JAMRIS 2012 Vol 6 No 2 by JAMRIS

pISSN 1897-8649 (PRINT) / eISSN 2080-2145 (ONLINE)

VOLUME 6

N째 2

2012

www.jamris.org

JOURNAL of AUTOMATION, MOBILE ROBOTICS & INTELLIGENT SYSTEMS

Editor-in-Chief Janusz Kacprzyk

Executive Editor: Anna Ładan aladan@piap.pl

(Systems Research Institute, Polish Academy of Sciences; PIAP, Poland)

Associate Editors: Jacek Salach (Warsaw University of Technology, Poland) Maciej Trojnacki (Warsaw University of Technology and PIAP, Poland)

Co-Editors: Oscar Castillo (Tijuana Institute of Technology, Mexico)

Dimitar Filev

Statistical Editor: Małgorzata Kaliczyńska (PIAP, Poland)

(Research & Advanced Engineering, Ford Motor Company, USA)

Kaoru Hirota Editorial Office: Industrial Research Institute for Automation and Measurements PIAP Al. Jerozolimskie 202, 02-486 Warsaw, POLAND Tel. +48-22-8740109, office@jamris.org

(Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology, Japan)

Witold Pedrycz (ECERF, University of Alberta, Canada)

Roman Szewczyk (PIAP, Warsaw University of Technology, Poland)

Editorial Board: Chairman: Janusz Kacprzyk (Polish Academy of Sciences; PIAP, Poland) Mariusz Andrzejczak (BUMAR, Poland) Plamen Angelov (Lancaster University, UK) Zenn Bien (Korea Advanced Institute of Science and Technology, Korea) Adam Borkowski (Polish Academy of Sciences, Poland) Wolfgang Borutzky (Fachhochschule Bonn-Rhein-Sieg, Germany) Chin Chen Chang (Feng Chia University, Taiwan) Jorge Manuel Miranda Dias (University of Coimbra, Portugal) Bogdan Gabryś (Bournemouth University, UK) Jan Jabłkowski (PIAP, Poland) Stanisław Kaczanowski (PIAP, Poland) Tadeusz Kaczorek (Warsaw University of Technology, Poland) Marian P. Kaźmierkowski (Warsaw University of Technology, Poland) Józef Korbicz (University of Zielona Góra, Poland) Krzysztof Kozłowski (Poznań University of Technology, Poland) Eckart Kramer (Fachhochschule Eberswalde, Germany) Piotr Kulczycki (Cracow University of Technology, Poland) Andrew Kusiak (University of Iowa, USA) Mark Last (Ben–Gurion University of the Negev, Israel) Anthony Maciejewski (Colorado State University, USA) Krzysztof Malinowski (Warsaw University of Technology, Poland) Andrzej Masłowski (Warsaw University of Technology, Poland)

Patricia Melin (Tijuana Institute of Technology, Mexico) Tadeusz Missala (PIAP, Poland) Fazel Naghdy (University of Wollongong, Australia) Zbigniew Nahorski (Polish Academy of Science, Poland) Antoni Niederliński (Silesian University of Technology, Poland) Witold Pedrycz (University of Alberta, Canada) Duc Truong Pham (Cardiff University, UK) Lech Polkowski (Polish-Japanese Institute of Information Technology, Poland) Alain Pruski (University of Metz, France) Leszek Rutkowski (Częstochowa University of Technology, Poland) Klaus Schilling (Julius-Maximilians-University Würzburg, Germany) Ryszard Tadeusiewicz (AGH University of Science and Technology in Cracow, Poland)

Stanisław Tarasiewicz (University of Laval, Canada) Piotr Tatjewski (Warsaw University of Technology, Poland) Władysław Torbicz (Polish Academy of Sciences, Poland) Leszek Trybus (Rzeszów University of Technology, Poland) René Wamkeue (University of Québec, Canada) Janusz Zalewski (Florida Gulf Coast University, USA) Marek Zaremba (University of Québec, Canada) Teresa Zielińska (Warsaw University of Technology, Poland)

Publisher: Industrial Research Institute for Automation and Measurements PIAP

If in doubt about the proper edition of contributions, please contact the Executive Editor. Articles are reviewed, excluding advertisements and descriptions of products. The Editor does not take the responsibility for contents of advertisements, inserts etc. The Editor reserves the right to make relevant revisions, abbreviations and adjustments to the articles.

JOURNAL of AUTOMATION, MOBILE ROBOTICS & INTELLIGENT SYSTEMS VOLUME 6, N° 2, 2012

CONTENTS 3

Improvement of Control and Supervision of Web Connected Mobile Robots Using PPU Computation Janusz Bedkowski, Andrzej Masłowski 8

Implementation of the Chaotic Mobile Robot for the Complex Missions Ashraf Anwar Fahmy 13

Non-invasive Identification of Servo Drive Parameters Reimund Neugebauer, Arvid Hellmich, Stefan Hofmann, Holger Schlegel 17

Human-machine Interface for Presentation Robot Jiri Krejsa, Vit Ondrousek 22

Novel Developments in Dimensional Nanometrology in the Context of Geometrical Product Specifications and Verification (GPS) M. Numan Durakbasa, P. Herbert Osanna, Gökcen Bas, Pınar Demircioglu, Mehmet Cakmakci, Adriana Hornikova 26

Simultaneous Measurement of Two Parameters by Double Current Supplied Bridge Adam Idzkowski, Jaroslaw Makal, Zygmunt L. Warsza 32

Decentralized PI Controller for Multi Motors Web Winding System Bousmaha Bouchiba, Abdeldjebar Hazzab, Hachemi Glaoui, Fellah Med-Karim, Ismaïl Khalil Bousserhane, Pierre Sicard

Methods of Decreasing the Influence of the Factors Disturbing the Reliability of Leak Detection Systems Mateusz Turkowski, Andrzej Bratek

Articles

Modelling of Mechatronic Devices Supported by 3D Engineering Software Jakub Wierciak, Ksawery Szykiedans, Aleksandra Binder-Czajka 47

The Process of Aluminium Moulds Warming in the Car Industry Jaroslav Mlýnek, Radek Srb 51

Analyses of Micro Moulding Process of the Microelements from Ceramic Powders Andrzej Skalski, Dionizy Biało, Waldemar Wisniewski, Lech Paszkowski 55

Using Reverse Engineering in Archaeology: Ceramic Pottery Reconstruction Calin Neamtu, Sorin Popescu, Daniela Popescu, Razvan Mateescu

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Improvement of Control and Supervision of Web Connected Mobile Robots Using PPU Computation Submitted 27th June 2011; accepted 16th September 2011

Janusz Be˛dkowski, Andrzej Masłowski

Abstract: The paper concerns the research related to the improvement of control and supervision of web connected mobile robots using Physic Processing Unit (PPU). PPU computations taken into the consideration include rigid body dynamics, collision detection and raycasting. The result is improved by Human Machine Interface that allows performing semantic simulation during multi robot task execution. Semantic simulation engine provides tools to implement the mobile robot simulation, which is based on real data delivered by robot’s observations in INDOOR environment. The supervision of real objects such as robots is performed by association with its virtual representation in the simulation, therefore events such as object intersection, robot orientation - pitch and roll are able to be monitored. The simulation can be integrated with real part of the system with an assumption of robust localization of real entities, therefore Augmented Reality capabilities are available. Keywords: semantic mapping, Human Machine Interface, mobile robot

1. Introduction

The main problem undertaken in the paper is improved control and supervision of web connected mobile robots [2], for example, inspection intervention robot system, with an application of Physic Processing Unit (PPU). The main applications of multi-robot inspection intervention system are actions in a disaster area, covering all the consequences of fire, chemical hazards, and for example the effect of terrorist attack. The environment of the system forces short time of inspection, and determines basic goals for the system. This provides clearly defined working conditions, the criteria for checking correctness of control and supervision algorithms. Many studies have shown extensive technical development in the area of mobile robotics. There have been many solutions [8] for technical issues related to unique mechatronics designs of mobile robots. Many new robots have high mobility [11] in difficult terrain. In addition, number of robots equipped with sophisticated sensors [12] increases, which enhances the effectiveness for example search and detection of victims [13], [14]. The paper concerns semantic mapping related to robot ontology [16] that represents a neutral knowledge representation capturing relevant information about robots, their capabilities and environment. It can assist in the development and testing effective technologies for sensing, mobility, navigation, planning integration and opera-

tor interaction. In general, ontologies make all pertinent knowledge about a domain [15] explicit and are represented in a computer-interpretable fashion that allows software to reason over that knowledge to infer with additional information. Applied ontology in semantic simulation engine [1] allows automatic generation of virtual environments based on robot’s observations and to perform semantic simulation based on a PPU capabilities. Physics engines dedicated for video games typically contain two core components, a collision detection/collision system, and the dynamics simulation component responsible for solving the forces affecting the simulated objects. Modern physics engines may also contain fluid simulations, animation control systems and asset integration tools. There are three major paradigms for the physical simulation of solids: - Penalty methods, where interactions are commonly modeled as mass-spring systems. This type of engine is popular for deformable, or soft-body physics. - Constraint based methods, where constraint equations are solved that estimate physical laws. - Impulse based methods, where impulses are applied to object interactions. Finally, hybrid methods are possible that combine aspects of the above paradigms. An early academic PPU research project [3], [4] named SPARTA (Simulation of Physics on A Real-Time Architecture) was carried out at Penn State [5] and University of Georgia. This was a simple FPGA based PPU that was limited to two dimensions. First commercial available dedicated PPU PhysX from Ageia was appeared in February 2006 [6]. The unit was most effective in accelerating particle systems, with only a small performance improvement measured for rigid body physics. PhysX is a proprietary realtime physics engine middleware SDK acquired by Ageia [7]. Computer programs such as modern video games supporting hardware acceleration by PhysX can be accelerated by either a PhysX PPU or a CUDAenabled GeForce GPU. Thus offloading physics calculations from the CPU allow it to perform other tasks instead – resulting in a smoother gaming experience and additional visual effects. An analysis published in [10] shows that most of the code used in PhysX applications is based on x87 instructions without any multi-threading optimization. This could cause significant performance drops when running PhysX code on the CPU. It was suggested that a PhysX rewritten using SSE instructions may substantially lessen the performance discrepancy between CPU PhysX and GPU PhysX. The paper is organized as follows: section 2 describes multi robotic system structure. Section 3 is related to Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

PPU computation applied for rigid body simulation/ collision detection, raycasting and triggers. Section 4 describes proposed semantic simulation engine with an implementation of semantic Human Machine Interface (main improvement). Conclusions are noted in section 5.

2. Multi robotic system structure

The main object of research is a web connected mobile robot system, for example, the inspection – intervention system consisting of a mobile base station, a group of autonomous robots and remote-controlled robot, equipped with a manipulator with n degrees of freedom. Figure 1 shows structure of such a system.

Fig. 1. Inspection – intervention system The main tasks of the system is an inspection and intervention of hazardous environment for human activities. More details concerning base station, mobile robots and supervision and control can be found in [2].

3. PPU computation

Physic Processing Unit (PPU) is a dedicated microprocessor for physic calculations. With central processing unit (CPU) and general-purpose computing on graphics processing units (GPGPU) PPU builds modern computers capable to execute sophisticated computational tasks in parallel. Examples of calculations involving a PPU include rigid body dynamics, soft body dynamics, collision detection, fluid dynamics, hair and clothing simulation, finite element analysis, and fracturing of objects. The main goal for PPU is to decrease CPU computation. In this paper we are focused on NVIDIA Ageia PhysX chip that is a complete one designed, marketed and supported PPU available with GeForce graphic cards. PhysX is an open standard and gives an opportunity to develop advanced software for mobile robotics for example Microsoft Robotics Developer Studio (MRDS) [9]. 3.1. Actors – basic elements of the simulation An actor typically contains shapes sometimes one, sometimes more, sometimes none. The shapes represent the physical space that actor occupies in the simulation, and they are the core of collision detection. It is possible to create an actor without shapes, which will then behave as a “ghost” and collides with nothing (though it may still be jointed to other actors and affected by gravity and other applied forces or torques). In semantic simulation all semantic entities are related to actors. Geometric at4

Articles

VOLUME 6,

N° 2

2012

tributes correspond to a shape of actor, for example wall will be simulated using rectangular prism. Robot is represented in the simulation by set of actors connected via joints. Each actor is associated with a shape – rectangular prism (box shape). Relationships between semantic entities are modeled using joints, for example solid wall orthogonal to floor corresponds to rectangular prism connected to a plane via fixed joint.

3.2. Continuous collision detection

It is useful for fast moving objects. With traditional collision detection, fast moving objects pass through other objects during a single time step. This effect is known as tunneling [17]. To deal with this problem, a technique known as Continuous Collision Detection (CCD) is necessary. Instead of testing for collision at discrete points, it tests an extruded volume which represents the object’s motion during the whole time step. If a collision is detected, the time of impact can be computed and the object’s motion constrained appropriately. The example is shown in Figure 2:

Fig. 2. An example of Continuous Collision Detection (CCD)

3.3. Raycasting Raycasting is basically a collision detection using rays. However, instead of attaching the line segment to an actor like other shapes, it gets cast into the scene at the user’s request, stabbing one or more shapes in the process. It has many uses, for example picking up objects in the Human Machine Interface to assign new goal for robot in 3D environment. In our robotic application raycasting is used also for laser system measurement simulation (Figure 3).

Fig. 3. Raycasting example – laser measurement system simulation

3.4. Triggers A trigger is a shape that permits other shapes to pass through it. Each shape passing through it can create an

Journal of Automation, Mobile Robotics & Intelligent Systems

event for the user when it enters, leaves, or simply stays inside the shape. Triggers are used to implement the monitoring of robot’s position.

4. Semantic simulation engine

The concept of semantic simulation engine applied in mobile robot system is a new idea, and its strength lies on the semantic map integration with mobile robot simulator. Semantic simulation engine is composed of data registration modules, semantic entities identification (data segmentation) modules and semantic simulation module. It provides software tools for the implementation of the mobile robot simulation based on real data delivered by robot and processed on-line using parallel computation. Semantic entities identification modules can classify door, walls, door, ceiling, stairs in INDOOR environment. Data can be delivered by robot’s observation based on modern sensors such as laser measurement system 3D and RGB-D cameras. Real objects are associated with virtual entities of simulated environment. Semantic entities identification module uses semantic net shown in Figure 4. From computation point of view image processing methods are applied for automatic semantic elements identification based on images derived from projected 3D cloud of points onto OXY plane. An example of automatic generated semantic entities such as walls, door and stairs is shown in Figure 5. More details concerning semantic prerequisites generation can be found in [1].

VOLUME 6,

N° 2

2012

can be extended by another objects, what is dependent on robust and accurate 3D scene analysis), robot simulator nodes (entities) Lrs={robot, rigid body object, soft body object…}, semantic map relationships between the entities Rsm= {parallel, orthogonal, above, under, equal height, available inside, connected via joint…}, robot simulator relationships between the entities Rrs = {connected via joint, position…}, semantic map events Esm, that are the same as a robot simulator events: Esm =Ers = {movement, collision between two entities started, collision between two entities stopped, collision between two entities continued, broken joint…}. The entities from semantic map correspond to actors in PhysX. Lsm is transformed into Lrs based on spatial model generated based on registered 3D scans i.e. walls, doors and stairs correspond to actors with BOX shapes. Rsm are transformed into Rrs with remark that doors are connected to walls via revolute joints. All entities/relations Rsm has the same initial location in Rrs, the location of each actor/ entity may change during simulation. The transformation from Esm to Ers effects that events related to entities from semantic map correspond to the events related to actors representing proper entities. Following events can be noticed during simulation: robot can touch each entity, open/close the door, climb the stairs, enter empty space of the door, damage itself (broken joint between actors in robot arm), brake the joint that connects door to the wall. It is noteworthy to mention that all robot simulator’s semantic events are useful for semantic HMI, where computer can monitor simulation events. From robot control point of view semantic simulation is used also for robot path planning using “empty spaces” (see Figure 6). “Empty spaces” compose graph, therefore it is possible to use classic path planning methods.

4.2. Semantic HMI

To demonstrate proposed approach Figure 6 shows basic robot task – going forward via corridor.

Fig. 4. Semantic net

Fig. 5. Semantic entities generated from 3D cloud of points

4.1. Semantic simulation

Accurate positioning of mobile robots, accurate encoders of inspection robot arm and satisfying tracking system are needed to update virtual entities position during real robot task execution. The semantic simulation is composed of: semantic map nodes(entities) Lsm={Wall, Wall above door, Floor, Ceiling, Door, Free space for door, Stairs…}, (It is important to noticed that the Lsm set

Fig. 6. Semantic HMI example. Robot, walls (green rectangular prisms), empty spaces (blue cubes) Semantic HMI visualizes robot position represented by rectangular shapes, environment objects – in this case walls and empty spaces. Semantic HMI supervises intersection of robot shapes with environment shapes to avoid potential collisions. HMI monitors robot’s position using triggers – empty spaces. Defined goal from operator console’s point of view is shown in Figure 7 (where raycaArticles

Journal of Automation, Mobile Robotics & Intelligent Systems

Fig. 7. Defining goal (red cube) in operator console

VOLUME 6,

N° 2

2012

sting method was used to pick up cubic shape). Figures 8 shows monitoring of the robot’s position, where robot is visiting new empty spaces and generate semantic events via triggers. Achieved goal by robot is shown in Figure 9, it is also related with a semantic event generated automatically by a trigger. The main difference comparing to a classic HMI proposed semantic HMI generates linguistic description of robot status, for example: Robot collides with wall. Robot is inside empty space. Robot achieved goal.

