gas pipeline magnetic flux leakage density signal," Int. Conf. Mechatronics Autom., pp. 3652-3656, 2007. [61] S. Saha, S. Mukhopadhyay, U. Mahapatra, S. Bhattacharya, and G. P. Srivastava, "Empirical structure for characterizing metal loss defects from radial magnetic flux leakage signal," NDT E Int., vol. 43, no. 6, pp. 507-512, 2010. [62] R. K. Amineh, et. al., "A space mapping methodology for defect characterization from magnetic flux leakage measurements," IEEE Trans. Magn., vol. 44, no. 8, pp. 2058-2065, 2008. [63] S. Mukhopadhyay and G. P. Srivastava, "Characterization of metal loss defects from magnetic flux leakage signals with discrete wavelet transform," NDT E Int., vol. 33, no. 1, pp. 57-65, 2000. [64] Y. Wang, Y. Xu, B. Wang, S. Ding, J. Xu, and M. Zheng, "Research on metal atmospheric storage tank inspection method for standard in China," Proc. ASME 2009 Press. Vessel. Pip. Div. Conf., pp. 1-6, 2017. [65] N. Pearson, "Discrimination of top and bottom discontinuities with MFL and the surface topology air-gap reluctance system (STARS)," Non-Destructive Testing – Australia, vol. 49(2), pp. 48-51, 2012. [66] L. Chen, X. Li, X. Li, and Z. Huang, "Signal extraction using ensemble empirical mode decomposition and sparsity in pipeline magnetic flux leakage nondestructive evaluation," Rev. Sci. Instrum., vol. 80, no. 2, p. 25105, 2009. [67] N. R. Pearson, M. A. Boat, R. H. Priewald, M. J. Pate, and J. S. D. Mason, "A study of MFL signals from a spectrum of defect geometries," 18th World Conf. Nondestruct. Test., pp. 16-20, 2012. [68] N. R. Pearson, J. S. D. Mason, and R. H. Priewald, "The influence of maintenance on the life cycle of above-ground storage tanks," Insight Non-Destructive Test. Cond. Monit., vol. 54, no. 6, pp. 311-315, 2012. [69] J. S. D. Mason, N. Pearson, and A. R. Ramırez, "Experimental study to differentiate between top and bottom defects for MFL tank floor inspections," NDT&E Int., vol. 42, pp. 16-21, 2009. [70] C. Mandache and L. Clapham, "A model for magnetic flux leakage signal," J. Phys. D. Appl. Phys., vol. 2427, 2003.

Theory and development of magnetic flux leakage sensor for flaws detection: A review (Nor Afandi Sharif)

216



ISSN: 2252-8814

[71] W. T. Ming, "Master of science (Physic)," Universiti Teknologi Malaysia, 2004. [72] K. Mandal and D. L. Atherton, "A study of magnetic flux-leakage signals," J. Phys. D. Appl. Phys., vol. 31, pp. 3211-3217, 1999. [73] S. Nagu, "Finite element modeling of magnetic flux leakage technique in plates with defect and without defect," Int. J. Mod. Eng. Res., vol. 3, pp. 3452-3455, 2013. [74] Z. Guoguang, "Defect quantitative recognition technology of circumferential magnetic flux leakage inspection in pipeline," Int. Conf. Comput. Appl. Syst. Model. (ICCASM), pp. 311-314, 2010. [75] Y. Li, J. Wilson, and G. Y. Tian, "Experiment and simulation study of 3D magnetic field sensing for magnetic flux leakage defect characterisation," NDT E Int., vol. 40, no. 2, pp. 179-184, 2007. [76] S. M. Dutta, F. H. Ghorbel, and R. K. Stanley, "Dipole modeling of magnetic flux leakage," IEEE Trans. Magn., vol. 45, no. 4, pp. 1959-1965, 2009. [77] G. Kopp and H. Willems, "Sizing limits of metal loss anomalies using tri-axial MFL measurements: A model study," NDT E Int., vol. 55, pp. 75-81, 2013. [78] X. Song, S. Huang, and W. Zhao, "Optimization of the magnetic circuit in the MFL inspection system for storagetank floors," Russ. J. Nondestruct. Test., vol. 43, no. 5, pp. 326-331, 2007. [79] F. Qingshan, J. Sutherland, B. Gu, Y. Wei, and C. Tao, "Evolution of Triax Magnetic Flux Leakage Inspection for Mitigation of Spiral Weld Anomalies," 6th Pipeline Technol. Conf., 2011. [80] Z. Guoguang and L. Penghui, "Signal processing technology of circumferential magnetic flux leakage inspection in pipeline," Third Int. Conf. Meas. Technol. Mechatronics Autom. Signal, no. 5, pp. 5-8, 2011. [81] D. A. T. Garcia, "Modelling and interpetation of magnetic flux leakge," RICE University, Houston, Texas, 2012. [82] Y. Li, G. Y. Tian, and S. Ward, "Numerical simulation on magnetic flux leakage evaluation at high speed," NDT E Int., vol. 39, no. 5, pp. 367-373, 2006. [83] D. L. Atherton and M. G. Daly, "Finite element calculation of magnetic flux leakage detector signals," NDT Int., vol. 20, no. 4, pp. 235-238, 1987. [84] M. Li and D. A. Lowther, "The application of topological gradients to defect identification in magnetic flux leakagetype NDT," IEEE Transactions on Magnetics, vol. 46, no. 8, pp. 3221-3224, 2010. [85] H. Tsuboi, N. Seshima, I. Sebestyen, J. Pavo, S. Gyimothy, and A. Gasparics, "Transient eddy current analysis of pulsed eddy current testing by finite element method," IEEE Trans. Magn., vol. 40, no. 2, pp. 1330-1333, 2004. [86] S. Tupsie, A. Isaramongkolrak, and P. Pao-la-or, "Analysis of electromagnetic field effects using FEM for transmission lines transposition," Eng. Technol., vol. 3, no. 2, pp. 870-874, 2009. [87] J. Daniel and R. M. Abudhahir, "Characterization of defects in magnetic flux leakage (MFL) Images using wavelet transform and neural network," 2014 Int. Conf. on Electronics and Communication Systems (ICECS), 2014. [88] W. Zhang, Y. B. Shi, and Y. J. Li, "Magnetic coil parameters design of oil casing damage detector based on magnetic flux leakage," Proc. 2011 IEEE Int. Conf. Appl. Supercond. Electromagn. Devices, pp. 50-53, 2011. [89] W. S. Singh, B. P. C. Rao, S. Thirunavukkarasu, and T. Jayakumar, "Flexible GMR sensor array for magnetic flux leakage testing of steel track ropes," J. Sensors, vol. 2012, pp. 1-6, 2012. [90] M. Augustyniak and Z. Usarek, "Finite element method applied in electromagnetic NDTE - a Review," J. Nondestruct. Eval., pp. 1-16, 2016. [91] F. Ji, C. Wang, S. Sun, and W. Wang, "Application of 3-D FEM in the simulation analysis for MFL signals," Insight Non-Destructive Test. Cond. Monit., vol. 51, no. 1, pp. 32-35, 2009. [92] G. Ting-yan, Y. Yong-liang, and H. Tian-yu, "The interaction of multiple magnetic circuits in magnetic flux leakage (MFL) inspection," 2014 IEEE Far East Forum on Nondestructive Evaluation/Testing, 2014. [93] Z. Zakaria et. al., "Simulation of magnetic flux leakage (MFL) analysis using FEMM software," 2010 IEEE Symp. Ind. Electron. Appl. (ISIEA 2010), pp. 481-486, 2010. [94] S. W. Lin, K. C. Ying, S. C. Chen, and Z. J. Lee, "Particle swarm optimization for parameter determination and feature selection of support vector machines," Expert Syst. Appl., vol. 35, no. 4, pp. 1817-1824, 2008. [95] D. Y. Singh and A. S. Chauhan, "Neural networks in data mining," J. Theor. Appl. Inf. Technol., vol. 5, no. 1, pp. 1-154, 2009. [96] A. Abdelgahani, et. al., "Solution of inverse problems in electromagnetic NDT using neural networks," Przegląd Elektrotechniczny (Electrical Rev.), vol. 87, no. 9, pp. 330-333, 2011. [97] Z. Yang, G. Dai, W. Li, and Y. Jiang, "Research on magnetic flux leakage signals quantity technology of tank floor corrosion defects based on artificial neural network," Fifth Int. Conf. Nat. Comput. Res., 2009. [98] J. Chen, S. Huang, and W. Zhao, "Three-dimensional defect inversion from magnetic flux leakage signals using iterative neural network," IET Sci. Meas. Technol., vol. 9, pp. 418-426, 2015. [99] Y. Zhang, Z. Ye, and C. Wang, "A fast method for rectangular crack sizes reconstruction in magnetic flux leakage testing," NDT E Int., vol. 42, no. 5, pp. 369-375, 2009. [100] P. Ramuhalli, L. Udpa, and S. S. Udpa, "Neural network-based inversion algorithms in magnetic flux leakage nondestructive evaluation," J. Appl. Phys., vol. 93, no. 10, pp. 8274-8276, 2003. [101] H. Wenhua and Q. Peiwen, "2-D defect reconstruction from MFL signals based on genetic optimization algorithm," IEEE Int. Conf. Ind. Technol., pp. 508-513, 2005.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 208 – 216

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 217~224 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp217-224



217

Active power loss reduction by opposition based kidney search algorithm K. Lenin Department of EEE, Prasad V. Potluri Siddhartha Institute of Technology, India

Article Info

ABSTRACT

Article history:

In this work Opposition based Kidney Search Algorithm (OKS) is used to solve the optimal reactive power problem. Kidney search algorithm imitates the various sequences of functions done by biological kidney. Opposition based learning (OBL) stratagem is engaged to commence the algorithm. This is to make certain high-quality of preliminary population and to expand the exploration steps in case of stagnation of the most excellent solutions. Opposition based learning (OBL) is one of the influential optimization tools to boost the convergence speed of different optimization techniques. The thriving implementation of the OBL engages evaluation of opposite population and existing population in the similar generation to discover the superior candidate solution of a given reactive power problem. Proposed Opposition based Kidney Search Algorithm (OKS) has been tested in standard IEEE 14, 30, 57,118,300 bus test systems and simulation results show that the proposed algorithm reduced the real power loss efficiently.

Received Apr 30, 2019 Revised Jul 11, 2019 Accepted Jul 27, 2019 Keywords: Opposition based kidney search algorithm Optimal reactive power Transmission loss

Corresponding Author: K. Lenin, Department of EEE, Prasad V. Potluri Siddhartha Institute of Technology, Kanuru, Vijayawada, Andhra Pradesh - 520007. Email: gklenin@gmail.com

INTRODUCTION Reactive power problem plays a key role in secure and economic operations of power system. Optimal reactive power problem has been solved by variety of types of methods [1-6]. Nevertheless, numerous scientific difficulties are found while solving problem due to an assortment of constraints. Evolutionary techniques [7-15] are applied to solve the reactive power problem, but the main problem is many algorithms get stuck in local optimal solution & failed to balance the Exploration & Exploitation during the search of global solution. In this work Opposition based Kidney Search Algorithm (OKS) is used to solve the optimal reactive power problem. Kidney search algorithm imitates the various sequences of functions done by biological kidney. In preliminary segment, a capricious population of feasible solutions is formed and re-absorption, secretion, excretion are replicated in the exploration procedure to verify different conditions well-established to the algorithm. Opposition based learning (OBL) stratagem is engaged to commence the algorithm. This is to make certain high-quality of preliminary population and to expand the exploration steps in case of stagnation of the most excellent solutions. Opposition based learning (OBL) is one of the influential optimization tools to boost the convergence speed of different optimization techniques. The thriving implementation of the OBL engages evaluation of opposite population and existing population in the similar generation to discover the superior candidate solution of a given reactive power problem. In all oppositional based optimization; the conception of OBL is used in the initialization procedure and as well as in each iteration using the generation jumping rate, Jr. Proposed Opposition based Kidney Search Algorithm (OKS) has been tested in standard IEEE 14, 30, 57,118,300 bus test systems and simulation results show the projected algorithm reduced the real power loss comprehensively. Journal homepage: http://iaescore.com/online/index.php/IJAAS

218 2.



ISSN: 2252-8814

PROBLEM FORMULATION Objective of the problem is to reduce the true power loss F=P =∑

∈

V + V − 2V V cosθ

(1)

Voltage deviation given as follows F = P + ω × Voltage Deviation

(2)

Voltage deviation given by Voltage Deviation = ∑

|V − 1|

(3)

Constraint (Equality) (4)

P =P +P Constraints (Inequality) P

≤P

(5)

≤P ,i ∈ N

(6)

≤Q ≤Q

≤V ≤V

,i ∈ N

(7)

≤T ≤T

,i ∈ N

(8)

≤Q ≤Q

,i ∈ N

(9)

OPPOSITION BASED KIDNEY SEARCH ALGORITHM Kidney search algorithm imitates the various sequences of functions done by biological kidney. Filtration, Re-absorption, Secretion, Excretion plays key function in the function of the kidney. In preliminary segment, a capricious population of feasible solutions is formed and re-absorption, secretion, excretion are replicated in the exploration procedure to verify different conditions well-established to the algorithm. Algorithm is built to perk up the exploration even a potential solution stirred to waste (W) and it will be fetch back to the filtered blood (FB). Glomerular filtration rate (GFR) test is employed to authenticate the robustness of kidney. The test roughly gives the capability of blood that pass all the way through the glomeruli every minute. Depending on the GFR test outcome which is less than 15 or falls between 15 and 60 or is more than 60 a meticulous action will be accomplished. This process executed to perk up the rate of exploration and to discover the optimal solution. The GFR testing process is added at the ending of iterations. When GFR level is less than 15, the method is recurring with the population in Filtered Blood. When GFR level is between 15 and 60, development of realistic solutions in Filtered blood is applied as a treatment for abridged kidney function. This sequence augments the searching capability and is designed to assist the algorithm in detection of improved solution. If the GFR level is larger than 60, then kidney function is ordinary, in which case no extra development is added to algorithm. Movement equation as follows 𝑍

= 𝑍 + 𝑟𝑎𝑛𝑑(𝑍

−𝑍)

(10)

Filtering of the solutions is done with a filtration rate and Calculation of the filtration rate (𝑙 ) is done using the following equation 𝑙 =𝛽×

∑

(

)

(11)

𝛽 is a constant value between 0 and 1 and is attuned in advance. s represents the size of the population. 𝑓(𝑦 ) represents an objective function of solution y at ith iteration [16]. In every iteration, Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 217 – 224