5. Conclusion

In this paper an improvement of supervision and control of web connected mobile robots using Physic Processing Unit (PPU) computation is shown. The approach is related with semantic simulation engine that allows building robot simulation based on robot’s observations (laser measurement system 3D, RGB-D cameras). Proposed ontological approach was used for creating semantic Human Machine Interface that is more ergonomic that potentially will improve human – robot interactions. The main achievement is the implementation of a semantic HMI that can be integrated with modern mobile robotic systems. HMI can integrate classic methods for path planning with sophisticated semantic map building, where decisions are made in high conceptual level. This research is showing an idea of replacing the geometrical description of the robot environment by ontology – based semantic representation, where objects are described by its attributes and relations, therefore the implementation is closer to a human natural understanding of the world. We believe that future systems will be more ergonomic by usage presented software techniques.

Acknowledgements This work is performed during postdoctoral scholarship (CAS/19/POKL 15.05.2011-15.11.2011) in Royal Military Academy, Unmanned Ground Vehicle Center, Brussels, Belgium, funded by Center for Advanced Studies, Warsaw University of Technology (project: “Priority IV of Human Capital Operational Programme from European Social Fund”).

Authors Fig. 8. Monitoring robot position

Janusz Będkowski* – Industrial Research Institute for Automation and Measurements PIAP, Al. Jerozolimskie 202, 02-486 Warsaw, Poland; Institute of Automation and Robotics, Warsaw University of Technology, św. A. Boboli 8, 02-525 Warsaw, Poland, januszbedkowski@gmail.com Andrzej Masłowski – Industrial Research Institute for Automation and Measurements Al. Jerozolimskie 202, 02-486 Warsaw, Poland; Institute of Automation and Robotics, Warsaw University of Technology, św. A. Boboli 8, 02-525 Warsaw, Poland, amaslowski@piap.pl a.maslowski@mchtr.pw.edu.pl

Fig. 9. Achieving goal by a robot 6

Articles

*Corresponding author

Journal of Automation, Mobile Robotics & Intelligent Systems

References:

[1] J. Będkowski, A. Masłowski, “Semantic simulation engine for mobile robotic applications”, Pomiary, Automatyka, Robotyka, 2/2011, Automation 2011, 6th-8th April, Warsaw, pp. 333-343, [2] J. Bedkowski, A. Masłowski, “Methodology of control and supervision of web connected mobile robots with CUDA technology application”, Journal of Automation, Mobile Robotics and Intelligent Systems, vol. 5, no. 2, 2011, pp. 3-11. [3] S. Yardi, B. Bishop, T. Kelliher, “HELLAS: A Specialized Architecture for Interactive Deformable Object Modeling”. In: ACM Southeast Conference, Melbourne, FL, 10th-12th March, 2006, pp. 56–61. [4] B. Bishop, T. Kelliher, “Specialized Hardware for Deformable Object Modeling”, IEEE Transactions on Circuits and Systems for Video Technology, vol. 13, no. 11, Nov. 2003, pp. 1074-1079. [5] “SPARTA Homepage”. Cse.psu.edu. http://www. cse.psu.edu/~mdl/sparta/. Retrieved 2010-08-16. [6] “Exclusive: ASUS Debuts AGEIA PhysX Hardware”. AnandTech. http://www.anandtech.com/ show/2001/4. Retrieved 2010-08-16. [7] NVIDIA Corporation (2008-02-13). NVIDIA completes Acquisition of AGEIA Technologies. Press release. http://www.nvidia.com/object/ io_1202895129984.html. Retrieved 2009-03-18. [8] J. Casper, R. R. Murphy, “Human-robot interactions during the robot-assisted urban search and rescue response at the World Trade Center”. IEEE Transactions on Systems, Man and Cybernetics, Part B, vol. 33, 2003, pp. 367-385.

VOLUME 6,

N° 2

2012

[9] http://msdn.microsoft.com/pl-pl/robotics/ [10] Real World Technologies – PhysX87: Software Deficiency [11] B. Shah, H. Choset, Survey on urban search and rescue robotics. Technical report, Carnegie Mellon University, Pittsburg, Pennsylvania, 2003. [12] M. Strand, R. Dillmann, “3D-Environment Modeling using an autonomous Robot”. In: IARP-Workshop on Environmental Maintenance & Protection, CD-ROM, 2008. [13] G. De Cubber, G. Marton, “Human Victim Detection”. In: International Workshop on Robotics for risky interventions and Environmental Surveillance, RISE’2009,Brussels, 12th-14th January 2009, CD-ROM. [14] H. Sugiyama, T. Tsujioka, M. Murata, “Victim detection system for urban search and rescue based on active network operation”. In: iidem, Design and application of hybrid intelligent systems, IOS Press, Amsterdam, 2003, pp. 1104-1113. [15] F. Harmelen, D. McGuiness, “OWL Web Ontology Language Overview,” W3C web site: http://www. w3.org/TR/2004/REC-owl-features-20040210/, 2004. [16] C. Schlenoff, E. Messina, “A Robot Ontology for Urban Search and Rescue”. In: Proceedings of the 2005 CIKM Conference: Workshop on Research in Knowledge Representation for Autonomous Systems, 31st October – 5th November, 2005, Bremen, Germany. [17] NVIDIA PhysX SDK 2.8 – Introduction.

Articles

FOR THE COMPLEX MISSIONS Submitted 3rd June 2011; accepted 10th January 2012

Journal of Automation, Mobile Robotics & Intelligent Systems

Ashraf Anwar Fahmy

VOLUME 6,

N° 2

2012

Abstract:

robots designed for scanning of unknown workspaces with borders and barriers of Mobile robotics, after decades of unknown shape, as in patrol or cleaning continuous development, keeps up as an purposes. The aim of this paper is to implement intensive research issue because of its everthe chaotic behavior of the mobile robot and increasing application to different domains and to evaluate the performance of the Arnold, its economical and technological relevance. Submitted 3rd June 2011; accepted 10th January 2012 Chua’s circuit equations, as Lorenz, and the Interesting applications can be seen in robots a controller of the mobile robot, from point of performing floor-cleaning tasks, executing view of performance index k, which reflects industrial transportation, exploring volcanoes, how high the coverage area, and the evenness scanning areas to find explosive devices, and Ashraf Anwar Fahmy E, which reflects the degree of variation index so on. a chaotic signal for an autonomous in covering the areas between species. The mobile robot is to increase and to take paper structure is as follows: The next section advantage of coverage areas resulting from its presents thetomobile robot model. The dedicated the evaluation criteria to bechaotic applied. Section Abstract: travelling paths. The chaotic behavior of the mobile controllers are illustrated in section mobile robot is achieved by adding nonlinear 5 is reserved to the simulation results. Finally3.section 6 Mobile robotics, after decades of continuous development, Section 4 is dedicated to the evaluation criteria into theresearch robot kinematic equations, concludes this paper. keepsequations up as an intensive issue because of its evto be applied. Section 5 is reserved to the like Arnold, Lorenz,toand Chua’s equations, er-increasing application different domains and itsthat ecosimulation results. Finally section 6 concludes this are well known equations for had a chaotic nomical and technological relevance. Interesting applica2. The Mobile Robot Model paper. behavior. The performance of the three guiding tions signals can be seen in robots performing floor-cleaning tasks, The mobile robot used is shown in Fig. 1. Let the linfor robotics system is evaluated in the executing transportation, exploring volcanoes, of Robot the robot v [m/s], and the the angular ve2. ear Thevelocity Mobile Model senseindustrial of the wide area coverage, the evenness scanning areas findtotal explosive devices, and so on. a chalocity w [rad/s], be the inputs to the system, and the state index, andtothe trajectory distance. The mobile robot used is shown in Fig. 1. Let otic signal for an autonomous mobile robot is to increase equation of the mobile robot is written as follows: the linear velocity of the robot v [m/s], and the Keywords : Chaos, chaotic motion, and to take advantage of coverage areas resultingchaotic from its the angular velocity w [rad/s], be the inputs to mobile robot, Chua’s circuit, Arnold and Lorenz 0 travelling paths. The chaotic behavior of the mobile robot is  x  state cos qequation the system, and the of the equations      u achieved by adding nonlinear equations into the robot kinemobile robot is written as= follows: y sin 0 q   w x� cos θ 0 υ   matic equations, like Arnold, Lorenz, and Chua’s equations,    0 1 q     1. Introduction �y� � � � sin θ 0� � � (1) that are well known equations for had a chaotic behavior. (1) ω �θ 0 1 The performance of the three guidingone signals for robotics Where x [m], and y [m] is the position of the mobile The chaos characterizes of mysterious Where x [m], and y [m] is the position of the rich behaviorsin of dynamical system is evaluated thenonlinear sense of the wide area systems. coverage, robot, is the angle of the robot. mobile robot, �� is the angle of the robot. Many index, research efforts have been paid to the evenness and the total trajectory distance. establish the mathematical theory behind Y chaos. chaos, Applications of chaos are mobile also being Keywords: chaotic motion, chaotic robot, �(t) studied and include, for example, controlling Chua’s circuit, Arnold and Lorenz equations chaos and chaotic neural networks. This paper follows a method to impart chaotic behavior to �(�) a mobile robot. This is achieved by designing 1. Introduction a controller which ensures chaotic motion [1, 2]. �(�) Further on chaotic trajectories of The chaos investigations characterizes one of mysterious rich behayr same type of the robot Many using research other viorsthe of nonlinear dynamical systems. equations were carried out in [3–11]. The main efforts have been paid to establish the mathematical theobjective in exploiting chaotic signals for an ory behind chaos. Applications of chaos are also being autonomous mobile robot is to increase and to X studied and include, for example, controlling chaos and take advantage of coverage areas resulting xr chaotic neural networks.paths. This paper a method from its travelling Largefollows coverage areasto impart chaotic behavior a mobile robot. Thissuch is achieFig. 1. Geometry of the robot motion on the Cartesian are desirable for to many applications as

Implementation of The Chaotic Mobile Robot for The Complex Missions

ved by designing a controller which ensures chaotic motion [1, 2]. Further investigations on chaotic trajectories of the same type of the robot using other equations were carried out in [3–11]. The main objective in exploiting chaotic signals for an autonomous mobile robot is to increase and to take advantage of coverage areas resulting from its travelling paths. Large coverage areas are desirable for many applications such as robots designed for scanning of unknown workspaces with borders and barriers of unknown shape, as in patrol or cleaning purposes. The aim of this paper is to implement the chaotic behavior of the mobile robot and to evaluate the performance of the Arnold, Lorenz, and the Chua’s circuit equations, as a controller of the mobile robot, from point of view of performance index k, which reflects how high the coverage area, and the evenness index E, which reflects the degree of variation in covering the areas between species. The paper structure is as follows: The next section presents the mobile robot model. The chaotic mobile controllers are illustrated in section 3. Section 4 is 8

Articles

plane

3. The Chaotic Mobile Robot Controller

In order to generate chaotic motions of the mobile robot, this is achieved by designing a controller which ensures chaotic motion. The type of chaotic patterns employed to generate the robot trajectory are the Arnold, the Lorenz, and the Chua’s circuit equations.

3.1. The Arnold equation

The equation of the Arnold is written as follows: (2)

Where A, B, and C are constants. It is known that the Arnold equation shows periodic motion when one of the constants, for example C, is 0 or small and shows chaotic motion when C is large [10]. The chaotic pattern of the

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

2012

60 50

Arnold equation, for the following parameters: A = 0.27, B = 0.135, C = 0.135 and initial conditions: x10 = 4, x20 = 3.5, x30 = 0, is shown in Fig. 2 After integration the Arnold equation (2) into the controller of the mobile robot equation (1), the state equation of the mobile robot becomes:

N° 2

40 30 20 10

0 50

-50

-20

-10

-5

Fig. 4. The Lorenz attractor in 3D space

-10 -15 25

10 X

-5

-10

-30

-25

-15

-20

-10

-5

The parametric values in the differential equation (4) are needed in order to generate a chaotic behavior. The Lorenz attractor is shown in Fig. 4 In the same way that was made with the Arnold equation, we coupled the Lorenz equation with the mobile robot equation, and the integrated system will be:

Fig. 2. Arnold chaotic pattern in 3-D space

X = −10 X + 10Y = 28 X − Y − X Z Y = − 8 Z + XY Z 3 x = v cos Z

(3) The integrated system of the Arnold equation with the mobile robot equation with appropriate adjusting parameters and initial conditions guaranteed that a chaotic orbit of the Arnold equation behaves chaotically. The resultant workspace coverage trajectory of the mobile robot at iteration n = 10000 is shown in Fig. 3

y = v sin Z The resultant workspace coverage trajectory of the mobile robot at initial conditions: X0 = 1, Y0 = 0, Z0 = 1, x0 = 1, y0 = 0 at iteration n = 10000 is shown in Fig. 5 Workspace coverage trjectory-Lornz

500 Workspace coverage trajectory-Arnold

400 20

300 yr[m]

yr[m]

200

100 -10

0 -20

-30 -40

-100 -200 -30

-20

-10

xr[m]

Fig. 3. Workspace coverage trajectory of the chaotic mobile robot, Arnold

3.2. The Lorenz equation

The Lorenz attractor is generated by the differential equation given by:

= −10 X + 10Y X = 28 X − Y − X Z Y

-100

-150

(4)

xr[m]

-50

Fig. 5. Workspace coverage trajectory of the chaotic mobile robot, Lorenz

3.3. The Chua’s circuit

The chaotic controller used herein as a trajectory generator is Chua’s circuit which is low cost and easy to construct for trajectory generators. The general equations of Chua’s circuit are:

= – ( X − X − f ( X )) X 1 2 1 1

X 2 = X1 − X 2 − X 3

= − 8 Z + XY Z 3 Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

Where: F ( X1 ) = bX1 +

VOLUME 6,

1 (a − b)[ X1 + 1| − X1 − 1 ] 2

The evaluation criteria are set according to the application purpose. Since we would like to use the robot in wandering around area in the area of no maps, the chaotic trajectory should cover the entire areas of patrolling as much as possible. The following three performances criteria are to be considered to evaluate the coverage rate of the chaotic mobile robot, namely the performance index k, the evenness index E, and the distance of the trajectory D. a) A performance index K representing a ratio of areas that the trajectory passes through or used space (Au), over the total working area (At)

These parameters generate double scroll attractor shown in Fig. 6. 4

2 0 -2 -4 0.5 0 -0.5

-1

-2

-3

Fig. 6. Chua’s attractor in the 3D space The integrated system of the Chua’s circuit equation as a controller of the mobile robot will be as follows:

X 2 = X1 − X 2 − X 3

X3 = − b X 2 x = v cos X 2 y = vsin X 2 The resultant workspace coverage trajectory of the mobile robot at iteration n = 10000 is shown in Fig. 7.

KQ =

3.4. Workspace coverage constrain

The workspace coverage trajectory of the three mentioned controllers, Arnold, Lorenz, and Chua’s equations are studied and analyzed. We can deduce from Fig. 5, Fig. 6, and Fig. 7 that each controller has a specific workspace coverage trajectory where the chaotic mobile robot moves in. We candidate three areas: (20 x 20, 40 x 40, and 60 x 60) for each controller as follows in Table 1.

(8)

A uQ A tQ

(9)

Where is the performance index of the Qth quadrant, is the area used by the trajectory in the Qth quadrant. In our case, we have

A tQ =

At 4

(10)

Table1. Workspace coverage constrain for the controllers

Workspace coverage trajectory-Chua's circuit

20 y[m]

Au At

The used area and the total area can be calculated by the following algorithm: • We divide the specified area into (NxN) pixels. • Initially, assign the value 0 for all pixels. • We get the x-coordinate and y-coordinate of the pixels which passes through the trajectory of the robot. • We assign the value 1 for each pixel passes through the trajectory. • We count the number of ones (pixels which passes through the trajectory) which is. • We count the number of zeroes Z (pixels which don’t passes through the trajectory). • The total area is the sum of and Z. Similarly, let us consider a rectangular shape area, Fig. 8. The total area can be partitioned into four quarter, denoted Q = 1, 2, 3, 4. The quantitative measurement of the trajectory can be evaluated by using the following equation:

= – ( X − X − f ( X )) X 1 2 1 1

2012

4. Evaluation Criteria

a = 9, b = 100/7, a = -5/7, b = -8/7.

N° 2

Controller

Chua’s work-

Lorenz’s work-

Arnold’s work-

Area

space constrain

20x20

X: 0

Y: 0

X: 0

X: - 40

X: 0

Y: 0

Y: 200

240

Y: 0

X: 20

X: -50

X: -30

Y: -60

Y: 200

260

Y: -30

-20

-40

-60 -20

40x40 0

40 x[m]

100

Fig. 7. Workspace coverage trajectory of the chaotic mobile robot, Chua’s circuit 10

Articles

60x60

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

Equations (8)-(10) will be used as performance indices in section 4. b) An evenness index E refers to how close in numbers each species in an environment are. The evenness index can be represented in our situation by [12]. s

K Qln(k Q )

(11)

Q =1

ln(s)

Where s: No of species = 4 Quarters in our case. E is constrained between 0 and 1. The less variation in covering the areas between the species, the higher E is. c) The total distance of the trajectory D. The total distance of the generated trajectory of each controller should be taken in the account to measure the performance of the controller in coverage a certain area, and it can be calculated by the following formula:

D = (x i+1 − x i ) + (y i+1 − y i ) 2

6. Conclusion

In this paper, we proposed the implementation of chaotic behavior on a mobile robot, which implies a mobile robot with a controller that guarantees its chaotic motion. The Arnold, the Lorenz, and the Chua’s equations, which are known to show the chaotic behavior, were adopted as the chaotic dynamics to be integrated into the mobile robot and the behaviors of these equations were evaluated from point of view of performance index k, which reflects Trajectory of the chaotic mobile robot-"Arnold"

(12)

Where: Xi + 1 and Xi are the x- coordinates at successive instants & Yi + 1 and Yi are the y- coordinates at successive instants.

2012

The plot of the performance index K, the evenness index E, and the total distance of the trajectory versus iterations n of the simulation for the Arnold, Lornz, and Chua’s controller in the three areas (20 m x 20 m), (40 m x 40 m), and (60 m x 60 m) are depicted in Fig. 12, Fig. 13, and Fig. 14, respectively.

y[m]

∑ E = 1−

N° 2

20 Q=2

Q=1

y[m]

10 0

-5 -5

Au1

Au2 Q=3

-10

10 x[m]

Fig. 9. The trajectory of the chaotic mobile robot controlled by the Arnold equation

Q=4 Au4

-20

Trajectory of the chaotic mobile robot-"chua's circuit"

Au3 20

-10

-5

5 x[m]

Fig. 8. Partition of the specified area

15 y[m]

-30 -15

5. Simulation Results

-5 -5

10 x[m]

Fig.10. The trajectory of the chaotic mobile robot controlled by the Lorenz equation Trajectory of the chaotic mobile robot-"Lornz"

15 y[m]

In order to evaluate the performance of the three controllers, Arnold, Lorenz, and Chua’s equations, used to generate the chaotic motion of the mobile robot, we simulate the three systems of equations (3), (5), and (7) given in section 3, in three different workspace areas (20 m x 20 m), (40 m x 40 m), and (60 m x 60 m), as specified in Table 1 illustrated in section 3. We simulate the system of equations using the parameters given in section 2 for each controller and the velocity v of the robot is 1 m/s. We use the performance index K, the evenness index E and the total distance of the chaotic mobile robot D, as the evaluation criteria to distinguish the performance between the three controllers in the three specified workspace areas. The chaotic trajectory of the mobile robot for the three specified controllers, Arnold, Lorenz, and Chua’s equations, in (20 m x 20 m) workspace at iteration n = 6000, number of pixels to cover area = N x N, N = 2000, integration step h = 0.1 and the parameters given in section 3, are shown in Fig. 9, Fig. 10, and Fig. 11 respectively. The robot moves as if is reflected by the boundary „mirror mapping”.

-5 -5

10 x[m]

Fig.11. The trajectory of the chaotic mobile robot controlled by the Chua’s circuit Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

100 50

K % ,chua K % ,Lornz K% , Arnold

0 1000

2000

3000

4000

5000

6000

7000

8000

9000

VOLUME 6,

10000

Iterations "n" 100 50

E % ,chua E % ,Lornz E% , Arnold

0 1000

2000

0 1000

3000

4000

5000 6000 Iterations "n"

7000

8000

9000

10000

3000

4000

5000 6000 Iterations "n"

7000

8000

9000

10000

D, chua D, Lornz D, Arnold

2000

Fig. 12. Plot of K, E, and D of the controllers in (20 m x 20 m) workspace area 100 50 0 1000 100 50 0 1000 10000 5000 0 1000

4000

6000 5000 Iterations "n"

7000

8000

9000

10000

3000

4000

5000 6000 Iterations "n"

7000

8000

9000

10000

3000

4000

5000 6000 Iterations "n"

7000

8000

9000

10000

3000

E % ,chua E % ,Lornz E% , Arnold

2000

D, chua D, Lornz D, Arnold

2000

Fig. 13. Plot of K, E, and D of the controllers in (40 m x 40 m) workspace area 100 50

K % ,chua K % ,Lornz K% , Arnold

0 1000

2000

3000

4000

5000 6000 Iterations "n"

7000

8000

9000

10000

3000

4000

5000 6000 Iterations "n"

7000

8000

9000

10000

3000

4000

5000 6000 Iterations "n"

7000

8000

9000

10000

100 50

E % ,chua E % ,Lornz E% , Arnold

0 1000

2000

10000 5000 0 1000

D ,chua D ,Lornz D, Arnold

2000

Fig. 14. Plot of K, E, and D of the controllers in (60 m x 60 m) workspace area how high the coverage area, the evenness index E, which reflects the degree of variation in covering the areas between species, and the total distance D of the generated trajectory. The effects of workspace size on the rate of convergence were studied. The results show that in low area coverage (20 m ´ 20 m), the performance of the Chua’s equation as a controller is slightly better than the others, but the performance of the Lornz’s equation as a controller is the best among the other in high area coverage (40 m ´ 40 m), and (60 m ´ 60 m). The performance of the Arnold equation as a controller is the worst among the other controllers especially in high area coverage. The total distance of the trajectory is increased semi-linear with the time depending on the linear velocity of the mobile robot and the assumed obstacles and boundary area. 12

Articles

The effects of the shape of the workspace, and to compare the results of described simulations with other algorithms – not based on chaos motion, will be considered in the future work. Ashraf Anwar Fahmy – Assistant Professor, Department of Computer Engineering, College of Computers and Information Technology, Taif University, Taif, Saudia Arabia. His research interests are mainly in the area of tracking system, control, and robotics. E-mails: ashraaf@tu.edu.sa, ashrafmanwar90@yahoo.com

References

K % ,chua K % ,Lornz K% , Arnold

2000

2012

Author

10000 5000

N° 2

[1] Y. Nakamura and A. Sekiguchi, „Chaotic mobile robot”, IEEE Transaction on Robotics and Automation, vol. 17, no. 6, 2001, pp. 898–904. [2] A. Sekiguchi and Y. Nakamura, “The Chaotic mobile robot”. In: Proc. IEEE/RSJ. Int. Conf. Intelligent Robots and Systems, vol. 1, 1999, pp. 172-178. [3] A. Jansri, K. Klomkarn, and P. Sooraksa, “Further investigation on trajectory of chaotic guiding signals for robotics system”. In: Proc. Int. Symp. Communication and Information Technology, 2004, pp. 1166-1170. [4] A. Jansri, K. Klomkarn, and P. Sooraksa, “On comparison of attractors for chaotic mobile robots”. In: Proc. 30th Annual Conf. IEEE Industrial Electronics Society, IECON, vol. 3, 2004, pp. 2536-2541. [5] C. Chanvech, K. Klomkarn, and P. Sooraksa, “Combined chaotic attractors mobile robots”. In: Proc. SICE-ICASE Int. Joint Conf., 2006, pp. 3079-3082. [6] L. S. Martins-Filho, R. F. Machado, R. Rocha, and V. S. Vale, “Commanding mobile robots with chaos”. In: ABCM Symposium Series in Mechatronics, J. C. Adamowski, E. H. Tamai, E. Villani, and P. E. Miyagi (Eds.), vol. 1, ABCM, Rio de Janeiro, Brazil, 2004, pp. 40-46. [7] S. Martins et al.”, Kinematic control of mobile robots to produce chaotic trajectories”, ABCM Symposium Series in Mechatronics, vol. 2, 2006, pp. 258-264. [8] S. Martins et al., “Patrol Mobile Robots and Chaotic Trajectories”. In: Mathematical Problems in Engineering, vol. 2007, Article ID61543, 13 pages, 2007. [9] J. Palacin, J. A. Salse, I. Valganon, and X. Clua, “Building a mobile robot for a floor-cleaning operation in domestic environments”, IEEE Transactions on Instrumentation and Measurement, vol. 53, no. 5, 2004, pp. 1418–1424. [10] Pecora, L. M., and Carroll, T. L., “Driving systems with chaoticsignals”, The American Physical Society, vol. 44, no. 4, 1991, pp. 2374-2384. [11] P. Sooraksa and K. Klomkarn, “No-CPU chaotic robots from classroom to commerce”. In: IEEE Circuits and Systems Mmagazine, 10.1109/MCAS, 2010, pp. 46-53. [12] J. Nicolas et al., “A comparative analysis of evenness index sensitivity”, Int. Review Hydrobiology, vol. 88, no. 1, 2003, pp. 3-15.

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Non-invasive Identification of Servo Drive Parameters Submitted 27th June 2011; accepted 20th September 2011

Reimund Neugebauer, Arvid Hellmich, Stefan Hofmann, Holger Schlegel

Abstract: For the tuning of servo controllers as well as for monitoring functions, significant parameters of the controlled system are required. In contrast to identification methods with determined input signals, the paper focuses on the problem of identification with regular process movements (non-invasive identification), leading to a lack of power density in some frequency ranges. A nonlinear Least Squares (LS) approach with single mass system and friction characteristic is investigated regarding the accomplishable accuracy and necessary constraints. The proposed method is applicable on industrial motion controllers and has been carried out with a multitude of input sequences. To verify the performance of the approach, achieved experimental results for the model parameters are exposed. Keywords: Identification, Parametric Model, Natural Excitation, Least Squares Method, SIMOTION

1. Introduction

With an increasing number of installed position controlled servo drives in production machines and machine tools, the importance of parameter identification grows as well. The optimal setting and adoption of controller parameters requires automatic strategies, which again need informative models. Consequently, a wide range of identification methods from control engineering was adjusted to the needs of drive control [1, 2]. This paper focuses on the field of non-invasive (i.e. not interfering with the process) parameter identification and explores the possibility of using regular movements of the servo drives, so called “natural excitations” [2]. Hence, a non-invasive, online capable approach is presented in the following chapter. In addition, experimental results with a variety of input signals are shown in section 3. Furthermore one possible approach for excitation detection is discussed in the 4th chapter. The paper is closed with some conclusions.

2. Non-invasive parameter identification in drive control

In contrast to identification with determined test signals, there is a need for the identification of plant parameters during regular operation of machine tools or production machines. Consequently, established identi-

fication methods have to be explored, whether they provide convenient results despite the insufficiency of only natural excitation. Another fact to consider is the online capability of the approaches for industrial motion controllers or numerical controllers. For the example of the velocity control loop of a cascaded position control (Fig. 1), [2] and [3] compare the applicability of extended Kalman Filters and Basis Function Networks. Furthermore, [4] presents an identification approach for mechanical systems, where a Least Squares method is used. Although it is combined with a Pseudo Random Binary Signal (PRBS) excitation, it seems promising to consider this method for non invasive identification as well. Generally it is not necessary to take high model orders into account for non-invasive identification after a model structure choice and complete parameter estimation during commissioning. In truth it is worthwhile to narrow the models down to significant parameters. This concurs with several publications, where only a single mass system (characterized by a total moment of inertia) and a friction model are taken into account for the chosen example [4, 5, 6]. According to [2] there is a limitation in the noninvasive identification quality for higher order models. Hence, a single mass system with a total moment of inertia (Jtot) and a friction characteristic, consisting of a constant friction moment (MRC) and a speed depending friction coefficient (µR) is chosen (Fig. 2).

Fig. 1. Cascaded position control with velocity loop (grey)

Fig. 2. Scheme of the velocity loop [9] with: ωcom… command velocity, ωact… actual velocity, Mcom… command torque, Mact… actual torque, Mfric… friction torque, CVC... velocity controller, GCuL... closed current loop Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Table 1. Variety of input sequences Input sequence

Controller setting GVC Nr.

KP [Nm s/rad]

TN [ms]

Rectangle

1,309

6,85

Stairs

0,8

Sinus

0,3

1 ⋅ ( M act − M fric ) J tot

1 = ⋅ ( M act − M RC ⋅ sign(ω act ) − µ R ⋅ ω act ) J tot

= b1 ⋅ [ M act (k − 1) − M RC ⋅ sign(nact (k − 1)) ]

with:

Offset

100 ms

250

2 x Magnitude

500

3 x Magnitude

(1)

With respect to the intended online capability of the identification procedure, the introduced parametric model will be combined with the LS method. For the realization, equation (1) has to be discretizised with the sample time Tsample. The resulting difference equation (2) with parameters (3) and (4) is further transformed to match the nomenclature of [1] (5). Notice that the parameters a1 and b1 still depend on Jtot and µR directly: nact (k ) + a1 ⋅ nact (k − 1)

Magnitude [1/ min]

300ms

Trapezoid

ω act =

Time period [ms]

Fig. 3. Test rig with known moment of inertia

(2)

− µ R ⋅Tsample ⋅60

a1 = −e b1 =

J tot ⋅2π

1 ⋅ (1 + a1 ) µR

y (k ) = − a1 ⋅ y (k − 1)

+b1 ⋅ [u (k − 1) − M RC ⋅ sign( y (k − 1)) ]

(3) (4) (5)

This leads to the description of the LS problem with a data vector ψ and a parameter vector Θ:

ψ T (k ) = [ − y (k − 1) u (k − 1) sign( y (k − 1)) ] ΘT = [ a1 b1 b1 ⋅ M RC ]

Fig. 4. Identification results for the moment of inertia Jtot with velocity controller setting 1

(6) (7)

With this definition, the LS problem is carried out recursively [1]. The calculation is feasible on industrial controllers due to the relatively small order of the resulting matrices (max. 3).

3. Identification Results

Experimental results were obtained on a test rig with known moment of inertia, which is equipped with an industrial motion controller SIMOTION with a sample time of 500 µs (Fig. 3). Despite the aspired non-invasiveness of the presented identification approach, test signals were used to investigate and develop the method. On the one hand, this is done to evaluate the performance of the identification technique and to locate minimum requirements, detached from a restricting case of application. On the other hand, 14

Articles

Fig. 5. Identification results for the moment of inertia Jtot with velocity controller setting 2

Journal of Automation, Mobile Robotics & Intelligent Systems

it allows a comparison of the accuracy of the approach with other invasive identification methods. The experimental results in this chapter are based on variety of input sequences (Table 1). The identification procedure is started in a phase of acceleration or deceleration of the drive, when a minimum value of torque (20% of the nominal torque) is present. Consequently, some experiments do not lead to any results due to the lack of excitation. Filtering of the input and output signals with the same filter helps to improve the accuracy [1, 2, 4]. Hence, the identification approach was extended by a moving average filter with the length of 5 elements. The main focus is the determination of the total moment of inertia Jtot, which is displayed in the Figures 4 and 5. It is noticeable, that the shape of the input sequence has the biggest influence to the accuracy, followed by the present velocity controller setting. Another fact is that all experiments with a velocity offset lead to discrepancies of less than five percent. The biggest divergence is about -10%, while 86% of the experiments provide results with a divergence less than ±5%, which is a typical value for invasive identification as well.

VOLUME 6,

N° 2

2012

Fig. 7. Power density spectrum of a rectangular input sequence

Fig. 8. Divergence of moment of inertia over power density area ASUU  1 N −τ  Φ uu (τ ) =  ⋅ ∑ u (k ) ⋅ u (k + τ )  for 1 ≤ τ ≤ N (8) N k =1   Fig. 6. Identification results for the friction parameters

Suu (k ) = FFT (Φ uu (τ ))

(9)

In the next step, the area under the power density spectrum is calculated (eq. 10) to represent the level of excitation (orange area in Figure 7).

The Results for the identification of friction parameters are displayed in figure 6. The presented curves result from the reproduction of the friction characteristic based on the estimated parameters MRC and µR. Notice, that the identification results for those values are only relevant for the specific velocity range of the process movement. For a better illustration, three experiments with rectangular input sequences and the introduced magnitudes 1-3 (Table 1) were combined in Figure 6. Additionally, a measured friction characteristic was included for verification. The identified characteristics match the measured curve in the linearization points, but vary towards lower or higher velocity values. A more precise image of the friction in the mechanical system could be issued by carrying out further movements in different velocity ranges.

The results are proportioned with the identified moment of inertia of equation 3 (red area in Figure 8). Identification results with a divergence of less then ± 10% are rated as sufficient (green area) and were achieved for rather high excitation. Hence, a minimum size of the excitation representative is about ASUU=200. Whether this statement can be generalized for other drives and thereby serve as excitation detection, future investigations will show.

4. Excitation detection

5. Conclusions

A second aspect in developing a non invasive identification procedure is the detection of suitable process excitation. The multitude of experimental results provides a basis for finding an adequate criterion. One promising approach is the analysis of the power density spectra Suu. They are computed by fourier transforming the autocorrelation Φuu of each input sequence signal (with the length N) according to equations 8 and 9 [4]. One example is displayed in Figure 7 whereat the discrete vector Suu is converted and plotted over frequencies.