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



219

previous to integration the Filtration of blood (FB) and waste (W) will be population for the subsequent iteration, the algorithm compute the GFR level based on the fr in FB Glomerular filtration rate

= 120 −

∗

(12)

Define the Population Calculate approximate solution in the population Most excellent solution 𝑍 , is found By (11) find the Filtration rate- 𝑙 , Define waste (W) Define filtered blood (FB) Number of iteration will be found Do while (iteration < 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑠) For 𝑍 ; compute new 𝑍 by using (10) Check the value of 𝑍 using 𝑙 If 𝑍 allocated to W then place on re-absorption and produce 𝑍 by using (10) If re-absorption is not fulfilled then 𝑍 will not be part of FB Eradicate 𝑍 from W (excretion) Place randomly Z into W to exchange 𝑍 End if 𝑍 is reabsorbed Else If it is superior than the 𝑍 𝑖𝑛 𝐹𝐵 𝑍 is secreted Calculate the GFR level solutions in FB by using (12) 𝑖𝑓 15 < 𝐺𝐹𝑅 𝑙𝑒𝑣𝑒𝑙 < 60 ; 𝑡ℎ𝑒𝑛 𝑖𝑚𝑝𝑙𝑒𝑚𝑒𝑛𝑡 𝑚𝑜𝑣𝑒𝑚𝑒𝑛𝑡 𝑜𝑓 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛𝑠 𝑖𝑛 𝐹𝐵 End if 𝑖𝑓 𝐺𝐹𝑅 𝑙𝑒𝑣𝑒𝑙 < 15 ; 𝑡ℎ𝑒𝑛 𝑎𝑙𝑔𝑜𝑟𝑖𝑡ℎ𝑚 𝑝𝑟𝑜𝑐𝑒𝑒𝑑𝑒𝑑 𝑤𝑖𝑡ℎ 𝑡ℎ𝑒 𝑠𝑎𝑚𝑒 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑖𝑛 𝐹𝐵 End if End if End for Rank the 𝑍 from FB and modernize the 𝑍 Merge W and FB By (11) amend filtration rate 𝑙 End while Return 𝑍 In this work Opposition based Kidney Search Algorithm (OKS) is used to solve the problem. Opposition based learning (OBL) stratagem is engaged to commence the algorithm. This is to make certain high-quality of preliminary population and to expand the exploration steps in case of stagnation of the most excellent solutions. Opposition based learning (OBL) is one of the influential optimization tools to boost the convergence speed of different optimization techniques [17]. The thriving implementation of the OBL engages evaluation of opposite population and existing population in the similar generation to discover the superior candidate solution of a given reactive power problem. The conception of opposite number requirements is to be defined to explain OBL. Let 𝑁 (𝑁 ∈ [𝑥, 𝑦]) be a real number and the 𝑁 (opposite number) can be defined as follows 𝑁 =𝑥+𝑦−𝑁

(13)

In the exploration space it has been extended as 𝑁 =𝑥 +𝑦 −𝑁

(14)

Where (𝑁 , 𝑁 , . . 𝑁 ) is a point in the dimensional search space, 𝑁 ∈ [𝑥 , 𝑦 ], 𝑖 → {1,2,3, . . 𝑑} In all oppositional based optimization; the conception of OBL is used in the initialization procedure and as well as in each iteration using the generation jumping rate, Jr.

Active power loss reduction by opposition based kidney search algorithm (K. Lenin)

220 a. b. c. d. e. f. g. h. i. j. k. l.



ISSN: 2252-8814

Begin Engender OBL based population Calculate each “Z” in the population and fix the “Zbest” Produce new-fangled “Z” for “Zi” based on mutual information based switching Apply the filtration operator Is “Zi” assigned as “W”? ; if “Yes” apply the reabsorption operator; or check is “Zi” better than the “Zworst” – if yes then secrete “Zworst” from FB or secrete “Zi” Can “Znew” be assigned as FB? If yes remove “Zi” from W and insert a random “Z” into “W” Have all “Zs” have been met? If “yes” engender 𝑍̅ or else go to step “d” Is "𝑍̅ " better than the “Zworst” in FB? if “yes” replace “Zworst” with "𝑍̅ " Or else update “Zbest” , merge W ,FB modernize the filtration rate Is end criterion reached? if yes stop or else go to step “d”

SIMULATION STUDY At first in standard IEEE 14 bus system the validity of the proposed Opposition based Kidney Search Algorithm (OKS) has been tested, Table 1 shows the constraints of control variables Table 2 shows the limits of reactive power generators and comparison results are presented in Table 3. Table 1. Constraints of control variables System

Variables

IEEE 14 Bus

Generator Voltage Transformer Tap VAR Source

Minimum (PU) 0.95 o.9 0

Maximum (PU) 1.1 1.1 0.20

Table 2. Constrains of reactive power generators System

Variables 1 2 3 6 8

IEEE 14 Bus

Q Minimum (PU) 0 -40 0 -6 -6

Q Maximum (PU) 10 50 40 24 24

Table 3. Simulation results of IEEE −14 system Control variables 𝑉𝐺−1 𝑉𝐺−2 𝑉𝐺−3 𝑉𝐺−6 𝑉𝐺−8 𝑇𝑎𝑝 8 𝑇𝑎𝑝 9 𝑇𝑎𝑝 10 𝑄𝐶−9 𝑃𝐺 𝑄𝐺 (Mvar) Reduction in PLoss (%) Total PLoss (Mw) NR* - Not reported.

Base case 1.060 1.045 1.010 1.070 1.090 0.978 0.969 0.932 0.19 272.39 82.44 0 13.550

MPSO [18] 1.100 1.085 1.055 1.069 1.074 1.018 0.975 1.024 14.64 271.32 75.79 9.2 12.293

PSO [18] 1.100 1.086 1.056 1.067 1.060 1.019 0.988 1.008 0.185 271.32 76.79 9.1 12.315

EP [18] NR* 1.029 1.016 1.097 1.053 1.04 0.94 1.03 0.18 NR* NR* 1.5 13.346

SARGA [18] NR* 1.060 1.036 1.099 1.078 0.95 0.95 0.96 0.06 NR* NR* 2.5 13.216

OKS 1.012 1.028 1.024 1.016 1.019 0.910 0.902 0.915 0.146 271.09 75.17 18.75 11.009

Then the proposed Opposition based Kidney Search Algorithm (OKS) has been tested, in IEEE 30 Bus system. Table 4 shows the constraints of control variables, Table 5 shows the limits of reactive power generators and comparison results are presented in Table 6. Table 4. Constraints of control variables System

Variables

IEEE 30 Bus

Generator Voltage Transformer Tap VAR Source

Minimum (PU) 0.95 o.9 0

Maximum (PU) 1.1 1.1 0.20

Table 5. Constrains of reactive power generators System

Variables

IEEE 30 Bus

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 217 – 224

1 2 5 8 11 13

Q Minimum (PU) 0 -40 -40 -10 -6 -6

Q Maximum (PU) 10 50 40 40 24 24

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

221

Table 6. Simulation results of IEEE −30 system Control variables 𝑉𝐺−1 𝑉𝐺−2 𝑉𝐺−5 𝑉𝐺−8 𝑉𝐺−12 VG-13 Tap11 Tap12 Tap15 Tap36 QC10 QC24 𝑃𝐺 (MW) 𝑄𝐺 (Mvar) Reduction in PLoss (%) Total PLoss (Mw) NR* - Not reported.

Base case 1.060 1.045 1.010 1.010 1.082 1.071 0.978 0.969 0.932 0.968 0.19 0.043 300.9 133.9 0 17.55

MPSO [18] 1.101 1.086 1.047 1.057 1.048 1.068 0.983 1.023 1.020 0.988 0.077 0.119 299.54 130.83 8.4 16.07

PSO [18] 1.100 1.072 1.038 1.048 1.058 1.080 0.987 1.015 1.020 1.012 0.077 0.128 299.54 130.94 7.4 16.25

EP [18] NR* 1.097 1.049 1.033 1.092 1.091 1.01 1.03 1.07 0.99 0.19 0.04 NR* NR* 6.6 16.38

SARGA [18] NR* 1.094 1.053 1.059 1.099 1.099 0.99 1.03 0.98 0.96 0.19 0.04 NR* NR* 8.3 16.09

OKS 1.028 1.029 1.017 1.028 1.019 1.026 0.920 0.921 0.922 0.929 0.094 0.107 297.08 131.78 14.46 15.012

Then the proposed Opposition based Kidney Search Algorithm (OKS) has been tested, in IEEE 57 Bus system. Table 7 shows the constraints of control variables, Table 8 shows the limits of reactive power generators and comparison results are presented in Table 9. Table 7. Constraints of control variables System

Variables

IEEE 57 Bus

Generator Voltage Transformer Tap VAR Source

Minimum (PU) 0.95 0.9 0

Maximum (PU) 1.1 1.1 0.20

Table 8. Constrains of reactive power generators System IEEE 57 Bus

Variables 1 2 3 6 8 9 12

Q Minimum (PU) -140 -17 -10 -8 -140 -3 -150

Q Maximum (PU) 200 50 60 25 200 9 155

Table 9. Simulation results of IEEE −57 system Control variables 𝑉𝐺 1 𝑉𝐺 2 𝑉𝐺 3 𝑉𝐺 6 𝑉𝐺 8 𝑉𝐺 9 𝑉𝐺 12 𝑇𝑎𝑝 19 𝑇𝑎𝑝 20 𝑇𝑎𝑝 31 𝑇𝑎𝑝 35 𝑇𝑎𝑝 36 𝑇𝑎𝑝 37 𝑇𝑎𝑝 41 𝑇𝑎𝑝 46 𝑇𝑎𝑝 54 𝑇𝑎𝑝 58 𝑇𝑎𝑝 59 𝑇𝑎𝑝 65 𝑇𝑎𝑝 66 𝑇𝑎𝑝 71 𝑇𝑎𝑝 73 𝑇𝑎𝑝 76 𝑇𝑎𝑝 80 𝑄𝐶 18 𝑄𝐶 25 𝑄𝐶 53 𝑃𝐺 (MW) 𝑄𝐺 (Mvar) Reduction in PLoss (%) Total PLoss (Mw) NR* - Not reported.

Base case 1.040 1.010 0.985 0.980 1.005 0.980 1.015 0.970 0.978 1.043 1.000 1.000 1.043 0.967 0.975 0.955 0.955 0.900 0.930 0.895 0.958 0.958 0.980 0.940 0.1 0.059 0.063 1278.6 321.08 0 27.8

MPSO [18] 1.093 1.086 1.056 1.038 1.066 1.054 1.054 0.975 0.982 0.975 1.025 1.002 1.007 0.994 1.013 0.988 0.979 0.983 1.015 0.975 1.020 1.001 0.979 1.002 0.179 0.176 0.141 1274.4 272.27 15.4 23.51

PSO [18] 1.083 1.071 1.055 1.036 1.059 1.048 1.046 0.987 0.983 0.981 1.003 0.985 1.009 1.007 1.018 0.986 0.992 0.990 0.997 0.984 0.990 0.988 0.980 1.017 0.131 0.144 0.162 1274.8 276.58 14.1 23.86

CGA [18] 0.968 1.049 1.056 0.987 1.022 0.991 1.004 0.920 0.920 0.970 NR* NR* 0.900 0.910 1.100 0.940 0.950 1.030 1.090 0.900 0.900 1.000 0.960 1.000 0.084 0.008 0.053 1276 309.1 9.2 25.24

AGA [18] 1.027 1.011 1.033 1.001 1.051 1.051 1.057 1.030 1.020 1.060 NR* NR* 0.990 1.100 0.980 1.010 1.080 0.940 0.950 1.050 0.950 1.010 0.940 1.000 0.016 0.015 0.038 1275 304.4 11.6 24.56

OKS 1.012 1.018 1.028 1.027 1.029 1.030 1.040 0.900 0.909 0.906 1.018 1.014 1.008 0.940 1.010 0.920 0.930 0.921 1.005 0.931 1.008 1.010 0.942 1.004 0.152 0.141 0.120 1272.92 272.01 24.12 21.092

Active power loss reduction by opposition based kidney search algorithm (K. Lenin)

222



ISSN: 2252-8814

Then the proposed Opposition based Kidney Search Algorithm (OKS) has been tested, in IEEE 118 Bus system. Table 10 shows the constraints of control variables and the comparison results are presented in Table 11 as shown in appendix. Table 10. Constraints of control variables System IEEE 118 Bus

Variables Generator Voltage Transformer Tap VAR Source

Minimum (PU) 0.95 0.9 0

Maximum (PU) 1.1 1.1 0.20

Then IEEE 300 bus system [19] is used as test system to validate the performance of the Opposition based Kidney Search Algorithm (OKS). Table 12 shows the comparison of real power loss obtained after optimization. Table 12. Comparison of real power loss Parameter PLOSS (MW)

Method EGA [20] 646.2998

Method EEA [20] 650.6027

Method CSA [21] 635.8942

OKS 613.0974

CONCLUSION In this work Opposition based Kidney Search Algorithm (OKS) has been successfully applied for solving optimal reactive power problem. Opposition based learning (OBL) stratagem is engaged to commence the algorithm. The prosperous execution of the OBL employ assessment of opposite population and existing population in the analogous generation to find out the better candidate solution of a given reactive power problem. In all oppositional based optimization; the conception of OBL is used in the initialization procedure and as well as in each iteration using the generation jumping rate. Proposed Opposition based Kidney Search Algorithm (OKS) has been tested in standard IEEE 14, 30, 57,118,300 bus test systems and simulation results show that the proposed algorithm reduced the real power loss efficiently. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

[11] [12] [13]

K. Y. Lee., "Fuel-cost minimisation for both real and reactive-power dispatches," Proceedings Generation, Transmission and Distribution Conference, vol. 131(3), pp. 85-93, 1984. N. I. Deeb, "An efficient technique for reactive power dispatch using a revised linear programming approach," Electric Power System Research, vol. 15(2), pp. 121-134, 1998. M. R. Bjelogrlic, M. S. Calovic, and B. S. Babic, "Application of Newton’s optimal power flow in voltage/reactive power control," IEEE Trans Power System, vol. 5(4), pp. 1447-1454, 1990. S. Granville, "Optimal reactive dispatch through interior point methods," IEEE Transactions on Power System, vol. 9(1), pp. 136-146, 1994. N. Grudinin, "Reactive power optimization using successive quadratic programming method," IEEE Transactions on Power System, vol. 13(4), pp. 1219-1225, 1998. Ng Shin Mei, R. Sulaiman, M. H. Mustaffa, Z., and Daniyal, H., "Optimal reactive power dispatch solution by loss minimization using moth-flame optimization technique," Appl. Soft Comput., vol. 59, pp. 210-222, 2017. Chen, G. Liu, L. Zhang, Z, and Huang, S., "Optimal reactive power dispatch by improved GSA-based algorithm with the novel strategies to handle constraints," Appl. Soft Comput. 2017, vol. 50, pp. 58-70, 2017. Naderi, E, Narimani, H, Fathi, M, and Narimani, M.R., "A novel fuzzy adaptive configuration of particle swarm optimization to solve large-scale optimal reactive power dispatch," Appl. Soft Comput. 2017, vol. 53, pp. 441-456, 2017. Heidari, A.A, Ali Abbaspour, R, and Rezaee Jordehi, A., "Gaussian bare-bones water cycle algorithm for optimal reactive power dispatch in electrical power systems," Appl. Soft Comput. 2017, vol. 57, pp. 657-671, 2017. Mahaletchumi Morgan, Nor Rul Hasma Abdullah, Mohd Herwan Sulaiman, Mahfuzah Mustafa, and Rosdiyana Samad., "Benchmark Studies on Optimal Reactive Power Dispatch (ORPD) Based Multi-objective Evolutionary Programming (MOEP) Using Mutation Based on Adaptive Mutation Adapter (AMO) and Polynomial Mutation Operator (PMO)," Journal of Electrical Systems, 12-1, 2016. Rebecca Ng Shin Mei, Mohd Herwan Sulaiman, and Zuriani Mustaffa., "Ant Lion Optimizer for Optimal Reactive Power Dispatch Solution," Journal of Electrical Systems, Special Issue AMPE2015, pp. 68-74, 2016. P. Anbarasan and T. Jayabarathi, "Optimal reactive power dispatch problem solved by symbiotic organism search algorithm," Innovations in Power and Advanced Computing Technologies, pp. 1-8, 2017. Gagliano A. and Nocera F. "Analysis of the performances of electric energy storage in residential applications," International Journal of Heat and Technology, vol. 35(1), pp. S41-S48, 2017.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 217 – 224

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

223

[14] Caldera M, Ungaro P, Cammarata G, and Puglisi G., "Survey-based analysis of the electrical energy demand in Italian households," Mathematical Modelling of Engineering Problems, vol. 5(3), pp. 217-224, 2018. [15] M. Basu., "Quasi-oppositional differential evolution for optimal reactive power dispatch," Electrical Power and Energy Systems, vol. 78, pp. 29-40, 2016. [16] Jaddi, N. S., J. Alvankarian, and S. Abdullah., "Kidney-inspired algorithm for optimization problems," Communications in Nonlinear Science and Numerical Simulation, vol. 42, pp. 358-369, 2017. [17] Tizhoosh HR., "Opposition-based learning: a new scheme for machine intelligence," Proceeding of the international conference on computational intelligence for modeling, control and automation, pp. 695-701, 2005. [18] Ali Nasser Hussain, Ali Abdulabbas Abdullah, and Omar Muhammed Neda., "Modified Particle Swarm Optimization for Solution of Reactive Power Dispatch," Research Journal of Applied Sciences, Engineering and Technology, vol. 15(8), pp. 316-327, 2018. [19] IEEE, The IEEE-test systems, 1993. [Online] Available: http://www.ee.washington.edu/trsearch/pstca/ [20] S. S. Reddy, et al., "Faster evolutionary algorithm based optimal power flow using incremental variables," Electrical Power and Energy Systems, vol. 54, pp. 198-210, 2014. [21] S. Surender Reddy., "Optimal Reactive Power Scheduling Using Cuckoo Search Algorithm," International Journal of Electrical and Computer Engineering, vol. 7(5), pp. 2349-2356, 2017.

APPENDIX Table 11. Simulation results of IEEE −118 system 𝑉𝐺 1 𝑉𝐺 4 𝑉𝐺 6 𝑉𝐺 8 𝑉𝐺 10 𝑉𝐺 12 𝑉𝐺 15 𝑉𝐺 18 𝑉𝐺 19 𝑉𝐺 24 𝑉𝐺 25 𝑉𝐺 26 𝑉𝐺 27 𝑉𝐺31 𝑉𝐺 32 𝑉𝐺 34 𝑉𝐺 36 𝑉𝐺 40 𝑉𝐺 42 𝑉𝐺 46 𝑉𝐺 49 𝑉𝐺 54 𝑉𝐺 55 𝑉𝐺56 𝑉𝐺 59 𝑉𝐺 61 𝑉𝐺 62 𝑉𝐺 65 𝑉𝐺 66 𝑉𝐺 69 𝑉𝐺 70 𝑉𝐺 72 𝑉𝐺 73 𝑉𝐺 74 𝑉𝐺 76 𝑉𝐺 77 𝑉𝐺 80 𝑉𝐺 85 𝑉𝐺 87 𝑉𝐺 89 𝑉𝐺 90 𝑉𝐺 91 𝑉𝐺 92 𝑉𝐺 99 𝑉𝐺 100 𝑉𝐺 103 𝑉𝐺 104 𝑉𝐺 105

Base case 0.955 0.998 0.990 1.015 1.050 0.990 0.970 0.973 0.962 0.992 1.050 1.015 0.968 0.967 0.963 0.984 0.980 0.970 0.985 1.005 1.025 0.955 0.952 0.954 0.985 0.995 0.998 1.005 1.050 1.035 0.984 0.980 0.991 0.958 0.943 1.006 1.040 0.985 1.015 1.000 1.005 0.980 0.990 1.010 1.017 1.010 0.971 0.965

MPSO [18] 1.021 1.044 1.044 1.063 1.084 1.032 1.024 1.042 1.031 1.058 1.064 1.033 1.020 1.023 1.023 1.034 1.035 1.016 1.019 1.010 1.045 1.029 1.031 1.029 1.052 1.042 1.029 1.054 1.056 1.072 1.040 1.039 1.028 1.032 1.005 1.038 1.049 1.024 1.019 1.074 1.045 1.052 1.058 1.023 1.049 1.045 1.035 1.043

PSO [18] 1.019 1.038 1.044 1.039 1.040 1.029 1.020 1.016 1.015 1.033 1.059 1.049 1.021 1.012 1.018 1.023 1.014 1.015 1.015 1.017 1.030 1.020 1.017 1.018 1.042 1.029 1.029 1.042 1.054 1.058 1.031 1.039 1.015 1.029 1.021 1.026 1.038 1.024 1.022 1.061 1.032 1.033 1.038 1.037 1.037 1.031 1.031 1.029

PSO [18] 1.085 1.042 1.080 0.968 1.075 1.022 1.078 1.049 1.077 1.082 0.956 1.080 1.087 0.960 1.100 0.961 1.036 1.091 0.970 1.039 1.083 0.976 1.010 0.953 0.967 1.093 1.097 1.089 1.086 0.966 1.078 0.950 0.972 0.971 0.960 1.078 1.078 0.956 0.964 0.974 1.024 0.961 0.956 0.954 0.958 1.016 1.099 0.969

CLPSO [18] 1.033 1.055 0.975 0.966 0.981 1.009 0.978 1.079 1.080 1.028 1.030 0.987 1.015 0.961 0.985 1.015 1.084 0.983 1.051 0.975 0.983 0.963 0.971 1.025 1.000 1.077 1.048 0.968 0.964 0.957 0.976 1.024 0.965 1.073 1.030 1.027 0.985 0.983 1.088 0.989 0.990 1.028 0.976 1.088 0.961 0.961 1.012 1.068

OKS 1.012 1.016 1.028 1.019 1.012 1.028 1.019 1.006 1.015 1.014 1.013 1.022 0.909 0.906 0.905 1.014 1.003 0.950 1.008 1.010 1.011 0.912 0.929 0.944 0.932 0.910 0.922 1.006 1.049 1.012 1.010 1.008 1.009 1.002 1.006 1.008 1.004 1.010 1.002 1.031 1.010 1.009 1.018 1.005 1.003 1.009 1.017 1.028

Active power loss reduction by opposition based kidney search algorithm (K. Lenin)

224



ISSN: 2252-8814 Table 11. Simulation results of IEEE −118 system (continued)

𝑉𝐺 107 𝑉𝐺 110 𝑉𝐺 111 𝑉𝐺 112 𝑉𝐺 113 𝑉𝐺 116 𝑇𝑎𝑝 8 𝑇𝑎𝑝 32 𝑇𝑎𝑝 36 𝑇𝑎𝑝 51 𝑇𝑎𝑝 93 𝑇𝑎𝑝 95 𝑇𝑎𝑝 102 𝑇𝑎𝑝 107 𝑇𝑎𝑝 127 𝑄𝐶 34 𝑄𝐶 44 𝑄𝐶 45 𝑄𝐶 46 𝑄𝐶 48 𝑄𝐶 74 𝑄𝐶 79 𝑄𝐶 82 𝑄𝐶 83 𝑄𝐶 105 𝑄𝐶 107 𝑄𝐶 110 PG(MW) QG(MVAR) Reduction in PLOSS (%) Total PLOSS (Mw) NR* - Not reported.

Base case 0.952 0.973 0.980 0.975 0.993 1.005 0.985 0.960 0.960 0.935 0.960 0.985 0.935 0.935 0.935 0.140 0.100 0.100 0.100 0.150 0.120 0.200 0.200 0.100 0.200 0.060 0.060 4374.8 795.6 0 132.8

MPSO [18] 1.023 1.032 1.035 1.018 1.043 1.011 0.999 1.017 0.994 0.998 1.000 0.995 1.024 0.989 1.010 0.049 0.026 0.196 0.117 0.056 0.120 0.139 0.180 0.166 0.189 0.128 0.014 4359.3 604.3 11.7 117.19

PSO [18] 1.008 1.028 1.039 1.019 1.027 1.031 0.994 1.013 0.997 1.000 0.997 1.020 1.004 1.008 1.009 0.048 0.026 0.197 0.118 0.056 0.120 0.140 0.180 0.166 0.190 0.129 0.014 4361.4 653.5 10.1 119.34

PSO [18] 0.965 1.087 1.037 1.092 1.075 0.959 1.011 1.090 1.003 1.000 1.008 1.032 0.944 0.906 0.967 0.093 0.093 0.086 0.089 0.118 0.046 0.105 0.164 0.096 0.089 0.050 0.055 NR* * NR* 0.6 131.99

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 217 – 224

CLPSO [18] 0.976 1.041 0.979 0.976 0.972 1.033 1.004 1.060 1.000 1.000 0.992 1.007 1.061 0.930 0.957 0.117 0.098 0.094 0.026 0.028 0.005 0. 148 0.194 0.069 0.090 0.049 0.022 NR* NR* 1.3 130.96

OKS 1.012 1.016 1.019 1.092 1.016 1.018 0.932 1.004 0.949 0.912 1.018 0.930 1.012 0.930 1.014 0.010 0.024 0.110 0.109 0.020 0.112 0.109 0.140 0.106 0.110 0.121 0.015 4358.02 605.97 13.38 115.02

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 225~231 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp225-231



225

A novel impedance source fed H-type flying capacitor multilevel inverter C. R. Balamurugan, P. Vijayakumar, T. Sengolrajan Department of EEE, Karpagam College of Engineering, India

Article Info

ABSTRACT

Article history:

In this paper, simulation using MATLAB/SIMULINK is performed with bipolar triangular fixed amplitude multi-carrier Phase Disposition (PD) PWM strategy with sine wave, Third Harmonic Injection, 60-degree Pulse Width Modulation and stepped wave reference for the chosen impedance Source based H-Type flying capacitor Multilevel Inverter (ISBH-Type FCMLI). The root means square value of the fundamental component and Total Harmonic Distortion of the output voltage which are the most important performance indices for the chosen inverter topologies are evaluated presented and compared for various references through duty ratios. From the simulation results it is observed that for various references the THD is almost similar but the root mean square value in terms of voltage is more for THI, 60-degree PWM and stepped wave reference with phase disposition strategy. The results are obtained for ma (amplitude modulation index) < 1 (under amplitude modulation index), ma=1 (normal amplitude modulation index) and ma > 1 (over amplitude modulation index).

Received Apr 23, 2019 Revised Jun 7, 2019 Accepted Jul 14, 2019 Keywords: Five level Impedance source MLI Stepped wave THD THI Trapezoidal

Corresponding Author: C. R. Balamurugan, Department of EEE, Karpagam College of Enigneering, Myleripalayam Village, Othakkal Mandapam, Tamil Nadu 641032, India. Email: crbalain2010@gmail.com

INTRODUCTION Multilevel Inverter (MLI) is made up of multiple switches. Multilevel inverter [1] enables the use of environmental friendly energy sources like solar cells and fuel cells. The main feature of this MLI is its ability to reduce the voltage stress on each power device due to the utilization of multiple levels on the DC bus. Two switches of the same leg cannot be switched ON simultaneously which will lead to short circuit. It has very narrow output voltage range. Boosting of output is not possible. I.e the output side voltage is less than or equal to the input side voltage. Related gate drive is required for each switch. Provides second order filter, suppresses current and voltage ripples. Impedance source inverter has both inductor and capacitor in the dc link it provides constant high impedance voltage source. It provides impedance source coupling to the inverter on one port and DC source on other port. It’s a Transformer less network, so simple. A detailed literature survey was made on the proposed work [1-28]. Based on the diffeternt author works the study is made and identified the problems. The solution for the problems was focused to address the issue. The proposed work will be used for both static and dynamic loads applications like drives [7, 14]. 2.

IMPEDANCE SOURCE FED MULTILEVEL INVERTER Figure 1 displays the traditional inveter. The conventional system will give only less number of voltage levels with more amount of distortion in the supply [2]. Figure 2 shows the need for boosting the input voltage. Combination of passive elements is used to boost or buck the output. Figure 3 gives Journal homepage: http://iaescore.com/online/index.php/IJAAS

226



ISSN:2252-8814

the impedance network fed to the inveter circuit. The special type of switches used as bidirectional switches. Figure 4 displays the H-type FCMLI. Table 1 shows H-type flying capacitor multilevel inverter.