N /2

ASuu = ∑ Suu (k )

(10)

k =1

An identification method for the estimation of an order reduced parametric model was presented. The approach is online capable on industrial motion controllers and has been implemented as an automatic application. Experimental results for a variety of input sequences were presented to prove the independency of specific test signals and thus the suitability for non invasive identification. Furthermore one possibility for excitation detection was introduced and carried our for the available test rig. Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

Acknowledgements

The Cluster of Excellence „Energy-Efficient Product and Process Innovation in Production Engineering“ (eniPROD®) is funded by the European Union (European Regional Development Fund) and the Free State of Saxony.

Authors

Professor Reimund Neugebauer – head of the “Institute for Machine Tools and Production Processes”, Chemnitz University of Technology, Faculty of Mechanical Engineering, Institute for Machine Tools and Production Processes, Reichenhainer Str. 70, 09126 Chemnitz, Germany; wzm@mb.tu-chemnitz.de) Arvid Hellmich* – Chemnitz University of Technology, Faculty of Mechanical Engineering, Institute for Machine Tools and Production Processes, Reichenhainer Str. 70, 09126 Chemnitz, Germany; arvid.hellmich@mb.tu-chemnitz.de Stefan Hofmann – Chemnitz University of Technology, Faculty of Mechanical Engineering, Institute for Machine Tools and Production Processes, Reichenhainer Str. 70, 09126 Chemnitz, Germany; stefan.hofmann@mb.tu-chemnitz.de Dr. Holger Schlegel – head of the division “Control and Feedback Control Technology“, Faculty of Mechanical Engineering, Chemnitz University of Technology, Reichenhainer Str. 70, 09126 Chemnitz, Germany; holger. schlegel@mb.tu-chemnitz.de * Corresponding Author

References

[1] R. Isermann, “Identifikation dynamischer Systeme 1“, Springer, 1992. [2] S. Beineke. “Online-Schätzung von mechanischen Parametern, Kennlinien und Zustandsgrößen geregelter elektrischer Antriebe“, VDI Fortschritt-Berichte 816, 2000. (in German) [3] S. Beineke et. al., “Comparison of parameter Identification Schemes for Self-Commissioning Drive Control of Nonlinear Two-Mass Systems”. In: IEEE Industry Applications Society, Annual Meeting, New Orleans, 1997, pp. 493-500. [4] F. Schütte, “Automatisierte Reglerinbetriebnahme für elektrische Antriebe mit schwingungsfähiger Mechanik”, Dissertation Thesis, Shaker Verlag, 2003.

Articles

VOLUME 6,

N° 2

2012

[5] H. Wertz, S. Beineke, N. Fröhleke, “Computer aided commissioning of speed and position control for electrical drives with identification of mechanical load”. In: IEEE Industry Applications Conference 1999. [6] F. Mink, A. Bähr, S. Beineke, “Self-commissioning feedforward control for industrial servo drives”, Elektromotion 2009. [7] A. O’Dwyer, Handbook of PI and PID Controller Tuning Rules, Imperial College Press, 2003. [8] K. J. Åström, T. Hägglund, Advanced PID Control, ISA, Research Triangle Park, 2006 [9] R. Neugebauer et. al, “Identification of Parametric Models and Controller Design for Cascaded Position Control”. In: Proceedings of the 18th IFAC World Congress, Milan, 2011.

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Human-machine Interface for Presentation Robot Submitted 27th June 2011; accepted 27th September 2011

Jiri Krejsa, Vit Ondrousek

Abstract: The purpose of mobile presentation robot is to interact with the users providing the required information. This paper presents the approach used in the highest software layer of autonomous mobile robot Advee used commercially for presentation purposes. The requirements for the layer are discussed together with available means of the robot. The particular building blocks and overall structure of the module are shown and key features are described in detail together with the behavior definition. Given module serves successfully when exploited to variable environment represented by people with computer literacy of great variation as proved during module verification tests. Keywords: human-robot interface, mobile robot, presentation robot

1. Introduction

With mobile robots appearing nowadays outside research laboratories more than ever, human-robot interface (HRI) related issues attract great attention [1, 2]. Most of the research is focused on improving certain means of communication, especially the voice dialog (sound used both as inputs – natural language understanding and outputs – robot voice) [3, 4] and utilization of computer vision (human face recognition, face related higher features recognition, gesture recognition, etc.). Various means of communication can be combined to increase the level of interaction. In [5] the speech recognition is combined with user localization using two processed microphones signals to obtain the location of the source, that can be further utilized in the response of the robot. Face recognition is accompanied with gesture recognition in [6] bringing wider variety of possible inputs to the robot. Work of Perzanowski [7] is an example of multimodal HRI, combining speech and gesture recognition with external input from Personal Digital Assistant (PDA). This paper is focused on HRI of presentation robot Advee, whose purpose is to serve as an autonomous mobile source of information, transferable to people of different computer literacy. Such requirement brings the necessity of combining all robot means to get the redundancy in communication channels, so the less computer literate people can still get the message while more literate user is not repelled. The paper is organized as follows. After short introduction of the robot itself the requirements for HRI layer are summarized together with available means (HRI re-

Fig. 1. Presentation robot Advee lated inputs and outputs). Proposed structure of the system is discussed with attention on key modules. The verification made during the tests together with discussion conclude the paper.

2. Presentation robot

The purpose of presentation robot is to present certain information to the users in interactive manner. Further paragraphs describe the design of human-robot interface module for robot Advee. As the HRI is only a part of the robot system, here is a short introduction of the robot itself. Presentation robot Advee is 160 cm high, 80 kg heavy wheeled robot with Ackerman steering. It has rear wheels driven with single DC actuator through mechanical differential. Robot is powered with 8cell LiFePo4 accumulator giving it approximately 8 hours operational range. The localization of the robot is handled by fusion of robot motion controllers commands, odometry readings and scanner of infrared beacons placed in known fixed locations. Fusion is performed by extended Kalman filter [8]. The software operated on the robot can be divided into three layers: · low level provides interaction with hardware devices · middle level implements robot state estimation and path planning · high level handles the user interaction Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

The robot is equipped with a wide range of sensors and outputs, serving as the primary resources for HRI layer. The particular sources are described in detail below. The size of the robot and access to its main communication devices is illustrated in Fig. 1.

3. HRI module

VOLUME 6,

In order to design the HRI module properly the requirements must be stated first. The overall goal is to create the interface that is friendly and universal (can be used by variety of users), while the underlying engine is robust, modular and flexible. The requirements are listed in order of importance in Tab. 1. Tab.1. HRI requirements Requirement

Description

redundancy

information should be transferred to user in all possible ways in parallel (e.g. both visual and sound)

robustness

HRI module must be capable of operation even when minor hardware failure occur (e.g. camera fails)

flexibility

simple addition of new feature, sensor, etc

parametrization

overall behavior can be easily changed, certain features can be disabled/enabled upon request

adaptability

HRI should adapt to the user abilities, e.g. longer timeouts for slower users, etc.

modularity

several programmers can work independently on the project, particular modules can be tested independently

3.2. Robot means

Based on the requirements, what are the means of the robot? We can divide the means into two basic categories: inputs and outputs, where inputs represent the information from the user to the robot and outputs represent

Tab. 2. Robot HRI means Details

Inputs Screen capacitive touch screen

robust touchscreen returning touch coordinates

Voice microphone, soundcard

incoming sound can be recorded and further processed

Vision CCD camera

instant flow of images further processed (face detection, etc)

Lower level data from lower level of robot control bumpers, proximity (distances from obstacles, current locasensors, odometry, tion, etc.) Outputs Screen LCD screen

hi resolution screen to present visual information

Voice soundcard, amp, speakers

modulated voice, currently prerecorded

Print printer

thermal printers up to 112mm wide

Motion motion actuators

safe motion in given area, towards given goal

3.3. HRI building blocks

The key idea in designing the HRI module is to separate the interaction into independent blocks, that are sequentially activated by main HRI internal engine. Only single block is active at the time. The overall behavior of the robot can then be defined in a single routine, that controls blocks activation. Each block uses different means of the robot, however in most cases the majority of the means are used.

Vision

Sources Sound input

Sound output

LCD touch screen

Motion Proximity sensors

Fig. 2. Robot means physical location 18

Articles

SrcScreen SrcVision

Printer output

2012

robot reaction. Some of the means are bidirectional, e.g. the touch screen acts both as input and output. The overview of Advee’s means is shown in Tab. 2. and physical location of the means on the robot is shown in Fig. 2.

Type / device

3.1. Module requirements

N° 2

SrcSoundIn SrcSoundOut

Blocks

Motion Catcher Video Selector Active block

SrcPrinter

Map

SrcLowLevel

Games

…

Fig. 3. HRI building blocks

Journal of Automation, Mobile Robotics & Intelligent Systems

To enable access to the means, the particular means are encapsulated into so called sources. Programmatically the blocks and sources are represented by certain classes. Each block is inherited from BaseBlock class and each source is inherited from BaseSource class. BaseBlock contains links to all the sources instances, therefore any subsequent block inherited can access all the sources. As the sources can serve to single block only, the mechanism that assigns and releases sources is implemented. Sources are assigned prior to block activation and released once the block is finished. Currently implemented sources and blocks are shown in Fig. 3, with an example of Map as active block having access to all the sources. The sources directly depend on the means of the robot from previous chapter. The sources encapsulate the access methods and events generated by corresponding devices. The key sources are: Screen: shows the graphics and generates events when touch is detected. Vision: processes the camera images with cascade of boosted classifiers based on Haar-like features for frontal face detection. SoundIn: processes the incoming sound to detect simple patterns (no speech recognition yet) SoundOut: plays prerecorded sound files of modulated voice. As the sounds are prerecorded, the variety of sounds is available for given type of the sound to avoid repeating the same phrases. Printer: thermal printer with 112 mm wide output. LowLevel: encapsulates the methods and events of lower level software layers. In particular the motion of the robot, its estimated position (given by Extended Kalman Filter based localization technique), the motion planner outputs, proximity sensors calls, hardware monitor calls (battery status, temperatures, etc.), etc. Each block represents single operation in human-robot interaction. The key blocks are: Motion – block that is active when the robot is moving and not interacting with people. Motion can end for a number of reasons: people are detected in surrounding (combination of proximity sensors analysis, face detection of the camera images, etc), somebody touches the screen, etc. Catcher – block that verifies whether there is a person the robot can talk to. The block serves as decision maker for uncertain situations (low confidence in face detection, etc) Selector – block that serves as the menu allowing user to select from several options, usually activating another block. The options are shown graphically as buttons on the screen with both pictograms and text. The selection is accompanied with voice explanation of the options. Custom blocks – blocks responsible for certain feature of the data to present. The names are selfexplanatory, for example: Map shows interactive map of the surroundings with indicated position of the robot in the map. Video runs the video, interruption is allowed Games enables users to play games, usually followed with the prints of reward in the case of victory in the game. Print handles printing, guiding the user through

VOLUME 6,

N° 2

2012

Select block Assign sources Activate block processing Release sources no

Emergency yes Handle emergency Fig. 4. HRI engine routine

the print process, checking whether user withdrawn the print, etc. Within each block a finite state machine is implemented, handling the behavior of the block, in particular the responses to user inputs (screen taps, incoming sounds, etc.) and internal events (end of the output sound, changes in lower level data, etc).

3.4. HRI behavior definition

Whole behavior of the robot is determined by the sequence of blocks. The sequence depends partially on user and partially on behavior definition. HRI engine works in simple cycle shown in Fig. 4. Once the appropriate block is selected, all sources are assigned to the block, block is activated and starts to operate. Within the block usually finite state machine is implemented, managing the incoming events and producing required outputs.

Leave Motion

Selector Advee

Catcher

Map Schedule

Selector Main

Selector Videos

Print Custom

Quiz Selector Games

Selector Language

 Lang1  Lang2 …

Video Advee

Print About

 Video1  Video2 … Print Victory

 Game1  Game2  …

Print Victory

Fig. 5. Behavior signal flow Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

Once the block is done the sources are released and next block is selected. Each block can finish in a regular manner, or once an emergency event occurs (HW failure, etc.). In such a case the emergencies are handled separately before the main cycle is entered again. The behavior can be easily visualized with a flow chart, an example of the signal flow is shown in Fig. 5. The flow can be arbitrarily modified, certain features (blocks or block sequences) omitted or added.

3.5. Implementation details

High level layer of the robot software runs on Windows 7 OS. The HRI module is implemented in C# using Microsoft Visual Studio environment. Computationally demanding routines (e.g. image processing) are written in C and C++. The predecessor to all sources classes – BaseSource class implements virtual methods for essential operations with the sources (startup and cleanup). The predecessor to all block classes – BaseBlock class implements virtual event responses for all possible sources events, therefore any descendant block can respond to such events. Virtual methods for block activation and deactivation are also implemented in BaseBlock class, taking care of assigning the sources to the currently active block and releasing the sources when block is finished. The termination of block processing is announced with BlockDone virtual event, overridden in particular blocks. Lower layers of the software run on separate computer with Linux OS. Communication with the lower layers (position estimation, motion, path planning, etc.) is based on Lightweight Communication and Marshalling (LCM) library [9] and is fully encapsulated into SrcLowLevel source.

4. Verification

The HRI was initially verified using released version of HRI together with simulator of the lower layers of robot. During the tests the HRI was presented mainly to students and basic settings were tuned, such as timeouts for particular blocks, understandability of the graphics, etc. Once the basic setup was done the tests were performed using real robot on a number of people from various backgrounds and data in the form of questionnaires

VOLUME 6,

N° 2

2012

were collected afterwards together with the camera recordings. Based on the collected data the HRI was further modified, with the emphasis on redundancy of given information, e.g. all voice outputs (talk of the robot) are accompanied with the text on the screen of the same meaning. Further tests lead to minor modifications in rephrasing the speech output and changing the voice modulation for further clarification. During the use of the robot all the user inputs are logged and collected data are used for further analysis and improvements. In order to determine whether the HRI module works according to requirements, several quantities can be extracted from data collected during the interactions. Such variables values obtained on several events with different types of users can be compared giving indication of HRI quality (ideally the quantities should not vary). While there could be numerous number of such indicators with respect to certain features (e.g. time spent with the map, win rate in games, etc.) only several give general impression on success of the HRI. Those indicators are described below. ATI – average time of interaction with single user [sec]: interval between the start of the interaction and its end (user left the robot). Longer ATI means single user spent more time with the robot, it is only indirect indicator of how well the interaction went. CSR – user catching success rate [-]: rate between successful catch of user (interaction starts) and failure (user walks away). Lower CSR means that people are afraid of the robot, do not know how to start an interaction, etc. TIP – interaction timeout percentage [%]: percentage of interactions ended with timeout in arbitrary block other than main menu. Higher percentage means that user left the robot during interaction. PPE – person per event [-]: total number of interactions divided by total number of users. If the PPE is higher than 1 then some users used the robot repeatedly. BPU – blocks per user [-]: number of blocks activated per user in single interaction. Higher BPU means that user exploited more options during interaction. Data collected on several events with different audience are listed in Tab. 3. The particular events were: E1 – VIP audience E2 – general audience E3 – university students E4 – seniors Tab. 3. HRI indicators results

Fig. 6. HRI in action 20

Articles

ATI

48±11

73±19

85±21

96±71

CSR

0.89

0.75

0.97

0.18

TIP

PPE

1.12

1.51

1.05

BPU

5.2±1.2

7.3±2.1

8.7±2.4

3.2±1.9

Results clearly show that the type of the users does have an impact on given indicators values, however in none of the events the HRI failed. Very low user catching

Journal of Automation, Mobile Robotics & Intelligent Systems

success rate in event 4 deserves further analysis. From the questionnaires given to the users we can state that the reason for the user to avoid the robot was not the HRI (users knew what to do), but general reluctance towards the robot itself.

5. Conclusion and future work

Described human-robot interface communication module was successfully designed and implemented in mobile robot Advee. HRI is now fully operational and used in commercial application of the robot with over 200 hours of operation in interaction mode with a variety of audience. While HRI exhibits no problems, the parametrization and flexibility need improvements, that can probably be achieved only through the design of special script language for behavior description. Currently ongoing research is focused on extending the recognition capabilities (age, gender) of image processing unit. The gesture recognition was rejected as the cost of explaining how to use the gestures to the user overcomes the potential benefits.

VOLUME 6,

[6] [7] [8]

[9]

N° 2

2012

based on noise separation device”. In: RO-MAN 2009, pp. 328-333. M. Chang, J. Chou, “A friendly and intelligent human-robot interface system based on human face and hand gesture”. In: AIM 2009, pp.1856-1861. D Perzanowski, A.C. Schultz et al., “Building a multimodal human-robot interface”, IEEE Intelligent Systems, vol. 16, no. 1, 2001, pp. 16-21. J. Krejsa, S. Věchet, “Odometry-free mobile robot localization using bearing only beacons”. In: Proceedings of EPE-PEMC 2010 Conference, Macedonia, pp. T5/40-T5/45. A.S. Huang, E. Olson et al., “LCM: Lightweight Communications and Marshalling”. In: IROS 2010, Taipei, Taiwan, pp. 4057-4062.