Figure 1. Traditional multilevel inverter

Figure 2. Impedance network fed multilevel inverter

Vdc

Figure 3. Z source fed multilevel inverter

Sa1

Sb1

Sa2

Sb2 R

a C4

Sa3

Sb3

Sa4

Sb4

Figure 4. H-type FCMLI

Table 1. H-type flying capacitor multilevel inverter - switch states and output voltage levels Sa1 0 0 0 1 1 0 0 1 1 1

Sa2 0 0 1 0 1 0 1 1 0 1

Sa3 1 1 0 1 0 1 0 0 1 0

Sa4 1 1 1 0 0 1 1 0 0 0

Sb1 1 0 1 1 1 0 0 0 0 0

Sb2 1 1 1 1 1 0 0 1 0 0

Sb3 0 0 0 0 0 1 1 0 1 1

Sb4 0 1 0 0 0 1 1 1 1 1

Vao -1/2 Vdc -1/2 Vdc 0 0 1/2 Vdc -1/2 Vdc 0 1/2 Vdc 0 1/2 Vdc

Vbo 1/2Vdc 0 1/2 Vdc -1/2 Vdc 1/2 Vdc -1/2 Vdc -1/2 Vdc 0 -1/2 Vdc -1/2 Vdc

Vab=VRN -Vdc -1/2 Vdc -1/2 Vdc 1/2 Vdc 0 0 -1/2 Vdc 1/2 Vdc 1/2 Vdc Vdc

MODULATION SCHEME The gate pulse is generated by comparing the reference signal with carrier frequency. The scheme developed based on CFD technic (Control Freedom Degree) [25]. Figure 5 shows the carrier arrangement of sinusoidal, third harmonic, 60 degree and stepped wave reference with phase disposition carrier.

Figure 5. Sample carrier arrangement of sinusoidal, third harmonic, 60 degree and stepped wave reference with phase disposition carrier Int. J. of Adv. in Appl. Sci.Vol. 8, No. 3, September 2019: 225 – 231

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



227

Z-SOURCE BASED MULITILEVEL INVERTER Figure 6 shows the basic view of impedance source based H type flying capacitor multilevel inverter. The choice of multilevel inverter depends on the applications. This combination of Z source and multilevel inverter provides unique features [27-28]. Z source multilevel inverter [15, 17-24] composed of DC source, Z network and single H bride inverter [26]. Figure 6 displays the power circuit of Z source based multilevel inverter. Voltage stress problem in conventional Z source inverter is overcome by Z source multilevel inverter. Voltage source and current source inverters can either buck or boost the voltage but Z source based multilevel inverter can able to both buck and boost the given voltage. Figure 6 shows the cascaded five level Z-source based multilevel inverter (ZSMLI). It has separate DC sources for each module with same voltage level. Presence of impedance network between DC source and main circuit overcome the limitations of traditional inverters.In traditional voltage source inverter the AC output voltage is below the DC input voltage and the dead time has to be introduced for both upper and lower devices which lead to distortion of output waveform. But the Z-source multilevel inverter can buck and boost the given input voltage.

Figure 6. Z-source based multilevel inverter 5.

SIMULATION RESULTS The simulated output voltage is shown for only one sample value of ma=1. The following parameter values are used for simulation: Vdc = 100V, R(load) = 10 ohms, C1 = C2 = C3 and C4 = 1000 e-3 Farad, fc =2000 Hz and fm = 50 Hz. Figures shows the sample five level output voltage generated by PDPWM [10,12] strategy with sine, THI, 60 degree and stepped wave reference and its FFT plot is shown in below Figures. Table 3 to Table 6 show the comparison of %THD, VRMS (fundamental), Vpeak and DC componenet for PDPWM strategies with various references. 5.1. Impedance source multilevel inverter with sine reference Figure 7 (a) and Figure 7 (b) represent the output voltage and harmonic spectrum of Z-source based H-type FCMLI for sinusoidal reference with PDPWM Strategy [16]. An output voltage and THD obtained are shown in Figure 7 (a).

A novel impedance source fed H-type flying capacitor multilevel inverter (C. R. Balamurugan)

228



ISSN:2252-8814

(a) Output voltage

(b) Harmonic spectrum

Figure 7. Impedance source multilevel inverter with sine reference 5.2. Impedance source multilevel inverter with THI reference Figure 8 (a) and Figure 8 (b) show the output voltage waveform and harmonic spectrum of Z-source based H-type FCMLI for THI reference [11, 13] with PDPWM Strategiy. An output voltage and THD are obtained that is shown in Figure 8 (a) and Figure 8 (b).

(a) Output voltage

(b) Harmonic spectrum

Figure 8. Impedance source multilevel inverter with THI reference 5.3. Impedance source multilevel inverter with 60 degree reference Figure 9 shows the output voltage and harmonic spectrum of Impedance Source multilevel Inverter with 60 degree reference. An output voltage and THD are obtained which shown in Figure 9 (a) and Figure 9 (b).

(a) Output voltage

(b) Harmonic spectrum

Figure 9. Sample FFT plot for impedance source multilevel inverter with 60 degree reference Int. J. of Adv. in Appl. Sci.Vol. 8, No. 3, September 2019: 225 – 231

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

229

5.4. Impedance source multilevel inverter with stepped wave reference Figure 10 displays the output voltage and harmonic spectrum of Impedance Source multilevel Inverter with stepped wave reference. An output voltage and THD is obtained which shown in Figure 10 (a) and Figure 10 (b).

(a) Output voltage

(b) Harmonic spectrum

Figure 10. Impedance source multilevel inverter with stepped wave reference In this analyse presence of RMS output voltage and THD in the output waveform is observed for various topologies by varying modulation index. The following tables represent measurements across multilevel inverter. Table 2 shows the simulation values. Table 3 and Table 4 show the measurement across MLI for L = 3mH, C = 4700µF and L = 3mH, C = 470µF. Table 5 and Table 6 represent the measurement across MLI for L = 250µH, C = 500µF and L = 160µH, C =1000µF. Table 2. Simulation values chosen Component Input Voltage Impedance Network

H-Type FCMLI AC Filter Load

Values Chosen 100 V L1 = 3mH L2= 3 mH C1 = 4700 micro farad C2 = 4700 micro farad Bus Capacitor (C1) = 1000e-6 Bus Capacitor (C2) = 1000e-6 MOSFET Clamping capacitor C1 = C2 = 100 e3 L = 3e-3 H C = 4700e-6 R = 10 ohm

Table 3. Output voltage THD for various modulation indicies Stepped wave reference 8.89 9.16 9.83 10.48

Modulation Type

Modulation Index

Sine Reference

THI Reference

60 degree Reference

Over Modulation Index (ma>1)

1.4 1.3 1.2 1.1

9.11 9.39 9.76 10.27

8.26 8.33 8.38 8.44

8.38 8.40 8.52 8.71

7.78

9.02

8.91

11.10

0.9 0.8 0.7 0.6

7.54 15.63 17.26 18.45

10.69 13.05 14.66 15.98

10.40 13.06 14.67 16.83

12.48 14.96 16.75 18.76

Normal Modulation Index (ma =1) Under Modulation Index (ma<1)

A novel impedance source fed H-type flying capacitor multilevel inverter (C. R. Balamurugan)

230



ISSN:2252-8814 Table 4. Vrms (fundamental) output voltage for various modulation indices Modulation Index 1.4 1.3 1.2 1.1

Modulation Type Over Modulation Index (ma>1)

Sine Reference

THI Reference

60 degree Reference

135.2 132.4 127.9 121

141.8 141.6 141.1 140.2

141.1 140.8 140 138.6

Stepped wave reference 135.8 133 125.7 118.3

Normal Modulation Index (ma =1)

87.96

132.5

136.9

111

Under Modulation Index (ma<1)

0.9 0.8 0.7 0.6

80.73 52.63 31.53 28.78

105.3 71.9 46.33 28.62

112.8 76.9 50.76 24.97

86.84 55.99 33.24 15.66

Table 5. Vpeakoutput voltage for various modulation indices Modulation Index 1.4 1.3 1.2 1.1

Modulation Type Over Modulation Index (ma>1)

Sine Reference

THI Reference

60 degree Reference

191.2 187.2 180.9 171.2

200.6 200.2 199.5 198.3

199.6 199.1 198 196.1

Stepped wave reference 192.1 188.1 177.7 167.2

Normal Modulation Index (ma =1)

124.4

187.4

193.6

156.9

Under Modulation Index (ma<1)

0.9 0.8 0.7 0.6

114.2 74.5 44.59 30.52

148.9 101.7 65.52 48.21

159.5 108.7 71.79 35.31

122.8 79.18 47.01 22.15

Table 6. DC componenet output voltage for various modulation indices Modulation Type Over Modulation Index (ma>1)

Modulation Index 1.4 1.3 1.2 1.1

Sine Reference

THI Reference

60 degree Reference

0.90 0.66 0.08 0.95

2.87 2.94 2.24 1.95

2.29 2.33 1.95 1.47

Stepped wave reference 1.57 0.97 0.79 2.10

Normal Modulation Index (ma =1)

2.37

1.65

1.19

2.80

Under Modulation Index (ma<1)

0.9 0.8 0.7 0.6

7.54 13.54 15.84 17.98

3.99 13.67 17.45 21.09

3.73 12.57 16.40 19.99

6.35 12.24 14.06 16.99

CONCLUSION The performance of Impedance source based flying capacitor multilevel Inverter with sinusoidal, third harmonic injection, 60 degree and stepped wave reference and PDPWM strategyare analysed in this work. Simulation was carried out for the proposed topologies using PWM technique. RMS values of output voltage and THD are observed using FFT analysis in MATLAB/Simulink environment. From the analysis of each proposed topology by varying impedance network values (L and C) are developed.From the simulation results it is found that for differenent references the THD is almost similar but the root mean square value in terms of voltage is more for THI, 60 degree PWM and stepped wave reference with phase disposition strategy. The results are observed for various values of ma (Amplitude Modulation Index) like under amplitude modulation index, normal amplitude modulation index and over amplitude modulation index. REFERENCES [1] Rodriguez, J.S. Lai, and Fang Z. Peng, "Multilevel Inverters: A Survey of Topologies, Controls and Applications," IEEE Trans on Industrial Electronics, vol. 49(4), pp. 724-738, 2002. [2] J. S. Lai and F. Z. Peng, "Multilevel converters – A new breed of power converters," IEEE Transactions Industry Applications, vol. 32, pp. 509-517, 1996. [3] Leon M. Tolbert and T.G. Habetler, "Novel multilevel inverter carrier-based PWM methods," IEEE IAS Annual meeting, pp. 1424-1431, 1998. [4] Li Li, Yaguang Liu, and Pragasen Pillay, "Multilevel selective harmonic elimination PWM technique in seriesconnected voltage inverters," IEEE Trans. on Industry Applications, vol. 36(1), pp. 160-170, 2002.

Int. J. of Adv. in Appl. Sci.Vol. 8, No. 3, September 2019: 225 – 231

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



231

[5] Brendan Peter McGrath and Donald Grahame Holmes, "Multicarrier PWM strategies for multilevel inverters," IEEE Transactions on Industrial Electronics, vol. 49(4), 2002. [6] P.C. Loh, D.G. Holmes, and T.A. Lipo, "Synchronization of distributed PWM cascaded multilevel inverters with minimal harmonic distortion and common mode voltage," Proc. IEEE PESC, pp. 177-182, 2003. [7] D. Krug, S. Bernet, and S. Saeed Fazel, "Comparison of 2.3-kV medium-voltage multilevel converters for industrial medium-voltage drives," IEEE transactions on Industrial Electronics, vol. 54(6), 2007. [8] M. Ghasem Hosseini Aghdam, S. Hamid Fathi, and B. Gharehpetian, "Harmonic optimization techniques in multiLevel voltage-Source inverter with unequal DC sources," Journal of Power Electronics, vol. 8(2), 2008. [9] Zhong Du, M. Tolbert, B. Ozpineci, and N. Chiasson, "Fundamental frequency switching strategies of a seven-level hybrid cascaded H-bridge multilevel inverter," IEEE Transactions on Power Electronics, vol. 24(1), 2009. [10] Palanivel P and S. Sekher, "Phase Shifted Carrier Pulse Width Modulation for Three Phase Multilevel Inverter to Minimize THD and Enhance Output Voltage Performance," Journal of Electrical Systems, vol. 6(2), pp. 1-13, 2010. [11] Bambang Sujanarko, "Advanced carrier based Pulse Width Modulation in asymmetric cascaded multilevel inverter," International Journal of Electrical & Computer Sciences, vol. 10(6), 2010. [12] Hussein A. Konber and Osama I. EL-Hamrawy, "Implementing a three phase nine-level cascaded multilevel inverter with low harmonics values," ProcsMEPCON’10, Cairo University, Egypt, December 19-21, 2010. [13] Faete Filho, Leon M. Tolbert, Yue Cao and Burak Ozpineci, "Real-time selective harmonic minimization for multilevel inverters connected to solar panels using artificial neural network angle generation," IEEE Transactions on Industry Applications, vol. 47(5), pp. 2117-2124, 2011. [14] Ashrafi B, Niroomand M, and Ashrafi Nia, "Novel reduced parts on-lineuninterruptible power supply," Procs IEEE 2012 International Power Engineering and Optimization Conference, pp. 252-257, 2012. [15] Yushan Liu, Baoming Ge, Haitham Abu-Rub, Fang ZhengPeng, "A modular multilevel space vector modulation for photovoltaic quasi-Z-Source cascade multilevel inverter," Proceedings of Twenty-Eighth Annual IEEE Conference on Applied Power Electronics APEC and Exposition, pp. 714-718, 2013. [16] Yushan Liu, Baoming Ge, Haitham Abu-Rub, Fang ZhengPeng, "Phase-shifted pulse-width-amplitude modulation for quasi-Z-source cascade multilevel inverter-based photovoltaic power system," IET Power Electronics, vol. 7(6), pp. 1444–1456, 2014. [17] B. Manjunatha, "Advanced Pulse Width Modulation Techniques for Z Source Multi Level Inverter," International Journal of Electrical, Computer, Energetic, Electronic and Communication Engineering, vol. 9(3), pp. 359-364, 2015. [18] M. Trabelsi, H. Abu-Rub, and BaomingGe,"1-MW quasi-Zsource based multilevel PV energy conversion system," Proceedings of IEEE International Conference on Industrial Technology (ICIT), pp. 224–229, 2016. [19] K. Vijayalakshmi and C.R. Balamurugan, "A Review on Z-Source Based Multilevel Inverter with Reduced Number of Switches," 2016 International Conference on Engineering and Technology (ICET). Karpagam College of Engineering, Coimbatore, December 16-17, 2016. [20] K. Vijayalakshmi and C. R. Balamurugan, "Investigations on Z-source based cascaded five level inverter," International Journal of Signal Processing, Image Processing and Pattern Recognition, vol. 9(12), pp. 37-50, 2016. [21] K. Vijayalakshmi and C. R. Balamurugan,"Simulation and Analysis of Improved Switched Inductor Quasi Z-Source Based Multilevel Inverter with Reduced Number of Switches," International Conference on Emerging Trends in Science, Engineering & Technology, Jerusalem College of Engineering, Chennai, March 18-19, 2017. [22] S.M. Revathi and C.R. Balamurugan, "A Review on Various Z-Source Fed Multilevel Inverter," 10th International Conference on Recent Innovations in Science, Engineering and Management, Dhruva Institute of Engineering and Technology, Nalgonda, July 7, 2017. [23] K. Vijayalakshmi and C.R. Balamurugan, "Z Source Multilevel Inverter Based on Embedded Controller," TELKOMIKA Indonesian Journal of Electrical Engineering and Computer Science, vol. 6(1), pp. 1-8, 2017. [24] C.R. Balamurugan and K. Vijayalakshmi, "Comparative Analysis of Various Z-source Based Five Level Cascaded H-bridge Multilevel Inverter," Bulletin of Electrical Engineering and Informatics (BEEI), vol. 7(1), pp. 1-14, 2018. [25] C.R. Balamurugan and R. Bensraj, "Analysis of Various Carriers Overlapping PWM Strategies for a Single Phase Ternary Multilevel Inverter," International Journal of Applied Power Engineering, vol. 7(1), pp. 27-39, 2018. [26] C.R. Balamurugan, S.P. Natarajan, V. Padmathilagam, and T.S. Anandhi, "Design of a New Three Phase Hybrid Hbridge and H-Type FCMLI for Various PWM Strategies," International Journal of Advances in Applied Science (IJAAS), vol. 4(3), pp. 82-89, 2015. [27] C.R. Balamurugan, S.P. Natarajan, and T.S. Anandhi, "Performance Evaluation of 3Φ Asymmetrical MLI with Reduced Switch Count," TELKOMIKA Indonesian Journal of Electrical Engineering and Computer Science, vol. 3(3), pp. 671-680, 2015. [28] C.R. Balamurugan, S.P. Natarajan, T.S. Anandhi, and R. Besnraj, "Hardware Implementation of Cascaded Hybrid Multilevel Inverter with Reduced Number of Switches," TELKOMIKA Indonesian Journal of Electrical Engineering and Computer Science, vol. 3(2), pp. 314-322, 2016.