Acknowledgements

Published results were acquired with the support of the CAS under the research plan AV0Z20760514 and with the support of the OPEC, project number CZ.1.07/2.3.00/09.0162. Authors would like to thank Petr Schreiber from Bender Robotics for his work on HRI coding.

Authors

Jiri Krejsa* – Institute of Thermomechanics AS CR v.v.i., Brno branch, Technická 2, Brno, 616 69, Czech Republic, krejsa@fme.vutbr.cz. Vit Ondrousek – Brno University of Technology, Faculty of Mechanical Engineering, Brno, 616 69, Czech Republic, ondrousek@fme.vutbr.cz *Corresponding author

References

[1] M. W. Kadous, R. K.-M. Sheh, C. Sammut, “Eﬀective user interface design for rescue robotics”. In: ACM Conference on Human-Robot Interaction, Salt Lake City, UT, USA: ACM Press, 2006. [2] K. Dautenhahn, M. Walters et al., “How may I serve you? A robot companion approaching a seated person in a helping context”. In: ACM Conference on Human-Robot Interaction (HRI), Salt Lake City, UT, USA: ACM Press, 2006 [3] M. Skubic, D. Perzanowski et al., “Spatial language for human-robot dialogs”, IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, vol. 34, issue 2, 2004, pp. 154167. [4] V. Kulyukin “Human-Robot Interaction Through Gesture-Free Spoken Dialogue”, Autonomous Robots, vol. 16, no. 3, 2004, pp. 239-257. [5] K. Park, S. Lee et al., “Human-robot interface using robust speech recognition and user localization Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Novel Developments in Dimensional Nanometrology in the Context of Geometrical Product Specifications and Verification (GPS) Submitted 10th October 2011; accepted 12th November 2011

M. Numan Durakbasa, P. Herbert Osanna, Gökcen Bas, Pınar Demircioglu, Mehmet Cakmakci, Adriana Hornikova

Abstract:

Adequate knowledge in the areas of intelligent coordinate metrology and design are important presuppositions to achieve waste free production and low costs of manufacturing with higher quality and accuracy at the same time. This is of extreme importance in present time of worldwide international competition in industry and production and at the same time increasingly higher costs of energy and raw material. The prescription and consumption of material and energy to achieve the necessary and required workpiece accuracy in series manufacturing depends to a great extent from the (geometrical) workpiece tolerances of any kind (roughness, form, positional, dimensional) which are prescribed for the production and the fulfillment of these tolerances and therefore for the function of the produced workpieces and their fitness for practical application and none the less of the economy of production altogether. This requirement is of great importance at the time being which is characterized as described above. Keywords: precision metrology, design, manufacturing, workpiece tolerancing

1. Introductory Remarks

If the workpiece geometry of machined parts is considered as a whole there exist interactions between the different features forming the periphery of the part. But also within the surface of every single feature there exist interactions between geometrical deviations of different kind and different order. If we take these deviations of dimensions, roughness, form and position collectively the existing interactions are significant for the accuracy, high quality and the functions of the parts that should be accomplished during practical application. The most important parameters in determining the suitability of a technical part are its compatibility, functionality, performance and corrosion resistance. The precise assessment of wear, friction and miniaturization demands creation of nanometer scaled surface structures, surfaces with thin film deposition and ultra precision surface treatment with the utilization of new manufacturing and measurement instrumentation and techniques. These include micro and nanofabrication of surface patterns and topographies by the use of laser machining, photolithographic techniques, and electron beam and colloidal lithography to produce controlled structures on technical surfaces in size ranging from 10 nm to 100 µm. At the time being 3D surface measurement is already proved to be an important tool 22

Articles

in several areas of surface analysis including wear, indentation, topography, contact problems and functional behavior of surfaces (see Figure 1).

Fig. 1. 3-D observation of a machined workpiece surface using a digital microscope [1] The needs of the industry for ultra-high precision engineering and workpieces with a surface roughness less than few nanometers call for measurement instrumentation that can be applied reliably in modern production engineering, together with international standards defining parameters and tolerances in the nanometer scale. The requirements on the measurement systems and the measurement strategy to determine suitable parameters, time, costing and the guarantee of a predetermined process stability by means of measurable and correlated parameters come into focus.

2. Intelligent Design and Advanced Measurement Techniques

As the tolerances of workpieces and their features decrease, the interaction and correlation between dimensional tolerances and surface finish become more important [2, 3]. To achieve surface finishes and part tolerances in the sub-micrometer and nanometer level, it is necessary to incorporate very sophisticated instrumentation and metrology into the design [4]. In the same period the standards governing product design and manufacturing have undergone basic international harmonisation. Focal points of interest included; workpiece microgeometry [5] and geometrical deviations [6], as well as tolerancing principles such as; Independency principle, Duality principle, Functional control principle, Feature principle, General specification principle, Rigid work-

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

and

piece principle, Responsibility principle according to in-ternational GPS Standard [7], which applies to the interpretation of GPS indications on all types of drawings. In many countries, the above-mentioned international standards have been adopted also on a national level, similar as the new international standards about quality assurance, quality management and environmental management [8, 9, 10]. In this respect the general term “Geometrical Product Specifications and Verification – GPS” has become recently well-known for the area of mechanical engineering. It defines on a technical drawing the shape (geometry), dimensions and surface characteristics of the workpiece under discussion. In this way the optimal function of the respective part is supposed to be guaranteed considering certain manufacturing tolerances. Nevertheless workpieces will be produced, which do not fulfill these requirements. Therefore workpieces are measured in order to compare them with the specifications. There is a need to relate between actual workpieces and: • the workpiece imagined by the designer, • the workpiece as manufactured, • the knowledge about the workpiece as measured. In order to establish this relationship between design, production and measurement and to clarify the mutual importance, standards have been developed in the area of GPS. Comprehensive knowledge in this area is an important presupposition to achieve economic design, construction, production, metrology and quality management. The concept of the GPS includes: – several types of standards, some are dealing with the fundamental rules of specification, some are dealing with global principles and definitions and some of them are dealing directly with the geometric characteristics; – different geometric characteristics such as size, distance, angle, form, location, orientation, roughness; – workpiece characteristics as results of different manufacturing processes and the characteristics of specific machine elements

- occurs at several steps of the product life cycle, in the

development of a product, design, manufacturing, metrology, quality assurance, etc. This concept is represented in Figure 2, showing four different group of GPS standards and designated as the “GPS-matrix-model” (Figure 2) [14].

3. Precision Metrology in Modern Production Environment

An important development as far as workpiece metrology is concerned is the big general advance of coordinate metrology which also happened in the same period of time as computer-aided metrology and “GPS” in general. Three dimensional coordinate measuring machines (3D-CMMs) allow to measure deviations of dimensions, form and position very accurately with only one measuring device [15]. Besides measuring accuracy the number of workpieces to be measured is important when choosing the measuring device. Especially when workpiece tolerances are more accurate than tolerance grade IT5 (e.g. 11 µm for 50 mm) it is necessary to make use of coordinate metrology. This is also possible for big series of workpieces. CMMs are referred to as those measuring instruments giving physical representations of a three dimensional rectilinear Cartesian coordinate system. – The nature of coordinate metrology can be defined as: – The geometrical features of the workpiece to be measured are touched in several measuring points using a coordinate measuring device. – The coordinates of the measuring points are used to compute the mathematical geometry of the workpiece with help of the control computer of the CMM. At the time being coordinate metrology is a very important tool to solve problems of production metrology of nearly any kind especially when high flexibility and high accuracy are demanded at same type of workpiece. One of the essential requirements in coordinate metrology is the computation of associated features from the

The Global GPS standards

General GPS Matrix General GPS chains of standards

The Fundamental GPS Standards

The Size chain of standards The Distance chain of standards The Radius chain of standards The Angle chain of standards The Form of a line The Form of a surface The Orientation chain of standards The Location chain of standards

The Circular run-out chain of standards The Total run-out chain of standards The Datum chain of standards The Roughness chain of standards The Waviness profile chain of standards The Primary profile chain of standards The Surface defects chain of standards The Edges chain of standards

Complementary GPS Matrix Complementary GPS chains of standards Process specific tolerance standards Machine element geometry standards

Fig. 2. The GPS-matrix-model – GPS Masterplan – Overview Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

probed contact points. Besides coordinate metrology modern optoelectronic methods are important measurement tools in computer integrated production plants and also as basic tools for global quality management and quality assurance activities. Their efficient use and correct calibration are crucial requirements for quality management in this environment. Presently exists the general development from micro technology to Nanotechnology. Nano technology describes new innovative manufacturing technologies, finishes, tolerances and especially measurement technique in the nanometer range [11] which especially is called Nanometrology. In pursuance of this aim since about 1982 new high resolution and high precision measuring devices have been developed, especially Scanning Tunneling Microscopy (STM) and Atomic Force or Scanning Probe Microscopy [12, 13]. For highest demands these methods make it possible to explore miniature structures and in general very accurate and small industrially produced parts and structures. With these measuring instruments lateral resolutions up to 1 nm are achieved and in perpendicular direction it is possible to achieve atomic resolution. Actual developments give evidence that in the near future it will be possible that such instruments will be used in high level industrial quality control laboratories especially also in advanced SMEs. Figure 3 shows the measurement data of the surface structure of a workpiece surface after precision machining.

VOLUME 6,

N° 2

2012

– evaluation of uncertainty, etc. Generally there are many definitions and concepts in Geometrical Product Specifications and Verification (GPS) but one of them which has been presented some years ago named “Skin-Model” [15] was studied in the frame of some research projects. The “Skin-Model” presents a new description for Geometrical Product Specifications and Verification (GPS) with its associated details and on its basis every workpiece can be geometrically defined and considered by applying manipulations of the workpiece geometry. This determination is based on mathematical rules and definitions. It means that according to this determination every workpiece can be designed and on the other hand according to the design it can be measured very clearly [16]. The Skin-Model is based on some general and basic definitions and it uses some tools which are named “Operations” which can be compared with mathematical operations as in mathematics and especially in arithmetic (Figure 4).

Ideal features Non-ideal features Fig. 4. Ideal model and “Skin-Model” of a workpiece By application of appropriate evaluation software the measuring results can be transformed into suitable data format that can be used for further calculation and study. In the next step the measuring results will be evaluated with applying computer programs, which are available for statistical evaluations. Now the quality control department will be able to analyse this phase easily and can give necessary advice or notice to relevant other departments especially in the developing and planning domain. This gives principal ideas in respect of the evaluation of measurement results with regard to GPS.

Fig.3. High accuracy 3D topography measurement using the AFM after surface finishing

4. Workpiece Quality in the Context of GPS

In general it is the intention of the manufacturer to satisfy all requirements that are demanded for a product. Every manufacturer is looking forward to satisfy all the criteria that its product must have. Technical means, tools and methods are used to ensure the consistency of product characteristics. One of its important characteristic or feature is geometrical specification or it is better to say Geometrical Product Specifications (GPS). Geometrical Product Specifications are a means to transform function dependent demands into produced workpieces and parts based on: – mathematical rules and methods, – consideration of macro and micro geometry, – possibilities for measuring of quantities and especially toleranced quantities and 24

Articles

5. Conclusion and Outlook to Future Developments

The ideas presented in this publication explain in principal the correlation between different geometrical deviations and the manufacturing conditions. This can help to achieve lower manufacturing costs and at the same time higher quality, accuracy and efficiency in present production. At the atomic level metrology and fabrication are closely related. As a still more futuristic development this may perhaps make possible the design and production of miniature measurement instruments or devices for medical treatment or operations in human beings that might operate autonomously in the micro- or even nanoworld. The speed and reliability that can be achieved make any idea of mass manufacturing, now, or in the foreseeable future, preposterous. But in any case nanometrology has become technical reality and pico- and femtometrology will not be impossible in the future.

Journal of Automation, Mobile Robotics & Intelligent Systems

The results of the presented study can be seen as a further step in the direction of a comprehensive analysis of workpiece geometry and it is fully in line with research work already carried out in the past [17]. By the described successful application of intelligent tolerancing and metrology for the solution of measurement problems of various kinds also new challenges are put onto precision production measurement technology especially in the area of GPS.

Authors Numan M. Durakbasa – Vienna University of Technology (TU-Wien), Head of Department of Interchangeable Manufacturing and Industrial Metrology (Austauschbau und Messtechnik/Produktionsmesstechnik & Qualität), Institute of Production Engineering and Laser Technology, Vienna University of Technology (TU-Wien), Karlspl. 13/3113, 1040 Vienna, Austria, durakbasa@mail.ift.tuwien.ac.at P. Herbert Osanna – Vienna University of Technology (TU-Wien), Department of Interchangeable Manufacturing and Industrial Metrology (Austauschbau und Messtechnik/Produktionsmesstechnik & Qualität), Institute of Production Engineering and Laser Technology, Vienna University of Technology (TU-Wien), Karlspl. 13/3113, 1040 Vienna, Austria, osanna@mail.ift.tuwien.ac.at MSc Gökcen Bas – Vienna University of Technology (TU-Wien), Department of Interchangeable Manufacturing and Industrial Metrology (Austauschbau und Messtechnik/Produktionsmesstechnik & Qualität), Institute of Production Engineering and Laser Technology, Vienna University of Technology (TU-Wien), Karlspl. 13/3113,1040 Vienna, Austria, goekcen.bas@mail.ift. tuwien.ac.at Dr Pinar Demircioglu* – Adnan Menderes University,Faculty of Engineering, Department of Mechanical Engineering, Adnan Menderes University 09010, Aydin,Turkey, pinar.demircioglu@adu.edu.tr Assoc. Prof. Dr. Mehmet Cakmakci – Dokuz EylulUniversity, Engineering Faculty, Industrial EngineeringDepartment, Izmir, Turkey, mehmet. cakmakci@deu.edu.tr Dr Adriana Hornikova – University of Economics in Bratislava, Faculty of Informatics Economics, Dept. Statistics, University of Economics in Bratislava, Bratislava, Slovakia, adriana.hornikova@euba.sk *Corresponding author

References

[1] Digital Microscope Keyence VHX-600, http:// www.digitalmicroscope.com/solutions/digital-microscope.php. [2] Osanna P. H., Durakbasa M.N., Kräuter L., “Industrial Metrology and Interchangeable Manufacturing under the Viewpoint of Nanotechnology and

VOLUME 6,

N° 2

2012

Nanometrology”, Bulgarian Academy of Sciences, Problems of Engineering Cybernetics and Robotics, vol. 59, 2008, pp.60-73. [3] Tabenkin A., “Effects of Form and Finish on Tolerances”, Quality, vol. 9, 1993. [4] D. Whitehouse, “Comparison between stylus and optical methods for measuring surfaces”, Annals CIRP, vol. 37, no. 2 , 1988, pp. 649-653. [5] EN ISO 4287:2009, Geometrical Product Specifications (GPS) – Surface texture: Profile method – Terms, definitions and surface texture parameters (ISO 4287:1997 + Cor 1:1998 + Cor 2:2005 + Amd 1:2009). [6] EN ISO 1101:2006: Geometrical Product Specifications (GPS) – Geometrical Tolerancing – Tolerances of Form, Orientation, Location and Run-out; EN ISO 1101/A1:2010 Geometrical Product Specifications (GPS) – Geometrical tolerancing –Tolerances of form, orientation, location and run-outAmendment 1: Representation of specifications in the form of a 3D model. [7] EN ISO 8015:2011 – Geometrical product specifications (GPS) – Fundamentals – Concepts, principles and rules. [8] EN ISO 9001: 2008: Quality Management Systems – Requirements. 2008-12. [9] EN ISO 14001:2009: Environmental management systems – Requirements with guidance for use (ISO 14001:2004 + Cor. 1:2009). [10] EN ISO 9004:2009: Managing for the sustained success of an organization – A quality management approach. [11] Taniguchi N., “On the Basic Concept of Nanotechnology”. In: Proc. Int. Conf. Prod. Eng., Tokyo, JSPE, 1974, part 2, pp. 18-23. [12] Binnig G., H. Rohrer, “The Scanning Tunnelling Microscope”, Sci. Amer., no. 253, 1985, pp. 40-46. [13] Binnig G., H. Rohrer, “Scanning Tunnelling Microscopy”, IBM J. Res. Develop., vol. 30, 1986, pp. 355-369. [14] ISO TR 14638: Geometrical product specification (GPS) – Masterplan, 1995. [15] Ballu A., Mathieu L., “Univocal Expression of Functional and Geometrical Tolerances for Design, Manufacturing and Inspection”. In: 4th CIRP Seminar on Computer Aided Tolerancing, Tokyo, Japan, 1995, pp. 31-46. [16] N.M. Durakbasa, A. Sadat, A. Nomak, “Dimensional and Geometrical Measurement and Interpretation of Measuring Results on the Basis of the Skin Model”, Measurement Science Review, vol. 1, no. 1, 2001, pp. 89-92. [17] P.H. Osanna, N.M. Durakbasa et al.: “Global Competitive Manufacturing on the Basis of Intelligent Metrology and Quality Management as Important Tools”. In: Proceedings of the International Symposium Tools and Methods of Competitive Engineering – TMCE 2002, Ed.: I. Horvath et al.,Wuhan, China, April 2002, ISBN 7-5609-2682-7, pp. 837-844.

Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Simultaneous Measurement of Two Parameters by Double Current Supplied Bridge Submitted 14th December 2010; accepted 22nd March 2011

Adam Idzkowski, Jaroslaw Makal, Zygmunt L. Warsza

Abstract:

A four-terminal (4T) bridge-circuit, with two voltage outputs is presented. This circuit, named as double current bridge is supplied by two equal current sources connected in parallel to opposite arms (2J) or by one such source switched between those arms (2x1J). Two output voltages from bridge diagonals as functions of arm resistance increments in absolute and in relative units are given. Also the example of this bridge application is proposed. Signal conditioning formulas of two-parameter measurement on the example of strain and temperature are discussed in detail. Some results obtained with the use of this bridge-circuit are briefly described. Keywords: multivariable measurement circuits, bridge circuit, strain measurement

1. Introduction

In the most applications the newest analog-to-digital (ADC) converters and digital parts of measurement systems assure satisfying resolution, speed and universality due to programming facilities. At present, an improvement of strain, pressure, force, torque or other measurements depends mainly on metrological properties of the input analog part of these systems. The sensor’s thermal error (drift of sensor’s offset and span) is compensated in the digital part of a conditioner by proper correction algorithms. For pressure measurement a piezoresistive sensor can be powered by an adjustable current source combined with a programmable-gain amplifier and external trimmable resistors (e.g. MAX1450) [1], or two amplifiers and two digitally controlled potentiometers [2], or four digital-to-analog converters (DAC) resulting in a temperature-depended bridge voltage (e.g. MAX1452) [3]. In a mass production of silicon piezoresistive-bridge pressure sensors, sensor-error correction is often affected by use of a laser or abrasive trimming machine, which trims resistors and thermistors in the signal conditioning circuit to the values required for offset and sensitivity compensation (e.g. in X-ducer piezoresistive pressure bridge-sensors [4], or NPC series of GE Novasensor pressure sensors [5]). Apart from well-known instrumentation for the measurement of single variables, the development of methods of continuous indirect multivariable measurement is urgently needed. High accuracy measurement of increments of input immittances of multi-terminal circuit and of some quantities affecting them is an example. Relevant problems have been considered on the example 26

Articles

of two-parameter simultaneous measurement of resistance increments of the four-terminal (4T) circuit [6]. Since 1998 Warsza has been proposing two types of structures for such measurement and for the primary signal conditioning on the input analogue part of instrumentation channels. One is the circuit of two four-arm classic bridges connected in cascade [7]. The other one has an unconventional supplying: the 4T circuit is supplied by two equal current sources in parallel to opposite arms – 2J or in practice by switching over two unequal sources between these arms – 2x2J or even only one – 2x1J. For all of these circuits Warsza proposed a common name: double current bridges. The circuits were described in [8], [9] and more extensively in [10]. To illustrate this concept of simultaneous two-parameter measurement, either one-axis strain and temperature or two-axis strain using strain gauges plugged in a double current bridge, an experimental bridge-circuit was built. It can be competitive to the other solutions [1]-[3] and it has following advantages: • there are two different output voltages depending on two measured quantities, i.e. strain and temperature, • the temperature reading and compensation in whole measuring span is realized without any additional sensor such as thermistor or thermistor-resistor parallel networks [4], [5]. Two output signals (representing temperature and strain) are interfaced to a microcontroller by ADCs. The 2J bridge circuits and the cascade bridge circuit are applicable for the GMR (Giant Magneto-Resistive) [12], [13] and other impedance sensors. An alternative idea than 2J bridges is the method of Pallas-Areny and collaborators [14], based on detecting changes in resistance by measuring the time needed to discharge a given capacitor.

2. Unbalanced double current bridge with 4 variable arms

This section describes the difference between double current bridges (2J) introduced by Warsza [6] and Wheatstone bridge. Both types are shown in Fig1. As it is shown in Fig. 1a,b the four resistance (4R) bridge is unconventionally powered by two equal current sources J connected in parallel to opposite bridge arms. Commonly known Wheatstone bridge is additionally given in Fig. 1c for comparison. The output voltages of bridges in Fig. 1 are: U DC = J

R1 R2 − R3 R4 ≡ J ⋅ t DC (ε i ) ∑ Ri

(1)

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

if R10 R40 = R20 R30 , and U DCW = J

Fig. 1. a) and b) – illustration of work in unbalance conditions: R1R2≠R3R4 and R1R4≠R2R3 of two outputs DC and AB of the four resistance bridge circuit unconventionally supplied by two equal current sources J, introduced in [6]; and for comparison: c) well known Wheatstone bridge supplied by current source J, unbalanced for: R1R3≠R2R4

R10 R30 (ε1 − ε 2 + ε 3 − ε 4 + ε1ε 3 − ε 2 ε 4 ) ≡ T0 ⋅ f (ε i ) ∑ Ri 0 ∑ Ri 0εi 1+ ∑ Ri 0 (6)

if R20 R40 = R30 R10 , Where: T0 AB , T0 DC , T0 – initial voltage sensitivities of circuits a- c; f DC (ε i ), f AB (ε i ), f (ε i ) – their unbalance functions. The above conditions for nominal initial values of resistances are fulfilled together if R10 = R20 = R30 = R40 = R0 . Assume that sensors are in all arms of the bridge and their resistance varies and that the sum of all relative resistance 4 increments is equal zero, i.e. ∑ ε i = 0 . In equations (4) and i =1 (5) one can notice that one pair of relative changes in resistance of the same index has the same sign and the other - opposite sign. Additionally, these changes are small when using pairs of strain gauges, then ε i ε j < ε i + ε j . Thus the equations can be simplified as follows.

U DC =

JR0 (ε1 + ε 2 − ε 3 − ε 4 + ε1ε 2 − ε 3ε 4 ) JR0 ≈ (ε1 + ε 2 − ε 3 − ε 4 ) 1 4 4 1 + ∑ εi 4

(7)

U AB =

JR0 (ε1 − ε 2 − ε 3 + ε 4 + ε1ε 4 − ε 2 ε 3 ) JR0 ≈ (ε1 − ε 2 − ε 3 + ε 4 ) 1 4 4 1 + ∑ εi 4

(8)

U AB = J

R1 R4 − R2 R3 ≡ J ⋅ t AB (ε i ) ∑ Ri

(2)

R1 R3 − R2 R4 = J ⋅ t (ε i ) ∑ Ri

(3)

U DCW = J

Where: Ri = Ri 0 (1 + ε i ) , ∑ Ri = R1 + R2 + R3 + R4 ; Ri 0 – the initial nominal resistance, ε i – the relative change in resistance, t, tDC, tAB – open-circuit voltage to current transmittances of DC and AB outputs. Forms (1), (2) of output voltages UDC, UAB of the 2 J bridge are similar to (3) for UDCW of unbalanced Wheatstone bridge supplied from current source, but the outputs DC or AB of the 2J circuit are balanced for the equal products of resistances in the neighboring bridge arms. After separation of resistance changes, the formulas mentioned above are U DC = J

More detailed information about this unconventional type of the bridge is given in [8]-[11]. In [6], [7] and [9] another methods to measure all separate arm increments of 4R bridge is also given. For example four measurements in sequence should be provided, i.e.: both voltages UDC, UAB when the bridge is supplied like in Fig. 1a and Fig. 1b, voltage UDCW for classic diagonal current supply of AB terminals (Fig. 1c) and increment in input open-circuit resistance RAB. Examples of 2J bridge-circuit application for 2-parameter measurements are presented in the next sections of this paper.

3. 2J Bridge with 2 active arms

The 2J bridge differs from the Wheatstone bridge in the way of supplying. It is quite easy to arrange it for simul-

R10 R20 (ε1 + ε 2 − ε 3 − ε 4 + ε1ε 2 − ε 3ε 4 ) ≡ T0 DC ⋅ f DC (ε i ) Ri 0 ε i ∑ Ri 0 ∑ 1+ ∑ Ri 0 (4)

if R10 R20 = R30 R40 and U AB = J

R10 R20 (ε1 − ε 2 − ε 3 + ε 4 + ε1ε 4 − ε 2 ε 3 ) ≡ T0 AB ⋅ f AB (ε i ) Ri 0 ε i ∑ Ri 0 ∑ 1+ ∑ Ri 0 (5)

Fig. 2. Double current bridge circuits for temperature ΔT and bending strain ε measurements Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

tanous two variable measurements (Fig. 2). The bridge has two outputs: A-B and D-C. Two equal current supply sources J are connected in parallel to opposite arms (R1, R3) of the bridge – circuit a). However it is difficult in practice to implement. If the excitations are not equal, equations (1) and (2) have additional components which are dependent on the difference ΔJ [6], [10]. The bridge can also be supplied by single current source J switched over to the same arms – circuit b). Then each of the output voltages is held, averaged in two cycles and measured, i.e.:

U DC = 0,5 U DC 1 + U DC 2

(

)

(9)

(

)

(10)

U AB = 0,5 U AB 1 + U AB 2

Such a supply does not cause the aforesaid problem of the different excitation currents. It ensures a compensation of thermoelectric voltages (of equal opposite values between output terminals) and the independence of the current J direction. Both circuit-bridges in Fig. 2. produce two output voltages (UDC and UAB) presented by (1) and (2) or (7) and (8). Assuming that: R3 = R30 , R4 = R40 (which is tantamount to ε 3 = ε 4 = 0 ) the output voltages are U DC = T0

U AB = T0

ε1 + ε 2 + ε1ε 2 ε1 + ε 2 1+

ε1 − ε 2 ε1 + ε 2

two parts ( ε1 = ε , + ε ,, , ε 2 = ε , − ε ,, , respectively). One part is increment due to temperature (15), the other one is the increment (or decrement) due to mechanical stress (16). If there are two strain gauges of the same type, the relative increments in temperature are of the same value and of the same sign. If strain gauges are glued to a beam such that the first gauge is stretched and the other compressed (Fig. 3a) then the increments due to mechanical stress are of the opposite signs.

(12)

If modules of the values |ε1|, |ε2| are small enough, i.e. ε1 ⋅ ε 2 << ε1 + ε 2 and ε1 + ε 2 << 4 (for absolute changes it is |∆R1+∆R2|<<2(R10 +R20), formulas (11), (12) are simplified to: U DC = T0 (ε1 + ε 2 ) (13) (14)

4. Example of application for strain and temperature measurements

Taking the 4T structure of the 2J bridge circuit into consideration, it can be applied to strain gauges. Two ways of placing the strain gauges on the beam are shown in Fig. 3. The first example is one-axis strain and temperature measurement by using two strain gauges A and B. The second example is 2-axis strain measurement when two forces (moments) are applied in two directions (it will be the purpose of future investigations). The increments in resistance of strain gauges consist of Articles

(15)

ε1" (ε b ) = −ε 2" (ε b ) = ε "

(16)

U DC = T0 (ε ' + ε ' ),

U AB = T0 (ε " + ε " )

(17)

Increments could be considered as linear for both measured quantities, i.e. ε’(ΔT) = αT ΔT; ε”(εb) = k·εb (where: αT – the temperature coefficient of resistance, ΔT – change of temperature, k = k0 (1 + α K ∆T ) – strain gauge factor, εb – bending strain). After substitution, both functions are as follows:

ε , = α T ⋅ ∆T =

where: T0 = 0 – the initial voltage sensitivity is equal 8 for both outputs. The first output voltage is proportional to the sum and the other one to the difference of increments. Also (13) and (14) become simpler, because UAB disappears in the denominators.

ε1' (∆T ) = ε 2' (∆T ) = ε '

From (13) and (14)

U AB = T0 (ε1 − ε 2 )

2012

Fig. 3. One-axis (a) and two-axis (b) strain measurement, A, B, C, D – strain gauges, FB – bending force, FS – stretching force, MB – bending moment

(11)

N° 2

ε ,, = k ⋅ ε b =

U DC 4 U DC = 2T0 JR0

U AB 4 U AB = 2T0 JR0

⇒ ∆T =

⇒ εb =

4U DC (18) JR0α T

4U AB JR0 k

(19)

The both measured quantities depend linearly on the output voltages UAB and UCD (respectively), supplying current J and the parameters of the gauges. Such an advantage of the circuit is difficult to achieve in other DC bridges [15], [16].

5. Real circuit and the experiment

The theory has been verified in the experimental circuit of a transducer given in Fig 4. The switched current source was constructed with the use of LM317 and four MOSFET switches (STP20NE06L), which has very low on-resistance RON = 0.06 Ω. The current excitation can be manually adjustable from 9 mA to 38 mA. The transistors work in pairs – two switched on and two switched off at the

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Fig. 6. Relative increments in resistance ε in response to the beam deflection X [11]

Fig. 4. Transducer system of double current bridge for measurement of strain and temperature

respectively, ε1(65, εb) – the increment due to mechanical stress in temperature 65°C, ε’’(εb) – the increment of mechanical stress after temperature compensation. The best straight line fit to data points presented in Fig. 6 are defined by :

ε ' (65) = 0.0022 X + 0.0710

(20)

ε1 (65, ε b ) = 0.0228 X + 0.0196

(21)

ε ' (23) = 0.0021X

(22)

ε '' (ε b ) = 0.0226 X

(23)

The nonlinearity error with respect to mechanical stress was 1.8% FSR maximum and 0.6% FSR in average. For the measurement of temperature changes the nonlinearity error is a little bigger – 2.5% FSR (maximum) and 1.0% FSR (average). Fig. 5. Laboratory stand (the cantilever had rectangle cross-section, width b = 20 mm, height h = 0.8 mm, length L = 200 mm, the distance between strain gauges and the place where force is fixed l = 180 mm, Young’s modulus E = 2.1⋅1011 N / m 2 ) same time. Their state of work is controlled by Atmega16 microcontroller port. The output voltages UAB and UDC (two of them of positive sign and another two of negative sign) are connected to 24-bit sigma-delta ADC (AD7718) via postconditioning module. It consists of instrumentation amplifiers (AD620AN) and ultra-precision voltage-dividers (MAX5491). The use of this module is necessary because AD7718 requires positive sign voltages of (0 – 2.56 V). The acquired voltage data from the circuit outputs were processed by the microcontroller. The measurements were taken for two temperatures of a cantilever beam (23 °C and 65 °C) while the beam was being bent by a micrometer screw (Fig. 5).

6. Experimental Results

The results of the experiment are shown in Fig. 6. It presents the diagrams of relative increments in resistance ε in response to the beam X deflection: ε’(23), ε’(65) – the increment in temperature calculated for 23°C and for 65°C

7. Bias and precision uncertainties

We assumed that our experimental results were obtained as several independent measurements combined in a single quantity for both measuring channels. K1 and K2 are the calibration factors to be determined. They are output quantities in the measurement models (24) and (25). The input quantities xi are: – voltages UAB and UCD - series of measurements reflect bias (offset) and precision (random) errors, – current J (single measurement – only bias component) – temperature ΔT (single measurement – only bias component), – beam length L (single measurement – only bias component), – beam height h (single measurement – only bias component), – beam deflection X (single measurement – only bias component), – parameters of strain gauges - k, αT, R0 (only bias components). K1 =

K2 =

8 L2 U AB ⋅ 3hkX JR0

(24)

U DC

(25)

⋅

α T ∆T JR0

Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

For uncertainties u ( xi ) that are small compared to xi statistical theorem states that combined standard measurement uncertainty (uncorrelated quantities) is:  N  ∂y  uc ( y ) =  ∑  ⋅ u ( xi )   i =1  ∂ xi

2 1/2

  

(26)

The uncertainties of K1 and K2 are: 1/2

 2u ( L)  2  u (U )  2  2u (h)  2  u (k )  2  u ( X )  2  AB   +  +  +  +  + uc ( K1 )  L   U AB   h   k   X   =  2 2 K1   u ( J )   u ( R0 )   + +         J   R0  

uc ( K 2 )  u (U CD )   u (α T )   u (∆T )   u ( J )   u ( R0 )  =   +  +  +  +  K2  U CD   α T   ∆T   J   R0  2

2 1/2

  

(27)

(28)

Assume Assume the values of relative uncertainties in Tab. 1, total uncertainties of the calibration factors K1 and K2 are: 1/2 uc ( K1 ) = ub ( K1 ) 2 + u p ( K1 ) 2  (29) K1

1/2 uc ( K 2 ) = ub ( K 2 ) 2 + u p ( K 2 ) 2  K2

(30)

where ub ( y ) - combined bias uncertainty, u p ( y ) – combined precision uncertainty. Table 1. Bias and precision uncertainties of input and output quantities

Quantity

Bias ub(xi)/xi

Precision up(xi)/ xi

UAB

0.1%

0.3%

0.5%

0.2%

0.5%

0.05%

0.1%

1.25%

UDC

0.1%

0.3%

ΔT

0.5%

αT

0.5%

0.2%

uC (K1 ) / K1

u b ( K1 )

u p ( K1 )

1.5%

1.46%

0.3%

uC ( K 2 ) / K1

ub ( K 2 )

u p (K 2 ) 2

1.0%

0.94%

0.3%

Articles

N° 2

2012

8. Conclusions

The 2J bridge which measures the real change in temperature of strain gauges in their localization and the mechanical stress is presented in this paper. The innovation of this method consists in a particular supplying of the circuit and in measuring the voltages on diagonals. The measured quantities depend on these values. The resistor bridge does not require an additional temperature sensor. The maximum nonlinearity error with respect to mechanical stress is 1.8% FSR and for the measurement of changes in temperature is a little bigger – 2.5% FSR. Although it is acceptable for many industrial applications the additional experimental tests and some upgrading is still needed. Combined, relative uncertainties of calibration factors K1 and K2 in this experiment are 1.5% and 1.0% (Tab. 1). Furthermore the method of 2-parameter measurement with the two current supply sources can be applied for semiconductor strain gauges. They have not only higher sensitivity than the metallic ones, but are more dependent on temperature. This kind of compensation could be replaced by 2-parameter measurement by the 2J-supply method with digital processing of their output signals. The 2J supplied bridge circuits could be implemented to design different types of MEMS sensors. It will be the aim of further work by the author. Acknowlwdgements This work was supported by Bialystok University of Technology research grant S/WE/3/08.

Authors

Adam Idźkowski – assistant professor at Dep. Of Electrical Eng., Białystok Technical University, ul. Wiejska 45D, 15-351 Białystok, Poland, adam_i@we.pb.edu.pl Jarosław Makal – adjunct professor at Dep. of Electrical Eng., Białystok Technical University, ul. Wiejska 45D, 15-351 Białystok, Poland, j.makal@pb.edu.pl Zygmunt L. Warsza* – professor emeritus at Technical Universities in Kielce and Radom (Dean of Transport Faculty), Poland; for the present professor at Industrial Research Institute for Automation and Measurements PIAP, Warsaw, Poland. President of Polish Metrological Society, zlw@op.pl *Corresponding author

References

[1] MAX1450, low cost 1%-accurate signal conditioner for piezorezistive sensors, Maxim product data sheet. [2] Sensor circuits and digitally controlled potentiometers, Intersil application note AN135, 2005. [3] Sensor temperature compensation using the four DAC signal conditioning architecture, Maxim application note 1839, 2002.

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

[4] C. Swartz, C. Derrington, J. Gragg, Temperature Compensation Methods For The Motorola X-ducer PressureSensor Element, Motorola Semiconductor application notes. [5] NPC-1210 series medium pressure sensor, Product data sheet of GE NovaSensor Inc. [6] Warsza Z. L., “Bridges Supplied by Two Current Sources–New Tool for Impedance Measurements and Signal Conditioning”. In: Proc. of IMEKO-TC-7 Symposium, Cracow, Poland, 2002, pp. 231-236. [7] Warsza Z. L., “Four-Terminal (4-T) Immitance Circuits (4T) in Multivariable Measurements”. In: Monography of Industrial Institute of Control and Measurement PIAP, Warsaw 2004. (in Polish). [8] Warsza Z. L., “Two Parameter (2D) Measurements in Four Terminal (4T) Impedance Bridges as the New Tool for Signal Conditioning part 1 and 2”. In: Proc. of the 14th International Symposium on New Technologies in Measurement and Instrumentation and 10th Workshop of IMEKO TC-4, Gdynia/Jurata, Poland, 2005, pp. 31-42. [9] Warsza Z. L. ., “Backgrounds of two variable (2D) measurements of resistance increments by bridge cascade circuit”. In: Proc. of SPIE, vol. 6347, part 2, 2006, 63472R. [10] Warsza Z. L., “Two Parameter (2D) Measurements in Double – current Supply Four-terminal Resistance Circuits”, Metrology and Measurement Systems, vol. XIII, no. 1, 2006, pp. 49-65. [11] A. Idźkowski, J. Makal, Z. L. Warsza, “Application of Double Current Bridge-Circuit for Simultaneous Measurements of Strain and Temperature”. In: IEEE Instrumentation, Measurement and Technology Conference (IMTC 2007), Warsaw, Poland, May 2007, pp. 1-4. [12] Sifuentes E., Casas O., Pallas-Areny R.,”Direct Interface for Magnetoresistive Sensors”, IEEE Instrumentation, Measurement and Technology Conference, (IMTC 2007), Warsaw, Poland, May 2007, pp. 1-6. [13] Bernieri A., Ferrigno L., Laracca M., Tamburrino A.,”Improving GMR Magnetometer Sensor Uncertainty by Implementing an Automatic Procedure for Calibration and Adjustment”. In: IEEE Instrumentation, Measurement and Technology Conference IMTC 2007, Warsaw, Poland, May 2007, pp. 1-6. [14] Jordana J., Pallas-Areny R., “A simple efficient interface circuit for piezoresistance pressure sensors”, Sensors and Actuators A-Physical, vol. 127, issue 1, 2006, pp. 69-73. [15] Sydenham P.H., Thorn R. (eds.), Handbook of Measuring System Design, John Wiley & Sons, New York 2005, Chapter 126: “Electrical Bridge Circuits – Basic Information” by Z. L. Warsza, pp. 867–877. [16] Dyer S. A. (ed.), Survey of Instrumentation and Measurement, Wiley-Interscience, John Wiley & Sons, New York 2001,Section: “Bridge Instruments” by J. Nicolas, pp. 309-326.

Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Decentralized PI Controller for Multimotors Web Winding System Submitted 27th October 2011; accepted 24th November 2011

Bousmaha Bouchiba, Abdeldjebar Hazzab, Hachemi Glaoui, Fellah Med-Karim, Ismaïl Khalil Bousserhane, Pierre Sicard

Abstract: Web winding systems allow the operations of unwinding and rewinding of various products including plastic films, sheets of paper, sheets, and fabrics. These operations are necessary for the development and the treatment of these products. Web winding systems generally consist of the same machine elements in spite of the diversity of the transported products. Due to the wide rang variation of the radius and inertia of the rollers the system dynamic change considerably during the winding/ unwinding process. Decentralized PI controller for web tension control and linear speed control are presented in this paper. The PI control method can be applied easily and is widely known, it has an important place in control applications. Simulation results show the effectiveness of the proposed linear speed and tension controller for web winding multi motors systems. Keywords: multi motors web winding system, PI controller, tension control, linear speed control

1. Introduction

Many types of materials are manufactured or processed in the form of a sheet or a web (textile, paper, metal, etc.) which then couples the processing rolls and the associated motor drives. The drives are required to work in synchronization to ensure quality processing and rewinding of the product. Tension is a very important web manufacturing and process setting. If severe tension variations occur, rupture of the material during processing or degradation of product quality can occur, resulting into significant economic losses due to material loss and reduced production rate. Therefore, in order to minimize a potential loss, the need arises to adequately control the tension within a predefined range in a mo-ving web processing section. Henceforth, due to their importance in industry, tension control problems have drawn the attention of many researchers. One problem is the establishment of a proper mathematical model. In [1], a mathematical model of a web span is developed, but this model does not predict the tension transfer. This problem was addressed in [2] and [3], with the assumption that the strain in the web is very small. However, the form of the nonlinear and coupling terms in the model are not always convenient for controller design so that other model structures, with comparable precision, are desirable. Several control strategies have been suggested to maintain quality and reduce sensitivity to external disturbances, including centralized multivariable control schemes for steel mill applications 32

Articles

[4], [5] and an H∞ control strategy to decouple web velocity and tension [3], [6]. Also, for tension regulation in a web transport system, [7] proposed a control method based on a unique active disturbance rejection control (ADRC) strategy, which actively compensates for dynamic changes in the system and unpredictable external disturbances. In [8] and [9], Port-Controlled Hamiltonian with Dissipation (PCHD) modeling is considered to develop stabilization strategies with a physical interpretation and motivation of the control action, interpreted as the realization of virtual dampers added to the system, which resulted into a type of dual action controller with velocity feedback and velocity error feedback terms. Some limited improvements were obtained in disturbance rejection properties and robustness with respect to some parameter variations. The conventional PI control dominates industry, it is simple and easy to implement [15]. Tuning of PI controllers is intuitive and is well accepted by practitioners. PIs can at most achieve a compromise in performance in terms of system response speed and stability, and this approach becomes insufficient at the increasingly high web velocities demanded by the industry and with thin or fragile materials. Nonlinearities that appear at high velocities, disturbance rejection properties and robustness to some parameter variations must be accounted for by the controller. A decentralized nonlinear PI controller is proposed to respond to this demand. The model of the winding system and in particular the model of the mechanical coupling are developed and presented in Section 2. Section 3 shows the controllers design for winding system. Section 4 shows the Simulation results using Matlab Simulink. Finally, the conclusion is drawn in Section 5.

2. System model

In this system, the motor M1 carries out unreeling and M3 is used to carry out winding, the motor M2 drives two rollers via gears “to grip” the band (Fig. 1). The stage of pinching off can make it possible to isolate two zones and to create a buffer zone [8, 9]. The objective of these systems is to maintain the linear speed constant and to control the tension in the band. The used motors are three phase induction motors type which each one is supplied by an inverter voltage controlled with Pulse Modulation Width (PWM) techniques. A model based on circuit equivalent equations is generally sufficient in order to make control synthesis. The electrical dynamic model of three-phase Yconnected induction motor can be expressed in the d-q synchronously rotating frame as [13]:

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N째 2

2012

Articles

IFOC

IFOC Linear speed IFOC Controller

IFOC

V2 V2

IFOC Linear speed IFOC Controller

Linear speed Linear speed 4. Si V2 Controller T Controller 3 Linear speed Linear speed Tension V2-ref Controller Controller 4. Si Controller

Linear speed Controller Linear speedT2 Tension Controller Controller Tension T2 Controller T2 Tension T2-ref Controller

V 4.2-ref Si V2-ref

Tension Controller Tension T3-ref Controller

4. Decentralized control web winding system Fig.Fig. 4. Decentralized control forfor web winding system T2-ref T3-ref T3-ref system 2-ref Fig.T4. Decentralized control for web winding Where the parameters of the PI controller are Fig. 4. Decentralized control for web winding system

Fig. 7. The of unwinder M1 Fig.inertia 7. Themoment inertia moment of unwinder M1

K P = 2 rL

(17)

K I = 2 r2 L

(18)

25 20

In the sequel, the decentralized structure shown on (Fig. 4) will be considered. The control structure is composed of 3 elementary controllers associated respectively to each motor. The cascade control configuration uses the tension as primary measurement and velocity as secondary measurement. The manipulated variable is the torque applied to the motors.

4. Simulation Results

The system was simulated in the MATLAB SIMULINK environment for a three motors winding system the winding system and the control system parameters are given in Appendix. As shown in Fig. 5, Fig. 8 and Fig. 11. An improvement of the linear speed is registered, and has follows

15 V1 Vref

0.3 0.2 0.2 0.1

2 3 Times (s)

the reference speed for PI controller after 0.3 sec, in all 0.04 0.03 motors, with the overshoot in linear speed (5%). Fig. 6 and Fig. 70.03 show that the radius and the inertia moment of 0.02 unwinder M1 decrease compared with the radius and the 0.02 of Motor M3 is increase shown in Fig. 9 inertia moment 0.01 and Fig. 10. 0.01 0 0 1 2 3 4 5 0 Times (s) 0 1 2 3 4 5 J2 [N.m] J2 [N.m]

[m/sec] ref 1

V ,V

Articles

0.4 0.3

2 3 4 5 Times (s) 1 2 3 4 5 Fig. 9. The radius of motor M2 Times (s) 0.05 9. TheM2 radius of motor M2 Fig. 9. The radius Fig. of motor 0.05 0.04

Fig. 5. The linear speed of unwinder M1 34

10 5

0.1 0 0 0 0

V2 Vref V2

15 10

Vref 05 0 1 2 3 4 5 0 Times (s) 0 1 2 3 4 5 Fig. 8. The linear speed of Motor M2 Times (s) Fig. 8. The linear speed of Motor M2 0.5 Fig. 8. The linear speed of Motor M2 0.5 0.4

20 15

J3 [N.m] J [N.m]

IFOC

2 3 4 5 Times (s) 1.242 1.24 Fig. 6. radius2of unwinder 0 The 1 3 4M1 5 1.24 Times (s) 0 1 2 3 4 5 0.022 Fig. 6. The radius of unwinder M1 Times (s) Fig. 6. The radius of unwinder M1 0.0219Fig. 6. The radius of unwinder M1 0.022 0.0218 0.022 0.0219 0.0217 0.0219 0.0218 0.0216 0.0218 0.0217 0.0215 0.0217 0 1 2 3 4 5 0.0216 Times (s) 0.0216 0.0215 1 moment 2 of unwinder 3 4 M1 5 Fig. 7. The0 inertia 0.0215 Times (s) 0 1 2 3 4 5 Fig. 7. The inertiaTimes moment (s)of unwinder M1

From follows ItFroa follows controll betterItp control tracking better p trackin 5

4 [N]

R [ ]

R [ ] R [ ]

R [ R] [ ]

,Vref [m/sec] V2,VVref 2 [m/sec]

R1 [m] R1 [m] R1 [m]

1.246 1.25 1.248 1.244 1.248 1.246 1.242 1.246 1.244 1.24 1.244 1.242 0

R2 [m] R2 [m]

Times (s) 50 Fig. 5.0The linear speed unwinder 1 2 of 3 4 M1 5 VOLUME 6, N째 2 2012 0 Times (s) 0 1 2 3 4 5 1.25 Fig. 5. The linear speed Times (s) of unwinder M1 1.248 Fig. 1.255. The linear speed of unwinder M1

J1 [N.m] J1 [N.m] J1 [N.m]

K I = 2 r L to (18) respectively each motor. (Fig.4) will be considered. The control structure is The cascade control configuration uses the tension as In the sequel, decentralizedcontrollers structure shown on composed of 3the elementary associated primary measurement and &velocity as secondaryis Journal of Automation, Mobile Robotics Intelligent Systems (Fig.4) will be considered. The control structure respectively to each motor. measurement. manipulated variable is the torque composed ofThe3control elementary controllers The cascade configuration uses theassociated tension as applied to the to motors. respectively each motor. primary measurement and velocity as secondary M2 The cascade The control configuration uses the as measurement. manipulated variable is tension the torque primary measurement and velocity as secondary applied to the motors. measurement. The manipulated M2 variable is the torque applied to the motors. M2

R [ ]

0.03 0.01

0.01

2 3 Times (s)

Figure.10 – The inertia moment of Motor M2. 0 Fig. 10. The25inertia moment 0 1 2 of Motor 3 M24 5 Times (s) 25 15

100

V3 Vref

2 3 Times (s)

V3 4 Vref

2 0

2 3 Times (s)

T2 Tref 4

*Cor

2 3 Times (s)

Refe [1] [2] 5

Fig. 15. The tension between Motor 2 and 3 Fig. 15. The tension between Motor 2 and 3 Table System parameters From the1. Figures (14-15), we can say that: the tension follows the reference tension with application of PI E 1.6e8 L1= L2= L3[m] 5 controller. 2 50 PI ƒn [Hz] control with S [m ] 2.75e-3 It appears clearly that the classical 4 R [m] 1.25 T = T [N] 1 1ref 2ref controller in linear speed control and tension control of20 R = R [m] 0.25 V [m/s] 2 3 2ref fers better performances in both of the overshoot control and the Jtracking 01=J02= error. However is easy to apply. p 2 0.022 J03[Kg.m2]

0.29

R3 [m] R [m] 3

3 1

4 0 14. The tension between Motor 1 and 2 Fig. 0 2 3 4 5 31 Times (s) T3 2 Fig. 14. The tension between Motor 1 Tref and 2 1

Fig. 11. The linear speed of winder M3 Fig. 11. The linear speed of winder M3

[3]

[4]

[5]

[6]

Table 1. System parameters

2 3 Times (s)

5. Conclusion

R [ ]R [ ]

0.24 Fig. of 0 12.1 The radius 2 3 winder 4 M35 Fig. 12. The radius of winder M3 Times (s) 0.0221 Fig. 12. The radius of winder M3 0.0221 0.0221 0.0221 0.0221 0.022 0.0221 0.022 0.022 0.022 1 2 3 4 5 0.022 0 Times (s) 0.022 Fig. 13.0The inertia moment 1 2 3of winder 4 M35 Times (s) From the figures (14-15), we can say that: the tension follows the tensionmoment with application PI controller. Fig.reference 13. The inertia of winderofM3 It appears clearly that the classical control with PI Fig. 13. The inertia moment of winder M3 From the (14-15), we can that:control the tension controller in figures linear speed control and say tension offers follows the reference tension application PI controller. better performances in both with of the overshootofcontrol and the It appears clearly that theto apply. classical control with PI tracking error. However is easy J3 [N.m]J [N.m] 3

R [ ]

T2 4 2 Figure.14 – The tension between Motor 1 and 2 Fig. 14. The tension between Motor Tref 1 and 2

Fig. 0 11. The linear speed of winder M3 0 1 2 3 4 5 Times (s)

0.28 0.29 0.27 0.28 0.26 0.27 0.25 0.26 0.24 0.25 0

2012

Fig. 13. The inertia moment of winder M3 0.022 50 1 2we can 3 say 4that: the 5 tension From the figures (14-15), Times (s) follows the reference tension with application of PI controller. 4 It appears that moment the classical control Fig. 13.clearly The inertia of winder M3 with PI 3 controller in linear speed control and tension control offers From the figures (14-15), we can say that: the tension better performances in both of the overshoot and the T2ofcontrol follows the reference tension with application PI controller. 2 tracking error. However is easy to apply. It appears clearly that the classicalTref control with PI controller in linear speed control and tension control offers 1 5 better performances in both of the overshoot control and the tracking4error. However is easy to apply. 0 0 1 2 3 4 5 Times (s) 5 3

155

N° 2

T3,Tref [N]

V3,Vref [m/sec] V3,Vref [m/sec]

20 Figure.10 – The inertia moment of Motor M2.

20 10

[N]

0.020

2 3 Times (s)

[N]

R [ ]

0.04 0.02

2 ref

2 3 4 5 Times (s) Fig. 9. The radius of motor M2

VOLUME 6,

T ,T

0.05 0.03

0.022 0.022 0.022

R [ ]

J2 [N.m] J [N.m] 2

0.04

2 3 4 5 Times (s) Fig. 9. The radius of motor M2

2 ref

0 0.05 0

[N]T ,T

0.1

2 ref

0.0221 0.022

J3 [N.