A novel impedance source fed H-type flying capacitor multilevel inverter (C. R. Balamurugan)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 232~242 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp232-242



232

Performance analysis of security framework for software defined network architectures D. Arivudainambi, K. A. Varun Kumar Department of Mathematics, Anna University, India

Article Info

ABSTRACT

Article history:

Software defined data centers (SDDC) and software defined networking (SDN) are two emerging areas in the field of cloud data centers. SDN based centrally controlled services takes a global view of the entire cloud infrastructure between SDDC and SDN, whereas Network Function Virtualization (NFV) is widely used for providing virtual networking between host and Internet Service Providers (ISP’s). Some Application as a Service used in NFV data centers have a wide range in building security services like Virtual firewalls, Intrusion Detection System (IDS), load balancing, bandwidth allocation and management. In this paper, a novel security framework is proposed to combat SDDC and SDN based on NFV security features. The proposed framework consists of a Virtual firewall and an efficient bandwidth manager to handle multiple heterogeneous application requests from different ISPs. Real time data were taken from an experiment for a week and A new simulation based proof of concept is admitted in this paper for validation of the proposed framework which was deployed in real time SDNs using Mininet and POX controller.

Received Apr 29, 2019 Revised Jul 26, 2019 Accepted Aug 14, 2019 Keywords: Bandwidth Denial of service attack Software defined network Traffic inspection Virtualization

Corresponding Author: K. A. Varun Kumar, Department of Mathematics, Anna University, Chennai, Tamil Nadu, 600025, India. Email: kavaruncse@gmail.com

INTRODUCTION Present day network infrastructure is fully loaded with network devices such as router, switches etc. Most of these devices constitute dedicated hardware and require manual configuration. Such configurations can lead enterprises to failure during provisioning and de-provisioning. Change management is a time consuming task that requires halting of adequate resources to make a change in a single hardware (network devices). Advanced technologies such as Cloudification of things, Internet of Things etc., force the networking organization under high pressure to be more efficient. As noted, the use of traditional Application Specific Integrated Circuits (ASIC) does not have the scalability to handle these technologies. Switching between traditional hardware centric data network (ASIC) is very resilient and time consuming. Hence, for overcoming the above mentioned issues in the traditional ASIC based network hardware, a software defined controller specific network architecture called Software Defined Network and Network Virtualization has been designed. Software Defined Network is a dynamic updation framework which controls and manages network devices, network related services using high level languages and application program interfaces. SDN architecture is a centralized platform that can economize on the Information Technology (IT) infrastructural cost. Hence SDN supports different kinds of network component for better flexibility and agility. In traditional networking device, a packet that arrive at the switches has some proprietary rules that takes the packet forwarded to its assigned destination. Switches treat all the packets arriving in the same allotted way. SDN has certain rules defined programmatically in the controller for packet handling. Network Journal homepage: http://iaescore.com/online/index.php/IJAAS

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



233

administrator has complete control over the switches. A dynamic change of rules is possible where traditional network devices do not have provision for dynamic change management. SDN has an additional support for packet Prioritizing, deprioritizing and restriction which gives granular level of control to the administrators. This could help administrators taking care of the traffic related issues in the network and to manage the traffic loads in a flexible manner. The SDN Security vector architecture is divided in to four sections namely southbound, northbound, eastbound and westbound as shown in Figure 1. Southbound Section is popular for its interaction between controllers and switches. Open Flow protocol is a southbound interface to the controller representing the bridge which connects the controller and forward plane such as switches. SDN architecture has an open standard non - vendor protocol which allows all the vendors to use of an open architecture. Hypervisor plays the key role in the present day SDN architecture which gaps the control between the controller and the protocols both Southbound and Northbound. Northbound Section is the most critical API in the SDN environment. A majority of networking components exist in this section. Pyretic and frenetic is a SDN specific policy program language which communicates with controllers in the northbound section. All the security applications such as virtualized firewalls, intrusion detection system have a common API for interacting with the controller. Some hacking methods may compromise API’s which cause intentional risks to SDN because of their API’s communication policy. East and West Section: Management of SDN architecture is done by east and west bound sections. The management plane is controlled by the distributed architecture in which instructions through the controller for managing data. The distributed architecture has some important functionalities including control, management, monitor and task distribution for different low level instances. This section has also to support multiple controllers where they can share the same tasks and nodes.

Figure 1. Software defined network security attack vectors A new paradigm architecture is introduced in SDN for attaining a centralized management of the entire network that may also offered potential risk on network security problems like DDOS attack. Attacks attempt at making a service unavailable to their legitimate users by tiring the network resources. DDoS attack is witnessing a rapid growth in frequency despite various solution found by researchers. Some effort in network security research community has been made to give the appearance as the most effective for data in potential issues or opportunities to observe DDoS attacks with in the new enterprise network setting that adopts SDN. Implementation of SDN could provide defense against the DDoS attacks in many cases. But it may have certain challenges too (i) Slow packet forwarding (ii) Dynamic network topology. The data plane in SDN normally forwards packets on the basis of the policies assigned by the control programs. In traditional network, Packet forwarding is done by a piece of hardware. But, in contrast, it is done by software in the SDN. So there is a delay in the forwarding of packets due to traffic overhead and network delay in communication between control programs. Migration of virtual machines plays a vital role in SDN. The main role of virtualizing network is the allocation of the VM’s to their clients, which needs to updation of the entire network topology after the occurrence of every migration. So, it needs Dynamic network topology for its operational model. But it becomes highly vulnerable. DDoS attacks can be performed easily because, In SDN, separation of control plane and data plane turns out to be a major considering that data plane usually Performance analysis of security framework for software defined network architectures (D. Arivudainambi)

234



ISSN: 2252-8814

asks the control plane to obtain a flow rule. When the data plane receives a new packet, it does not normally know how to handle it. This is a key feature which helps attackers to strike the target in the sdn with flooding of request. This leads to DDoS attack. So, the performance of the sdn network and incoming traffic is analyzed to search to discover any abnormal behoviorial changes. The rest of this paper is organized as follows. Related works have been studied and compared with the work in Section II. Analysis of the effect of SDN on DDOS attack through motivation towards this work has been explained in section III. Section IV elaborates on the proposed security framework and its main components. A description of the operations of security framework is provided. In other words the experimental setup and evaluation of the framework in terms of scalability and performance is done. Section VI provides the conclusion. 2.

RELATED WORK Zhou et al [1] has proposed a method to detect compromised SDN devices by introducing Backup controllers to the network by recognizing the inconsistent behavior of Primary controllers and Switches when they are trustless. A backup controller plays an auditor role where the information of any state is recorded. Varadharajan et al [2] has constructed an architechture for securing end-to-end services across multiple SDN Domains by the policy language where security policies are used to control the flow of information. It also defines the path and flow based security policies which are significant for the services. It demonstrates intra and inter domain communications with multiple SDN Domains. Fawcett et al [3] has delivered the concept of virtualization in SDN for effective detection and protection. TENNISON a novel distributed SDN Security framework which is capable of monitoring, rapid detection and remediation and creates an insight into the multiple controllers of the network attack. Liu et al [4] has designed a SDN-based data transfer security model Middlebox-Guard (M-G) which reduces network latency and manages the dataflow of the network to enusure it run safely based on Intergral Linear Program (ILP) algorithm to tackle switch volume constraints. Kalkan et al [5] has described and evaluated joint entropy-based security scheme (JESS) to enhance security against DDos Attacks where it detects and mitigates these kind of hazards. Since it relies on a statistical model, it mitigates not only known attacks but also unfamiliar attacks with efficient manner. Bing et.al. [6] have presented a DDoS attack mitigation system for both cloud and SDN computing where attack detection is done with the help of high programmable network monitoring. They also have proposed a graphical model that deals with the data shift problem for a new network paradigm. Shin et.al. [7] have provided a new technique to fingerprint the SDN network and launch the DDoS attack using the resources of the entire network or part thereof in the SDN. They have introduced the SDN scanner and feasible defense mechanism techniques. Chun et.al. [8] Presented an attack mitigation system for virtual machines. First, they started with the generation of an attack graph analytical model to enable counter measure selection and vulnerability detection. Their proposed frameworks significantly improve the effectiveness of attack detection. Kim et.al. [9] have focus on issues in network management. They have identified three major problems, namely dynamic network changes, support providing for network configurations in high level languages, control over performing network diagnosis and troubleshooting. Their prototype working in SDHN and Campus network states points out to the numerous advantages over deploying a SDN rather than the traditional network. Wang Yang et.al [10] have presented an Open Scass: An open chain as a service platform, in which they have integrated the Software Defined Network and Network function virtualization for the enforcement of a service chaining policy, which enables the improvement of the scalability, auto provisioning by integrating SDN and NFV. Banikazemi et.al. [11] designed a meridian framework which supports dynamic updates to virtual network and organization of different task on a large number of devices, irrespective of whether they concentrate on automatic controller of network devices through application Interface with the help of SDN. It can be integrated with many cloud controllers. Wang et.al. [12] have applied SDN technology to the cloud center to propose a dynamic load balancing of the cloud center. SDN monitors the network traffic, task scheduling and balances the load condition of the network over a long period of time and increases the throughput. Richard et.al. [13] Proposed a Unified Server Resource Management for Information and Communication technology (ICT), It is a Dynamic method for allocation of VM’s to cloud providers and their clients. This method can reduce cost by migrating 50% of virtual machines. Yan et.al. [14] have discussed the impact of DDoS attacks over SDN in cloud computing. They have discussed the implications of features in SDN like centralized controller, dynamic rule updation, software controlled traffic analyzer that play a vital role in defence against DDoS attacks in SDN. Munoz et.al. [15] present orchestration of both SDN and NFV for abstraction of virtualizing the tenant network. On the one hand, SDN slices the physical network into multiples of virtual network and on the other, NFV drives all hardware devices to virtualize. Then the virtualization of SDN controllers update their tenant instances to work independently with Individual controller that connects within a second. Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 232 – 242

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



235

Li et.al. [16] have discussed a controller minimization problem in SDN. They state that distributed controller allocation is more efficient than a general controller that is allocated to each VM. They have proved the presence of better security for VM’s through a distributed controller compared to a general controller. Izzat et.al. [17] have provided a widespread study on SDN security. They discuss security threats and their effects in SDN i.e., Ip Spoofing, Data Tampering, Refutation, Data Leakage, Denial of Service attacks, and Privilege Escalation. The authors have also surveyed different SDN security controls, such as intrusion detection systems/intrusion prevention system, firewalls, access control, network assessment, and policy management. They part out to several pathways of SDN evolving. Fei et.al. [18] made a complete study of the significant areas in SDN/OpenFlow operation, including the basic concepts, applications, virtualization, controller, language abstraction, security, Quality of Service (QoS), as well as its incorporation into wireless and optical networks. They discuss the advantages and disadvantages of different schemes, and discuss the future research developments in this area. They have expressed their optimism of the survey assistance to both industry and academia and Research and Development (R&D) people to realize the latest development of SDN/OpenFlow designs. Shin et.al. [19] have presented different vulnerabilities of SDN. The attackers could mount an inundation attack on the SDN controller with the goal of weakening the entire SDN. Therefore, the controller must have a suitable architecture to survive such an attack while the current operation is in progress. Scheduling based architecture is designed for the SDN controller that hints to effective attack quarantine and network defense during denial of service attacks. 3.

MOTIVATION SDN is divided into three main efficient layers, namely organization layer, control layer, and application layer, as shown in Figure 1. The possibility of deadly DDoS attacks thrown on these three layers of architecture does exist. Based on the potential targets, the DDoS attacks thrown on SDN can be categorized, namely control layer DDoS attacks, application layer DDoS attacks and infrastructure layer DDoS attacks. 3.1. Control layer DDoS attacks These controllers could possibly be seen as Single Point of Failure hazards for the network. So, they are a predominantly striking target for DDoS attack in the SDN design. The following approaches can launch flooding of control plane DDoS attacks: which can affect all the API’s. In a real time scenario, many contradictory flow rules from different requests may cause DDoS attacks on the control layer. Inside the operation of SDN, data plane normally make a request to control plane to follow flow rules when the data plane gets new network packets which is not known to the controller that how to handle. There are two possibilities for the managing a new flow when flow match does not exist in the flow table: either the entire packet or a portion of the packet header is communicated to the controller to give solution to the query [20]. With a huge capacity of network traffic, transmitting the complete packet to the controller would lodge high bandwidth [21]. 3.2. Application layer DDoS attacks DDoS attacks in application layer are of two types, first one is to attack some application within the SDN and (b) to attack north bound API. There is no clear separation of API’s and resources used in the SDN [22]. Hence DDoS attacks launched over the particular application in SDN can affect all other applications in SDN. 3.3. Infrastructure layer DDoS attacks There are two possible types of data plane DDoS attacks. The first is to attack some switches, once the switches get compromised, automatically all the resources in the SDN architecture become vulnerable. The other type is to attack south bound. For example, the packets must be stored in node memory and only header information of packets is transmitted to the controller until the flow table entry is returned. In this scenario, the attacker has an option to execute a DDoS attack on the node by recognizing the number of new and unknown flows. To reveal the viability of DDoS attacks, a scanning tool named SDN scanner has been introduced in to clearly exhibit the network that set up SDN. This technique can be easily performed by adjusting current network scanning tools (e.g., ICMP scanning and TCP SYN scanning). The attack can be directed to SDN network by a distant attacker, and can suggestively reduce the performance of an SDN network without needful high performance or high capacity devices [23]. One clear case of DDoS that can directly disturb the controller is flooding the controller with header information packets (new packets). Mostly a new packet that Performance analysis of security framework for software defined network architectures (D. Arivudainambi)

236



ISSN: 2252-8814

does not have the match in the flow table would be sent to the controller for processing. Most DDoS attacks use unidentified source addresses obtained through the spoofing technique, which interprets new incoming packets at the switch. There is also main drawback seen when the number of new incoming packets is larger than the secure channel’s bandwidth and reduces the controller’s handling power. In DDoS attacks, flooding of packets are directed to a host or a group of hosts in a network. If the source addresses of the incoming packets are unidentified, the switch does not find a match and has to advance the packet to the controller. Hence for addressing the problem referred to above efficient Software defined Intrusion Detection System is desirable. 4.

PROPOSED FRAMEWORK FOR COMBATTING DDOS IN SDN In general, data plane plays the role of handling and forwarding data packets to destination & Control plane is a key player in handling the logical functionalities of router firmware such as load balancing, virtualization and firewall. This design in firmware decelerates the performance of hardware and leads to a major time consuming process for solving the problem referred to above, horizontal integration called Software Defined Network has been introduced. The main property of SDN is a controller based dedicated software architecture which separates the control plane from the data plane. However, this central software management is highly prone to cyber-attacks due to its single umbrella property of control and handles components. Here, a novel security framework stated in Figure 4 is proposed to provide security by combining virtual firewall and bandwidth analyzer & Ford-Fulkerson algorithm [24] is applied to obtain the maximum flow between two access points in the network. The algorithm is as follows. Let G be a Graph with vertex and Edges G (V,E) and for each edge from each vertex, the flow be f(u,v) and capacity of the network be c(u,v). a. Capacity The flow along an edge cannot exceed its capacity as stated ƒ (u, v) ≤ c (u, v), (u, v) ϵ E b.

Skew symmetry The net flow from u to v must be the opposite of the net flow from v to u (u,v) ϵ E ƒ(u,v) = - ƒ(u,v)

Flow conservation The net flow to a node is zero, except for the source, which "produces" flow, and the sink, which "consumes" flow. u ϵ V : u ≠ s and u ≠ t => ∑ d.

∈

ƒ(𝑢, 𝑤) = 0

Value The flow leaving from s must be equal to the flow arriving at t ∑(

, )

ƒ(𝑠, 𝑢) = ∑(

, )∈

ƒ(𝑣, 𝑡)

Inputs Given a Network G = (V, E) with flow capacity c,a source node s,and a sink node t Output Compute a flow ∫ from s to t of maximum value ƒ (u,v) ← 0 for all edges (u,v) While there is a path p from s to t in Gf , such that cf(u,v) > 0 for all edges (u,v) Ꜫ p: Find cf (p) = min { cf(u,v):(u,v) Ꜫ p} For each edge (u,v) Ꜫ p ƒ (u.v)←∫(u,v) + cf(p) (send flow along the path) ƒ (u,v)←∫(u,v) – cf(p)(The flow might be returned later) For each pair (u,v) access points the path which gives the maximum flow has a entropy tag E u,v Define E = MAX E u,v (u, v) Ꜫ p Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 232 – 242

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



237

Fixing a threshold 0.5, if E > 0.5 then the corresponding path is suspicious. The algorithm is applied to maintain each flow within the network. The below mentioned tuple table is maintained after each of algorithm execution. 4.1. Virtual firewall Controller is one among the control components responsible for transforming Open flow switch (high level) logics into IP table rules. Firewall as shown in Figure 2 implemented in the security framework is a stateful firewall where it monitors all the active connections and performs both ingress and egress filtering on the basis of the rules defined in IP tables. The deployed firewall stated in Figure 3 can distinguish between the data packets coming from legitimate source or not. A sample set can be stated as follows Rule set1. ufw allow protocol tcp from IPTable_allow_list to any port 8008. 4.2. Bandwidth analyzer Attacks performed on top of layer 3 and layer 4 focus on bandwidth and resource utilization which leads to compromise in performance and security. Most of the attacks launched over are volumetric and stealthier target specific. Hence an optimized bandwidth analyzer is required for monitoring the bandwidth periodically or intervals. The bandwidth analyzer consists of a significant module called Traffic Inspection Model (TIM). TIM helps incisive inspection of the packet and extracts possible information from the packets.

Figure 2. Proposed virtual firewall architecture

Figure 3. Firewall operation based on rule set

Performance analysis of security framework for software defined network architectures (D. Arivudainambi)

238



ISSN: 2252-8814

Figure 4. Architecture diagram of proposed SDN security framework ALGORITHM Controller Define ports. Define host policies (host name, pid.uid); For each Switches S in network N do Monitor-enable==Controller 1||Controller 2 If traffic –generating host==INTRA_HOST Monitor-enable==Controller 1 Else if Traffic –generating host==INTER_HOST Monitor-enable==Controller 2 For each host do Inspect-packet==stateful host.Traffic.match (true); host.Uid.match (true); host.Uid.match (true); end end end Detection For each host in the network do For each Switch in the network do Packet.thershold==50 Packet.host==host.getUID, get NID Packet. Network==get. Network by Switch (SID) If packet. Generating! =host.getUID,get(host ID) flag (“Malicious host”) elseIf (Packet.host exceeds threshold) flag (“DDOS possible”) flag(host.getUID()) end end end 5.

EXPERIMENTAL SETUP The proposed security framework is tested and implemented in real time test bed. The proposed framework is developed with the complete use of python and deployed in a parrot operating system. SDN configuration as stated in Table 1 is implemented in the parrot Operating System (OS) using Mininet.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 232 – 242

Int. J. of Adv. in Appl. Sci.



ISSN: 2252-8814

239

Table 1. Configuration of host, controller and link state S.No

Host

1 2 3 4 5 6 7 8

AAh1 AAh2 ABh1 ABh2 BAh1 BAh2 BBh1 BBh2

Host address 10.1.1.1 10.1.1.2 10.1.2.1 10.1.2.2 10.10.10.1 10.10.10.2 10.10.20.1 10.10.20.2

Host MAC 0A:0A:00:00:00:01 0A:0A:00:00:00:02 0A:0B:00:00:00:01 0A:0B:00:00:00:02 0A:0B:0A:00:00:01 0A:0B:0A:00:00:02 0A:0B:0B:00:00:01 0A:0B:0B:00:00:02

Link state sAA sAA sAB sAB sBA sBA sBB sBB

Controller1 Ryu cA cA cA cA cA cA cA cA

Port 6633 6633 6633 6633 6633 6633 6633 6633

Controller2 Ryu cB cB cB cB cB cB cB cB

Port 6634 6634 6634 6634 6634 6634 6634 6634

5.1. Experimentation Initially Mininet was executed to get virtual network with virtual hosts, virtual switches, virtual controllers, virtual links etc. The controller with proposed security framework was developed in python and deployed as individual component. A new topology with ovsf switch setup was created with limited number of switches and its associated hosts and links. Mininet was executed using custom topology with parameters. Open flow switch support was always enabled and ryu controller component was used for managing and handling the switch. A clear view on basic SDN experimental configuration is stated in Table 2. For example S is the switch, source host is represented as BBh1, Destination host as ABh1, attack type as SNMP, protocol as TCP. Packet captured in normal scenario is 6435 and the attack scenario is 1074. 5.2. Packet capture TCP dump utility is installed and all the data packets moving in and around all interfaces were tapped and recorded into .pcap file [25]. These captured data helped extraction of the packets that deviated from the normal packets. The captured packets along with statistics (Normal and attack) are detailed in Table 2. Switch S22 was monitored for validating the experiments and they were considered for the attack scenario. Figure 5 shows the packet flow rate within the switch S22. The arrival of packet rate was estimated for every 15 minutes interval. In Figure 4, dark lines clearly show that number of requests is very high while comparing normal and attack scenario.

Figure 5. Packet flow rate within the switch S22

Performance analysis of security framework for software defined network architectures (D. Arivudainambi)



240

ISSN: 2252-8814 Table 2. Switch, host configuration, protocol and attack type of experimental setup

S.No 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

Number of Switch S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22 S1,S2,S11,S12,S21,S22

Source Host

Destination Host

BBh1 BBh1 BAh1 BAh1 ABh1 ABh1 ABh2 ABh2 BBh1 BBh1 BAh1 BAh1 ABh1 ABh1 ABh2 ABh2 BBh1 BBh1 BAh1 BAh1 ABh1 ABh1 ABh2 ABh2 BBh1 BBh1 BAh1 BAh1 ABh1 ABh1 ABh2 ABh2

AAh1 AAh2 AAh1 AAh2 AAh1 AAh2 AAh1 AAh2 ABh1 ABh2 ABh1 ABh2 ABh1 ABh2 ABh1 ABh2 BAh1 BAh2 BAh1 BAh2 BAh1 BAh2 BAh1 BAh2 BBh1 BBh2 BBh1 BBh2 BBh1 BBh2 BBh1 BBh2

Attack Type

Protocol

DDoS SNMP Land Flood UDP flood Hping3 POD Smurf DR-DoS SNMP Land Flood UDP flood Hping3 POD Smurf DR-DoS SNMP Land Flood UDP flood Hping3 POD Smurf DR-DoS SNMP Land Flood UDP flood Hping3 POD Smurf DR-DoS

TCP TCP Openflow Openflow Openflow Openflow TCP TCP TCP TCP Openflow Openflow Openflow Openflow TCP TCP TCP TCP Openflow Openflow Openflow Openflow TCP TCP TCP TCP Openflow Openflow Openflow Openflow TCP TCP

Packet captured Normal Scenario 6435 6435 17000 17000 17000 17000 6435 6435 6435 6435 17000 17000 17000 17000 6435 6435 6435 6435 17000 17000 17000 17000 6435 6435 6435 6435 17000 17000 17000 17000 6435 6435

Attack Scenario 1076 1076 15002 15002 15002 15002 1076 1076 1076 1076 15002 15002 15002 15002 1076 1076 1076 1076 15002 15002 15002 15002 1076 1076 1076 1076 15002 15002 15002 15002 1076 1076

5.3. Attack scenario Once the topology was created using mininet, interfaces other than eth0 were monitored and recorded. Interfaces other than eth0 and interface start S1-eth1....Sn-ethn belongs to software defined switches and hosts associated with the switches and its links were recorded using tcpdump utility. Any interface which was up in the control plane was taken with unlimited requests to send data packets to an individual switch or its associated hosts. This led to the performance of DDoS over a specific host within the connected switch. From Figure 6, red spline which ranges from 15 – 105 packets per second states the general packet flow handled by the controller from switch S22. Blue explains the packet flow during DDoS detection which in turn clearly shows that the flow of packet rate is completely reduced and falls between 10-15 packets per second. 5.4. Firewall The controller component consists of a major module firewall. Firewall is used for recording the internal host address (permanent and logical). Rule set is defined to the firewall for inspecting and detecting malicious traffic. Rule set is based on the behavior features of hosts and networks.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 232 – 242

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



241

Figure 6. Detection rate vs. packet flow rate within switches 5.5. Training phase The Experimentation is carried out and data transfer, routing, packet forwarding etc were performed within the hosts. This experimentation was carried out for a week with balanced payload and transfer. The recorded traffic was taken into consideration and processed for extracting the possible features from individual interfaces, connected switches, associated hosts, links. These extracted features were tabulated into a system with network level features and used for further classification. A final entropy was calculated and links with defined IP address and MAC address were registered in the IP tables (firewall rule table). 5.6. Testing phase In testing phase, the attack module written in python was executed for creating suspicious traffic from other interfaces. ie. Host H1 associated to Switch S2 created the attack payload over H2 host associated with switch S3. Once the attack module got executed, all the traffic moving in and around the interfaces was tapped and features were extracted. Based on the extracted features, the entropy was calculated for the individual switch, if the entropy exceeded, the system flagged the traffic as suspicious and the IP address and MAC address were registered automatically in the IP tables for denial. 6.

CONCLUSION This paper has proposed a novel security framework to combat DDoS using a virtual firewall and a bandwidth analyzer. The model presented shows attacks launched over SDN as highly sophisticated and hard in terms of scalability. The proposed framework is highly reliable in detecting the suspicious traffic and handling that traffic within the medium without compromising on its performance. The proposed framework works fine for fixed topology. In future the work can be enhanced to fit changing topology. Some of the key concerns like topology changing firewall rules, automating rule updation etc. can be taken into consideration and a large volume of work spent on the above addressed issues. REFERENCES [1] [2] [3] [4] [5] [6] [7]

Haifeng Zhou, Chunming Wu, Chengyu Yang, Pengfei Wang, Qi Yang, Zhouhao Lu, and Qiumei Cheng, “SDNRDCD: A Real-Time and Reliable Method for Detecting Compromised SDN Devices,” 2018 IEEE/ACM transactions on networking, Vol. 26, No. 5, 2018. Vijay Varadharajan, Kallol Karmakar, Uday Tupakula, and Michael Hitchens “A Policy-Based Security Architecture for Software-Defined Networks,” 2019 IEEE Transactions on information forensics and security, Vol. 14, No. 4, 2019. Lyndon Fawcett, Sandra Scott-Hayward, Matthew Broadbent, Andrew Wright, and Nicholas Race, “TENNISON: A Distributed SDN Framework for Scalable Network Security,” 2018 IEEE Journal on selected areas in communications, Vol. 36, No. 12, 2018. Yanbing Liu, Yao Kuang, Yunpeng Xiao, and Guangxia Xu “SDN-Based Data Transfer Security for Internet of Things,” 2018 IEEE Internet of things journal, Vol. 5, No. 1, 2018. Kübra Kalkan, Levent Altay, Gürkan Gür, and Fatih Alagöz “JESS: Joint Entropy-Based DDoS Defense Scheme in SDN,” 2018 IEEE Journal on selected areas in communications, Vol. 36, No. 10, 2018. Bing wang, Yoa Zeng, Wenjing liu, and Thomas Hou “DDoS Attack Protection in the Era of Cloud Computing and Software-Defined Networking,” 2014 IEEE 22nd International Conference on Network Protocols, Computer Networks 81, pp. 308–319, 2015. S. Shin and G. Gu, “Attacking software-defined networks: a first feasibility study,” Proc. the Second ACM SIGCOMM Workshop on Hot Topics in Software Defined Networking, pp. 165–166, 2013.

Performance analysis of security framework for software defined network architectures (D. Arivudainambi)

242 [8] [9] [10] [11] [12] [13] [14] [15]

[16] [17] [18] [19] [20] [21] [22] [23] [24] [25]



ISSN: 2252-8814

Ying-Dar Lin, Po-Ching Lin, Chih-Hung Yeh, Yao-Chun Wang, and Yuan-Cheng Lai “An extended SDN architecture for network function virtualization with a case study on intrusion prevention,” IEEE Network, Vol. 29, No. 3, pp. 48-53, 2015. Sungheon Lim, Seungnam Yang, Younghwa Kim, Sunhee Yang, and Hyogon Kim “Controller scheduling for continued SDN operation under DDoS attacks,” Electronics Letters, Vol. 51, No. 16, pp. 1251–1261, 2015. Wanfu Ding, Wen Qi, Jianping Wang, and Biao Chen, “OpenSCaaS: an open service chain as a service platform toward the integration of SDN and NFV,” IEEE Network, Vol. 29, No 3, pp. 30-35, 2015. Mohammad Banikazemi, David Olshefski, Anees Shaikh, John Tracey, and Guohui Wang, “Meridian: an SDN platform for cloud network services,” IEEE Communications Magazine, Vol. 51, No. 2, pp. 120–127, 2013. Wang Yon, Tao Xiaoling, He Qian, and Kuang Yuwen, “A dynamic load balancing method of cloud-center based on SDN,” China Communications, Vol. 13, No. 2, pp. 130-137, 2016. Richard Cziva, Simon Jouët, David Stapleton, Fung Po Tso, and Dimitrios P. Pezaros “SDN-Based Virtual Machine Management for Cloud Data Centers,” IEEE Transactions on Network and Service Management, Vol. 13, No. 2, pp. 212–225, 2016. Qiao Yan, F. Richard Yu, Qingxiang Gong, and Jianqiang Li, “Software-Defined Networking (SDN)and Distributed Denial of Service (DDoS) Attacks in Cloud Computing Environments: A Survey, Some Research Issues, and Challenges,” IEEE Communications Surveys & Tutorials, Vol. 18(1), pp. 602-622, 2016. Raul Munoz, Ricard Vilalta, Ramon Casellas, Ricardo Martinez, Thomas Szyrkowiec, Achim Autenrieth, Víctor Lopez, and Diego Lopez “Integrated SDN/NFV management and orchestration architecture for dynamic deployment of virtual SDN control instances for virtual tenant networks,” IEEE/OSA Journal of Optical Communications and Networking, Vol. 7, No. 11, pp. 62-70, 2015. He Li, Peng Li, Song Guo, and Amiya Nayak, “Byzantine-Resilient Secure Software-Defined Networks with Multiple Controllers in Cloud,” IEEE Transactions on C+loud Computing, Vol. 2, No. 4, pp. 436-447, 2014. Izzat Alsmadi and Dianxiang Xu, “Security of Software Defined Networks: A Survey,” Computers and security, Elsevier, Vol. 53, pp. 79-108, 2015. Fei Hu, Qi Hao, and Ke Boa, “A Survey on Software-Defined Network and OpenFlow: From Concept to Implementation,” IEEE Communications Surveys & Tutorials, Vol. 16, No. 4, pp. 1553-877X, 2014. Seungwon Shin and Guofei Gu, “Attacking Software-Defined Networks: A First Feasibility Study,” Proceedings of the second ACM SIGCOMM workshop on Hot topics in software defined networking, pp. 165-166, 2013. Mayur Channegowda, Reza Nejabati, and Dimitra Simeonidou, “Software-defined optical networks technology and infrastructure: Enabling software-defined optical network operations,” IEEE/OSA Journal of Optical Communications and Networking, Vol. 5, No. 10, pp. A274–A282, 2013. C. Dixon, D. Olshefski, V. Jain, C. DeCusatis, W. Felter, J. Carter, M. Banikazemi, V. Mann, J. M. Tracey, and R. Recio, “Software defined networking to support the software defined environment,” IBM Journal of Research and Development, Vol. 58, No. 2/3, pp. 3:1-3:14, 2014. Zilong Ye, Xiaojun Cao, Jianping Wang, Hongfang Yu, and Chunming Qiao, “Joint topology design and mapping of service function chains for efficient, scalable, and reliable network functions virtualization” IEEE Network, Vol. 30, No. 3, pp. 81-87, 2016. Shibo Luo, Jun Wu, Jianwua Li, and Longhua Gua, “A Multi-stage Attack Mitigation Mechanism for Softwaredefined Home Networks,” IEEE Transactions on Consumer Electronics, Vol. 62, No. 2, pp. 200–207, 2016. Wikipedia, [Online]. Available: https://en.wikipedia.org/wiki/Ford%E2%80%93Fulkerson_algorithm Haifeng Zhou, Chunming Wu, Ming Jiang, Boyang Zhou, Wen Gao, Tingting Pan, and Min Huang, “Evolving Defense Mechanism for Future Network Security,” IEEE Communications Magazine, Vol. 53, No. 4, 2014.

BIOGRAPHIES OF AUTHORS Arivudainambi D is currently an Associate professor in Department of Mathematics, Anna University, Chennai, Tamilnadu. He received his Post-Doctoral in University of Toronto in 2004. He received his Ph.D. degree from Anna University, in 2002 and His research interest includes Computer Networks, Queuing theory, Stochastic Processes and its applications, Operations Research, Cloud Computing, Wireless Sensor Networks, Evolutionary Algorithms, Adhoc Networks.

Varunkumar K.A. is currently pursuing his Ph.D. in Anna University, Chennai. His research interest includes Network Security, Malware Analysis, Cloud Security, Software Defined Network Security and Image Processing.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 232 – 242

International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 3, September 2019, pp. 243~250 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i3.pp243-250



243

Compact planar ultrawideband MIMO antenna for wireless applications P. Pavithra, A. Sriram, K. Kalimuthu Department of Electronics and Communication Engineering, SRM Institute of Science and Technology, India

Article Info

ABSTRACT

Article history:

A compact microstrip fed printed monopole MIMO antenna with ultrawideband (UWB) frequency response (S11< -10 dB for 3.1-10.6 GHz) is proposed in this paper. The proposed antenna is miniaturized and has a high isolation of > 23 dB between the ports compared to the existing UWB multiinput multi output (MIMO) antennas in the literature. The proposed antenna is built on FR4 substrate with thickness of 1.6 mm using all-digital single chip architecture and it is planar in geometry to be easily integrated with the other electronic components in the printed circuit board (PCB). The UWBMIMO antenna is analyzed using simulation and measurements and its performance is investigated. The antenna is extremely useful for low power short range communications and it provides high multipath immunity due to diversity.

Received Feb 22, 2019 Revised May 17, 2019 Accepted Jun 6, 2019 Keywords: MIMO antenna Mutual coupling Printed monopole antenna Ultrawideband (UWB)

Corresponding Author: P. Pavithra, Department of Electronics and Communication Engineering, SRM Institute of Science and Technology, Chennai-603203, India. Email: pavi.feb24@gmail.com

INTRODUCTION Ultra-wideband (UWB) technologies have widely drawn considerable attention due to the several advantages for communications and sensing applications due to the properties of low-power consumption, high data rate, robustness to the multipath environment, relatively low complexity and high time-domain resolution. UWB antennas cover the frequency range of 3.1 GHz to 10.6 GHz which is mainly assigned for the UWB indoor communication applications. Currently, there is a huge demand in increasing the channel capacity. Hence, the two or more, multiple antennas in a single terminal can provide higher data rate which increases the channel capacity without sacrificing the requirement of additional power and frequency bands. MIMO, Multiple-Input Multiple-Output systems are enriched withmultiple antennas, which can be used as both in transmitter as well as in receiver mode. These multiple antennas have the advantages of higher data rate which in turn enhances the channelcapacity, greater reliability of the system without any extra power or bandwidth. Since, multiple antennas occupy the same arrayof antenna structure, less spacing would be placed between them, and hence mutual coupling isincreased. Hence, high port isolation is recommended to reduce the losses. The combination of UWB technology with multi-input multi output (MIMO) is proposed in this paper. This system increases the channel capacity for several users to access several services at the same time. It also overcomes the disadvantage of multipath fading in which conventional UWB technologies are facing. This paper gives an effective solution for the limitation of short range communication which require low power consumption devices. But in MIMO technology, several factors of antenna are to be considered such as size, isolation and gain etc. Several techniques have been introduced such as slits [1] to enhance the impedance matching. A2 × 1 UWB MIMO antenna is designed in [2] with only limited bandwidth (3–6 GHz) which does not meet Journal homepage: http://iaescore.com/online/index.php/IJAAS

244



ISSN: 2252-8814

the FCC specifications and there are many parameters that influence the isolation and VSWR. Whereas, some of the antennas have complex geometry [3] to achieve UWB response. In order to reduce mutual coupling between MIMO antennas ,many configurations and structures have been implemented.Protruded stubs with two element MIMO are introducedin [4] to reduce the wideband mutual coupling between the radiators, the space is not used effectivelywhich results in larger size of antenna.Although many neutralization lines are implemented in [5] to reduce the wideband mutual coupling between the radiators, theisolaton reported is veryless at higher frequencies.The UWB band is covered thoughwrench shaped structures with slots [6]. In [7], a diversity MIMO antenna has been desiged in which protrudedstubs that are employed for the port isolation between the antenna elements to cover theentire UWB spectrum. In [8], a 2–port compact MIMO antenna using asymmetric coplanarstrip feeding configuration has been design with the structure of slots between the monopole andthe ground plane for the achievement of high port isolation of 20 dB. The Isolation of < -15 dB hasbeen achievced by designing defected ground plane and the introduction of the slits in the patchof the antenna as in [9]. In [10], isolation is not very high even without decoupling structures. Though various slots have been introduced in [11], lower UWB range is not covered. A highly compact single element antenna is designed for UWB applications in [12] which cannot be used for MIMO applications. MIMO antennas in [13, 14] occupy a larger board space to radiate a particular frequency range of 2.41-2.46 GHz and 2.2-2.9 GHz respectively. In summary, the size of the existing antennas is vey large when compared with proposed fourelement MIMO antenna. This paper addresses the issues of size, complexity and isolationof UWB MIMO antenna. 2. ATENNA DESIGN 2.1. Conventional single element UWB antenna design The geometry of a single element UWB antenna and detailed design parameters are shown in Figure 1 and Table 1 respectively. The proposed monopole antenna is printed on the FR4 - epoxy substrate with the thickness of 1.6 mm and dielectric loss tangent of 0.02. The total volume of the antenna is 40 × 40 × 1.6 mm3. A circular radiating patch is etched on the front side of the substrate and a partial ground plane is present on the back side.

(a)

(b)

Figure 1. (a) Top layer of single element antenna and (b) Bottom layer of single element antenna (all dimensions are in mm) Table 1. Dimensions of the single element UWB monopole antenna Sl. No. 1 2 3 4

Structure

Material

Substrate Patch Feedline Ground

FR4 – epoxy Copper Copper Copper

X 40 5.5(radius) 8.2 16.5

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 243 – 250

Size (mm) Y 40 3 6.7

Z 1.6 0.035 0.035 0.035

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



245

2.2. Four element UWB MIMO antenna Figure 2 depicts the geometry of the four element UWB MIMO antenna. The microstrip fed antenna is matched at 50Ω impedance which gives radiation with high reliability and impedance matching. The principle antenna element is considered as a circular quarter wave monopole with the edge to feed gap distance of 1.5 mm. This feed gap determines the impedance bandwidth of the antenna. The antenna elements are spatially constructed with high intrinsic isolation without any additional filtering requirements and that allows easy extension of number of elements in an array. The absence of decoupling circuit or isolating structures gives the compact size of 40 × 40 × 1.6 mm3. This is the most compact four element MIMO antenna realized till date.

Figure 2. Four element UWB MIMO antenna (all dimensions are in mm) 3. RESULTS AND DISCUSSION 3.1. Return loss Initially, single element UWB antenna is simulated and its performance in terms of return loss andgain are analyzed. Later, four element UWB MIMO antenna is implemented and the antenna’s performances are re-analyzed. The simulated results of the antenna is illustrated by the excitation given at port 1 while the other ports are matched with 50 Ω load. The proposed antenna has < -10 dB impedance bandwidth over the operating band. The simulated result of return loss plot of single element UWB antenna and four element UWB MIMO antenna is compared in Figure 3.

Figure 3. Simulated plot of return loss Compact planar ultrawideband MIMO antenna for wireless applications (P. Pavithra)

246



ISSN: 2252-8814

3.2. Isolation The main challenge in construction of UWB MIMO technology is to increase the number of elements in antenna array with high isolation and without introduction of any decoupling circuits. The isolation of the proposed antenna isgreater than 23 dB as observed over UWB band. 3.3. Fabricated antenna The proposed MIMO antenna is fabricated on a low cost FR4 substrate with the thickness of 1.6 mm which is shown in Figure 4. The fabricated antenna is testedforreturn loss and isolation parameters using Rohde and Schwarz vector network analyzer as shown in Figure 5. During measurement, one port is excited and other ports are matched with 50-Ω load impedance. Figure 6, depicts the comparison plot of simulated and measured results of return loss in which the antenna covers the UWB band of 3.1-10.6 GHz with S11< -10 dB and it can be observed that the measured S-parameters are very well matching with the simulation results. It can be observed from Figure 7, that the antenna has a high port isolation of > 23 dB over the entire UWB spectrum. The performance of proposed antenna in terms of gain and radiation pattern are tested using Anechoic chamber which is shown is Figure 8. The summary of the performance of the proposed antenna with the existing UWB MIMO antennas is tabulated in Table 2. It can be observed that the proposed antenna is better in terms of size, simplicity and isolation. Table 2. Comparison of proposed MIMO antenna with existing structures No. of Bandwidth Complexity Size (λ × λ) elements (GHz) (Yes/No) [1] 2 3.1- 10.9 0.4 × 0.4 Yes [2] 2 3.1-5 0.34 × 0.16 Yes [3] 2 2.4- 9.2 0.32 × 0.32 Yes [4] 2 3.4-12 0.45 × 0.45 Yes [5] 2 2.87-10.4 0.28 × 0.25 Yes [6] 2 4.2-9 0.49 × 0.53 Yes [7] 2 3.1-10.6 0.22 × 0.26 Yes [8] 4 3-10.6 0.6 ×0.5 Yes [9] 2 2.83-10.18 0.5×0.3 Yes [10] 2 3-11.5 0.48×0.28 Yes [11] 1 4.5-11.8 1×0.7 No [12] 1 3-14 0.3×0.3 Yes [13] 2 2.41-2.46 0.96×1 Yes [14] 2 2.2-2.9 0.51×0.58 Yes Proposed 4 3.1-10.6 0.4 × 0.4 No th th *Sij is the isolation between the i antenna element and j antenna Ref.

(a)

Decouplers (Yes/No) Yes Yes Yes Yes Yes Yes Yes Yes Yes No No No Yes No No

Sij* (dB) >15 >22 >20 >15 >20 >16 >18 >20 >15 >15 Nil Nil >27 >20 >23

(b)

Figure 4. (a) Front view of proposed antenna and (b) Back view of proposed antenna

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 243 – 250

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



247

Figure 5. Measurement of isolation using vector network analyzer

Figure 6. Simulated and measured results of return loss

Figure 7 (a). Simulated and measured results of S(1,2), S(1,3) and S(1,4)

Compact planar ultrawideband MIMO antenna for wireless applications (P. Pavithra)

248



ISSN: 2252-8814

Figure 7 (b). Simulated and measured results of S(2,3), S(2,4) and S(3,4)

Figure 8. Measurement of gain and radiation pattern in ananechoic chamber 3.4. Peak gain The peak gain of the proposed antenna is 3.1 dBiat 9.9 GHz and the gain varies from –2.3 dBi to 3.1 dBi over the UWB spectrum as shown in Figure 9.

Figure 9. Peak gain plot Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 243 – 250

Int. J. of Adv. in Appl. Sci.

ISSN: 2252-8814



249

3.5. Radiation pattern In Figure 10 (a) to Figure 10 (c), the radiation patterns of the proposed antenna at xz and yz planes are investigated at 3.3 GHz, 5.5 GHz and 7.5 GHz, respectively. It can be seen that the antenna exhibits omni directional radiation pattern at xz plane and a figure of eight pattern in the yz plane. The radiation patterns are stable at lower, middle and higher frequencies ensuring good UWB performance.

Figure 10 (a). Radiation Pattern at 3.3 GHz

Figure 10 (b). Radiation Pattern at 5.5 GHz

Figure 10 (c). Radiation Pattern at 7.5 GHz 4.

CONCLUSION A compact, cost – efficient UWB MIMO monopole antenna has been designed and investigated. The antenna has -10 dB impedance bandwidth from 3.1–10.6 GHz. The proposed antenna isdevoid of decoupling circuits and has high isolation of >23 dB with stable radiation patterns within the operating band. It has a very compact size of 40 × 40 mm and it can be scalable to many number of elements in an antenna array which would be utilized for future 5G communication systems. ACKNOWLEDGEMENTS The authors would like to thank the institute of IIITDM Kancheepuram and VIT University, Chennai for providing the Vector network analyzer and Anechoic chamber, respectively for testing of antenna.

Compact planar ultrawideband MIMO antenna for wireless applications (P. Pavithra)

250



ISSN: 2252-8814

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

M. Gulam Nabi Alsath and Malathi Kanagasabai, “Compact UWB monopole antenna for automotive communications,” IEEE Transactions on Antennas and Propagation, vol. 63(9), pp. 4204-4208, 2015. S. Zhang and G. Pedersen, “Mutual coupling reduction for UWB MIMO antennas with a wideband neutralization line,” IEEE Antennas Wireless Propag., Lett., vol. 15(5), pp. 166-169, 2015. Inderpreet Kaur Sokhi, R. Ramesh, and Usha Kiran, “Design of UWB-MIMO Antenna for Wireless Applications,” IEEE WiSPNET 2016 conference, 2016. Jianfeng Zhu, Botao Feng, Biao Peng, Shufang Li, and Li Deng, “Compact CPW UWB diversity slot antenna with dual band-notched characteristics,” Microwave and Optical Technology Letters, vol. 58(4), pp. 989-994, 2016. Minal Dhanvijay, Anuradha Pattekar, and Rajiv Kumar Gupta, “Compact Circular Ring Shaped Monopole UWB MIMO Antenna,” 2017 IEEE 3rd International Conference on Sensing, Signal Processing and Security, 2017. Ankit Kumar Verma, R. Nakkeeran, and Rigvendra Kumar Vardhan, “Design of 2×2 single sided wrench shaped UWB MIMO antenna with high isolation,” 2016 International Conference on Circuit, Power and Computing Technologies [ICCPCT], 2016. Chao-Ming Luo, Jing-Song Hong, and Lin-Lin Zhong, “Isolation Enhancement of a Very Compact UWB-MIMO Slot Antenna with Two Defected Ground Structures,” IEEE Antennas and Wireless Propag., Lett., vol. 14(4), pp. 1766–1769, 2015. Ahmed A. Ibrahim, Mahmoud A. Abdalla, and ohn L. Volakis, “4 Elements UWB MIMO Antenna for Wireless Applications,” Antennas and Propagation & USNC/URSI National Radio Science Meeting, 2017. Chetan P. Bambarkar, Sukanya Kulkarni, “Design of ultra-wideband (UWB) MIMO antenna,” 2017 International Conference on Intelligent Computing and Control (I2C2), 2017. Ahmed A. Ibrahim, Jan Machac, and Raed M. Shubair, “Compact UWB MIMO Antenna with Asymmetric Coplanar Strip Feeding Configuration,” 2017 IEEE 28th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), 2017. P. Kumar, M. Tripathy, and H. P. Sinha, “Wide Band CPW Fed Slotted Microstrip Antenna,” TELKOMNIKA Indonesian Journal of Electrical Engineering and Computer Science, vol. 15(1), pp. 114-119, 2015. Ammar Alhegazi, Zahriladha Zakaria, Noor Azwan Shairi, Sharif Ahmed, and Tole Sutikno, “UWB Filtenna with Electronically Reconfigurable Band Notch using Defected Microstrip Structure,” TELKOMNIKA Indonesian Journal of Electrical Engineering and Computer Science, vol. 8(2), pp. 302-307, 2017. Charles MacWright Thomas, Huda A. Majid, Zuhairiah Zainal Abidin, Samsul Haimi Dahlan, Mohamad Kamal A. Rahim, and Raimi Dewan, “A Study on V-Shaped Microstrip Patch MIMO Antenna,” TELKOMNIKA Indonesian Journal of Electrical Engineering and Computer Science, vol. 5(3), pp. 606-611, 2017. S. Salihah, M. H. Jamaluddin, R. Selvaraju, and M. N. Hafiz, “A MIMO H-shape Dielectric Resonator Antenna for 4G Applications,” TELKOMNIKA Indonesian Journal of Electrical Engineering and Computer Science, vol. 10(2), pp. 648-653, 2018.

Int. J. of Adv. in Appl. Sci. Vol. 8, No. 3, September 2019: 243 – 250

Institute of Advanced Engineering and Science Indonesia

: D2, Griya Ngoto Asri, Bangunharjo, Sewon, Yogyakarta 55187, Indonesia

Malaysia

: 51 Jalan TU 17, Taman Tasik Utama, 75450 Malacca, Malaysia

COPYRIGHT TRANSFER FORM (Please compile this form, sign and send by e-mail) Please complete and sign this form and send it back to us with the final version of your manuscript. It is required to obtain a written confirmation from authors in order to acquire copyrights for papers published in the International Journal of Advances in Applied

Sciences (IJAAS) Full Name and Title Organisation Address and postal code City Country Telephone/Fax E-mail Paper Title Authors

Copyright Transfer Statement The copyright to this article is transferred to Institute of Advanced Engineering and Science (IAES) if and when the article is accepted for publication. The undersigned hereby transfers any and all rights in and to the paper including without limitation all copyrights to IAES. The undersigned hereby represents and warrants that the paper is original and that he/she is the author of the paper, except for material that is clearly identified as to its original source, with permission notices from the copyright owners where required. The undersigned represents that he/she has the power and authority to make and execute this assignment. We declare that: 1. This paper has not been published in the same form elsewhere. 2. It will not be submitted anywhere else for publication prior to acceptance/rejection by this Journal. 3. A copyright permission is obtained for materials published elsewhere and which require this permission for reproduction. Furthermore, I/We hereby transfer the unlimited rights of publication of the above mentioned paper in whole to IAES. The copyright transfer covers the exclusive right to reproduce and distribute the article, including reprints, translations, photographic reproductions, microform, electronic form (offline, online) or any other reproductions of similar nature. The corresponding author signs for and accepts responsibility for releasing this material on behalf of any and all co-authors. This agreement is to be signed by at least one of the authors who have obtained the assent of the co-author(s) where applicable. After submission of this agreement signed by the corresponding author, changes of authorship or in the order of the authors listed will not be accepted.

Retained Rights/Terms and Conditions 1. 2.

Authors retain all proprietary rights in any process, procedure, or article of manufacture described in the Work. Authors may reproduce or authorize others to reproduce the Work or derivative works for the author’s personal use or for company use, provided that the source and the IAES copyright notice are indicated, the copies are not used in any way that implies IAES endorsement of a product or service of any employer, and the copies themselves are not offered for sale. Although authors are permitted to re-use all or portions of the Work in other works, this does not include granting third-party requests for reprinting, republishing, or other types of re-use.

Yours Sincerely,

Corresponding Author‘s Full Name and Signature Date: ……./……./…………

IAES Journals, http://iaescore.com/journals International Journal of Advances in Applied Sciences (IJAAS), email: iaes.editor@gmail.com, editor@iaesjournal.com

Guide for Authors International Journal of Advances in Applied Sciences (IJAAS) is a peer-reviewed and open access journal dedicated to publish significant research findings in the field of applied and theoretical sciences. The journal is designed to serve researchers, developers, professionals, graduate students and others interested in state-ofthe art research activities in applied science areas, which cover topics including: chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences. Papers are invited from anywhere in the world, and so authors are asked to ensure that sufficient context is provided for all readers to appreciate their contribution.

The Types of papers The Types of papers we publish. The types of papers that may be considered for inclusion are: 1) Original research; 2) Short communications and; 3) Review papers, which include meta-analysis and systematic review

How to submit your manuscript All manuscripts should be submitted online at http://iaescore.com/journals/index.php/IJAAS

General Guidelines 1) Use the IJAAS guide (http://iaescore.com/journals/ourfiles/IJAAS_guideforauthors_2019.docx) as template. 2) Ensure that each new paragraph is clearly indicated. Present tables and figure legends on separate pages at the end of the manuscript. 3) Number all pages consecutively. Manuscripts should also be spellchecked by the facility available in most good word-processing packages. 4) Extensive use of italics and emboldening within the text should be avoided. 5) Papers should be clear, precise and logical and should not normally exceed 3,000 words. 6) The Abstract should be informative and completely self-explanatory, provide a clear statement of the problem, the proposed approach or solution, and point out major findings and conclusions. The Abstract should be 150 to 250 words in length. The abstract should be written in the past tense. 7) The keyword list provides the opportunity to add keywords, used by the indexing and abstracting services, in addition to those already present in the title. Judicious use of keywords may increase the ease with which interested parties can locate our article. 8) The Introduction should provide a clear background, a clear statement of the problem, the relevant literature on the subject, the proposed approach or solution, and the new value of research which it is innovation. It should be understandable to colleagues from a broad range of scientific disciplines. 9) Explaining research chronological, including research design and research procedure. The description of the course of research should be supported references, so the explanation can be accepted scientifically. 10) Tables and Figures are presented center. 11) In the results and discussion section should be explained the results and at the same time is given the comprehensive discussion. 12) A good conclusion should provide a statement that what is expected, as stated in the "Introduction" section can ultimately result in "Results and Discussion" section, so there is compatibility. Moreover, it can also be added the prospect of the development of research results and application prospects of further studies into the next (based on the results and discussion). 13) References should be cited in text by numbering system (in Vancouver style), [1], [2] and so on. Only references cited in text should be listed at the end of the paper. One author should be designated as corresponding author and provide the following information: • E-mail address • Full postal address • Telephone and fax numbers Please note that any papers which fail to meet our requirements will be returned to the author for amendment. Only papers which are submitted in the correct style will be considered by the Editors.

International Journal of Advances in Applied Sciences (IJAAS) Institute of Advanced Engineering and Science (IAES) e-mail: ijaas@iaesjournal.com

IJAAS Journal Order Form Volume

Number

Amount

Price (Rp)

Price (USD)

Total

Name

Company

Address

City / State

Zip

Telephone/Fax

: ........................, ..........................

Signature: ..................................

Order form for subscription should be sent to the editorial office by fax or email

Payment by Bank Transfer Bank Account name (please be exact)/Beneficiary: LINA HANDAYANI Bank Name: CIMB NIAGA Bank Branch Office: Kusumanegara City: Yogyakarta Country :Indonesia Bank Account # : 5080104447117 Bank Swift : BNIAIDJAXXX >>> Please find the appropriate price in the price list on next page >>>

The price list for domestic and foreign subscribers

Volume

Number

Year

1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4 5 5 5 5 6 6 6 6 7 7 7 7 8 8 8 8

1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4

2012 2012 2012 2012 2013 2013 2013 2013 2014 2014 2014 2014 2015 2015 2015 2015 2016 2016 2016 2016 2017 2017 2017 2017 2018 2018 2018 2018 2019 2019 2019 2019

Price (IDR) for domestic subscribers 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00

Price (USD) for foreign subscribers 36 36 36 36 36 36 36 36 40 40 40 40 40 40 40 40 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44

The price included the printing, handling, packaging and postal delivery fees of the hardcopy to the address of the authors or subscribers (with Registered Mail). For foreign subscribers, an additional fee is charged if you would your order is mailed via Express Mail Service (EMS): - $25 for ASIA Continent - $35 for AUSTRALIA Continent - $35 for AFRICA Continent - $39 for AMERICA Continent - $39 for EUROPE Continent (No additional fee for delivering your order by Registered Mail).

International Journal of Advances in Applied Sciences (IJAAS) Volume 8, issue 3, Sep. 2019

Articles inside

Nor Afandi Sharif, Rizauddin Ramli, Abdullah Zawawi Mohamed, Mohd Zaki Nuawi

Hassan Farahan Rashag

K. A. Varun Kumar, D. Arivudainambi

C. R. Balamurugan, K. Vijayalakshmi