Journal of Automation, Mobile Robotics & Intelligent Systems 0.2

T ,T

R2 [m]

0.3 0.1

The objective of this paper a 1.6e8 L1= L2consists = L3[m] in developing 5 model of a winding system constituted of three motors that 2 ƒn [Hz] S [mis ] coupled mechanically 2.75e-3 by a strap whose50 tension is adjustable and to develop the and 4 R1 [m] 1.25 T1ref= methods T2ref [N] of analysis synthesis of the commands robust and their application to 20 a constant R2=Rsynchronize [m] V2ref [m/s] the0.25 three sequences and to maintain 3 mechanical tension between the rollers of the system. J01=J02=Computer J03[Kg. 2 simulations pshow the robustness and the 0.022 m 2] performance of the winding system with the PI controllers, however PI control dominates industry and it is simple and easy to implement.

5. Conclusion

The objective of this paper consists in developing a model of a winding system constituted of three moAUTHORS tors that is coupled mechanically by a strapHazzab, whose tenBousmaha Bouchiba*, Abdeldjebar Hachemi 1 Glaoui , Ismaïl Khalil Bousserhane of sion is adjustable and to develop the methods–ofLaboratory analysis command, Analysis and Optimization of the Electroand synthesis of the commands robust and their applicaEnergizing Systems, Faculty of Sciences technology, tion to synchronize the three sequences and toand maintain Béchar University B.P. 417 Béchar, 08000, Algeria. a constant mechanical tension between the rollers of the a_hazzab@yahoo.fr, ghachemi@yahoo.fr, system. ismail@yahoo.fr Fellah Med-Karim – Laboratory of ICEPS Intelligent 35 Articles Control and Electrical power Systems University of Sidi

[7]

[8]

[9]

[10]

[11]

Journal of Automation, Mobile Robotics & Intelligent Systems

Authors

Bousmaha Bouchiba*, Abdeldjebar Hazzab, Hachemi Glaoui, Ismaïl Khalil Bousserhane – Laboratory of command, Analysis and Optimization of the ElectroEnergizing Systems, Faculty of Sciences and technology, Béchar University B.P. 417 Béchar, 08000, Algeria. a_hazzab@yahoo.fr, ghachemi@yahoo.fr, ismail@yahoo.fr Fellah Med-Karim – Laboratory of ICEPS Intelligent Control and Electrical power Systems University of Sidi Bel-Abbes, Algeria. mkfellah@yahoo.fr Pierre Sicard – Department of electrical and computer engineering, Université du Québec à Trois-Rivières, Canada. pierre.sicard@uqtr.ca *Corresponding autor: bouchiba_bousmaha@yahoo.fr

References

[1] D. P. Cambell, Process Dynamics, Wiley, 1958, pp.113-156. [2] G. Brandenburg, “New Mathematical Model For Web Tension and Register Error”. In: Proceedings of the 3rd IFAC Conference on Instrumentation and Automation in The Paper, Rubber and Plastics, vol. 1, May 1976, pp. 411-438. [3] H. Koç, D. Knittel, M de Mathelin, G. Abba, “Modeling and Robust Control of Winding Systems for Elastic Webs”, IEEE Trans. Contr. Syst. Technol., vol. 10, March 2002, pp.197-208. [4] J. E. Geddes, M. Postlethwaite, “Improvements in Product Quality in Tandem Cold Rolling Using Robust Multivariable Control”, IEEE Trans. Contr. System. Technology, vol. 6, March 1998, pp. 257-267. [5] S. H. Jeon et al., “Decoupling Control of Bridle Rolls for Steel Mill Drive System”, IEEE Trans. Ind. Application, vol. 35, January/February 1999, pp. 119-125.

Articles

VOLUME 6,

N° 2

2012

[6] D. Knittel et al., “Tension Control for Winding Systems With Two-Degrees of Freedom H∞ Controllers”, IEEE Trans. Ind. Applicat. Syst., vol. 39, January/February 2003, pp. 113-120. [7] B.T. Boulter, Y. Hou, Z. Gao, F. Jiang., “Active Disturbance Rejection Control for Web Tension Regulation and Control”. In: IEEE Conference on Decision and Control, Orlando, USA, December 2001, pp. 4974- 4979. [8] F. Mokhtari, P. Sicard, N. Léchevin, “Damping Injection Control of Winding System Based on controlled Hamiltonian Systems”. In: 12th IFAC Symposium on Automation in Mining, Mineral and Metal Processing – IFAC MMM’07, Québec, Canada, August 2007, pp. 243-248. [9] F. Mokhtari, P. Sicard, N. Léchevin, “Stabilizing Winding Systems by Injection Damping Control Based on controlled Hamiltonian Systems”. In: Proc. of IEEE International Electric Machines and Drives Conference – IEMDC’07, Antalya, Turkey, May 2007, pp. 95-100. [10] Bousmaha Bouchiba, Abdeldejbar Hazzab, Hachemi Glaoui, Fellah Med-Karim, Ismaïl Khalil Bousserhane, “Sliding Mode Speed Control for Multi-Motors System” , JAMRIS, vol. 4, no. 3, 2010, pp. 50-54. [11] Christian Thiffault, Pierre Sicard, Alain Bouscayrol, “Tension Control Loop Using a Linear Actuator Based on The Energetic Macroscopic Representation”. In: CCECE 2004–CCGEI 2004, Niagara Falls, Canada, May 2004. [12] S. Charlemagne, A. Bouscayrol, Slama Belkhodja, J.P. Hautier, “Flatness based control of non-linear textile multimachine process”. In: Proc. of EPE’03, Toulouse, France, September 2003, CD-ROM. [13] Adlane Benlatreche Dominique Knittel “State Feedback Control with Full or Partial Integral Action for Large Scale Winding Systems”. In: Industry Applications Conference, Oct. 2005, vol. 2, pp. 973-978.

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Methods Of Decreasing The Influence Of The Factors Disturbing The Reliability Of Leak Detection Systems Submitted 27th June 2011; accepted 2nd September 2011

Mateusz Turkowski, Andrzej Bratek

Abstract: During implementation of leak detection and localization systems for liquid and gas pipelines the authors have met serious problems concerning reliability. The main problems were integrity (i.e. discontinuities) and quality of the input data acquired from telemetry systems, such as damages, bad calibration and drift of the measuring instruments and transmitters, bad balance between fluid entering and leaving the system and incorrect installation of temperature transmitters. The proposals how to overcome these problems have been presented. Keywords: leak detection, leak localization, reliability

1. Introduction The leak of the gas or liquid pipeline brings always large losses of various kind: suspending the product transport, the cost of reparation of the damage and loss of transported product. In case of explosive or/and flammable or/and dangerous to environment media (e.g. petroleum and other petrochemical products), the leak causes hazard for safety of the people and the equipment as well as environmental contamination. Events like that induce high social and financial costs, which are proportional to the leak intensity and duration. At all stages of pipeline building and operating must be therefore fulfilled the regulations and recommendations of numerous standards and regulations, whose purpose is to provide long-lasting operation of pipeline system. However, even if the pipeline has been designed and built very carefully, there is always a potential of leaks. If in spite of all precautions a leak happened, its effects can be minimized only by fast detection and localization of the leak point enabling quick dispatcher reaction (stopping pumping, closing the valves, organizing provisional damage repair etc.). Pipeline leak detection systems play therefore a key role in minimization of the leaks probability and impact. A lot of technologies of leak detection and localization are commercially available today, the background information about them has been presented in [1]. In case of long range pipelines most of the leak detections methods are analytical (internal) methods based on comparison of the pipeline mathematical model data with the real measurement data obtained from the telemetry or SCADA systems. For the purposes of the leak detection and localization systems the following parameters are measured and processed:

• pressure at the inlet, at the end, and at as much as possible points located along the pipe, i.e. at the valve stations, metering and regulating stations, terminals, • flow rate at the inlet and if possible at the outlet of the pipeline, • temperature usually at the same places that pressure measurement or minimum at inlet and outlet, • if possible density of the transported liquid at inlet of the pipeline. Most important parameters for the procedures of leak detections are pressure and flow-rate. Measurements of other variables (temperature, density) are auxiliary but they can significantly increase the system accuracy. Measured pressure p is compared with calculated pressure pcal, in the simplest case (static model) the following formula can be used for gas:

pcal =

p02 −

1.62114qst2 ρst2 ZRT λ L D5

(1)

where p0 – pressure at inlet of the pipeline, qst and ρst – flow rate and density at standard conditions, Z – compressibility coefficient, R – gas constant, T – absolute temperature λ - friction coefficient (function of Reynolds number and pipe roughness), L – pipeline length, D – pipe internal diameter. In most cases, however, much more complicated, but more accurate dynamic models in form of differential equations system are used [2, 3], i.e. for liquids:

∂w( x, t ) 1 ∂p( x, t ) + =0 ∂x E ∂t

(2)

∂p ( x , t ) ∂w ( x , t ) + ρ ( x) = ∂x ∂t λ ( x) ρ ( x) = − ρ ( x ) g sin α − w (t ) w (t ) 2D where: x – coordinate along the pipeline t – time w – average velocity of liquid E – elasticity modulus of liquid-pipeline system ρ – liquid density a – inclination angle of the pipe. Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

Apart of the physical models some neural, additive, fuzzy, swarm particles models are used, sometimes supported by data mining methods [4]. They do not need precise information about the pipeline, so they are suitable in all cases when the leak detection system has to be installed in old pipelines. Almost all methods are based on the constant comparison of the measured and calculated data. If the differences (residua) override a certain limit the alarm is generated and localization procedures are started. The algorithms of leak detection and localization are very sensitive to the data discontinuity, fluctuations of measuring signals, resulting from instruments noise, uncertainty [5] and systematic errors of the instruments. The quality of the measurement data is therefore of greatest importance. It is evident, that bad quality of the data will strongly influence the leak detection systems and can generate false alarms which can be as dangerous as the leak itself – the dispatcher cease to react even to the real leak alarms.

VOLUME 6,

N° 2

2012

Another example, for liquid pipeline, is presented in Fig. 2. This is an example of dedicated (not public, as in case of GPRS) telemetry system. The temporary disturbances are occasional, but even the short-lived disturbance of the signal used by the system can induce false alarms or leak localization faults. The changes of the operational conditions i.e. change of the pumped medium, switching the sending/receiving tank (see Fig. 2), start or stop of the compressor in gas compressor station change the parameters of the pipeline system. It can have negative influence on the performance of the leak detection system.

2. Discontinuity of the data

The data acquired from the telemetry or SCADA system usually contain a lot of discontinuities. The discontinuities are particularly frequent in the systems based on the GPRS connection, usually when the network is overloaded or the distance from the nearest antenna is great. Fig. 1 presents the example of data (pressures) from one of the gas metering and regulating stations installed in the gas pipelines system – object of the research. The number of discontinuities can be great, sometimes up to 30 – 35 % of all measurements. In case of data lack, the data from telemetry system has to be replaced by the approximated values. This can be fulfilled by the linear approximation – the straight line is calculated from the previous data with the use of least squares method. In case of long data absence the reconstruction is however suspended and the signal is no more taken into account by the leak detection system (Fig. 1).

Fig. 2. Disturbances of signals from 8 pressure transmitters installed in subsequent places on the liquid pipe The system should recognize, identify and distinguish between the disturbances due to operational conditions (i.e. starting or stopping pumping) and disturbances due to measuring transmitter failure. The measuring signals recognized as wrong, are eliminated from the leak detection process whereas occurrence of signal disturbances identified as caused by operational maneuvers need special treatment by the detection and localization algorithms. The identification of wrong signals is based on the observation of the momentary changes of the measured parameters ant their comparison with the values calculated from the model. The parameters witch have the values out of typical for the pipeline parameter ranges are rejected, as well as the values which speed of change is to high and not justified by the operational changes of pipeline parameters. The exclusion of the wrong transmitters can be fulfilled automatically or manually.

3. Transmitter calibration problems

Fig. 1. Pressure signal transmitted from the gas pipeline, possibilities of the signal correction 38

Articles

The serious problem is a bad calibration of the pressure transmitters. Sometimes the offset of the transmitter signal may caused that positive changes of pressure along the pipeline are observed, what is contradictory to the fundamentals of physics. In such cases the transmitter characteristics can be corrected with the use of the model data. It must be underlined that each calibration (which is seen by the system as the rapid pressure change) must be done in close cooperation with the detection system supervisor, which has to temporary exclude the calibrated transmitter from the system.

Journal of Automation, Mobile Robotics & Intelligent Systems

4 Efficiency of the filtration methods

Before the measuring signals are utilized through the system they have to be filtered to decrease the noise influence to the system efficiency. The various low pass filters were considered but the time averaging of the signals at various time constants has been proved as very useful. Such averaging can be written in the form

∑ y=

n −1 i =0

x (k − i) n

where n – number of averaged samples; multiplied by sampling interval represents averaging time constant, x – input signal, y – filtered signal. The effects of such data processing is presented in Fig. 3.

VOLUME 6,

N° 2

2012

p – pressure, w – fluid velocity, h – height, subscript i – number of the pipeline section. b) For the calculation of the friction coefficient the mean values of the pressures and velocities for the period equal about the time of stabilization of the disturbances propagation in the pipeline should be used. c) Periodically, during stable conditions, the consistency between model data and measured data for subsequent pressure measurement points should be checked. d) In case of recognition in the section i the difference between calculated and measured pressure greater then the accepted threshold value dP, the friction coefficient λ in the i-th section should be corrected. The threshold value should be determined taking into account the normally existing pressure variations. In order to do not overlook the real leak in each step only the small part of the correction should be introduced, according to the recursive formula (5) with the coefficient z which can be equal i.e. 0.01.

λi = ςλi −1 + (1 − ς ) λi

(5)

During such procedure of tunning, despite only the value of the friction coefficient l is modified, all the slowly changing parameters i.e. instrumentation drift between calibrations, are compensated.

Fig. 3 Diagram of the pressures at two gas station, rough and averaged. Averaging with time constant 15 s

Imbalance

Parallel to the analysis of the pressure distribution in the system the balance of the fluid flowing in and out of the system should be monitored. For this purpose some index proposed in [6] should be calculated; in case of gas according to the formula (6): n

τ (t ) = ∆Vst ,in (t ) − ∑ ∆Vst ,out (t ) − Vst ,acc (t ) i =1

The most effective method for liquid pipes was the double filtration with time averaging wit the use of time constant 1 s and 10 s. Lower time constant (1 s) was used to detect the occurrence of the leak fast and the higher (10 s) – to stabilize measurement results for the purpose of precise leak localization.

5. Tunning of the model

The data introduced in the model is never perfect. During operation i.e. the roughness of the pipeline due to corrosion or dirt can change (so the friction coefficient λ also changes). The system need therefore permanent tunning to assure consistency between model and the real pipeline system. The method of tunning can be described as follows: a) During stable state of the pipeline (constant flow, no technological operations) the values of the friction coefficients should be permanently computed from the formula (4). They should be calculated for subsequent pipe sections between pressure measurement points.

λi =

2D ρw 2 Li

[( pi − pi+1 ) + ρg (hi − hi+1 )]

where D – internal pipe diameter Li – length of pipeline section, g – acceleration due to the gravity, ρ – density,

The variable t(t) can be called corrected flow imbalance at the moment t. This is the difference between the volume of gas flowing into the pipeline system DVn,in(t) and the volume that has flown out of the sysn

tem

∑ ∆V i =1

st ,out

(t ) (n is the number of output stations),

minus the volume of the gas accumulated in the pipeline Vst,acc(t); subscript st denotes standard conditions. The term Vst,acc(t) depends on the gas temperature, pressure and composition and can be calculated with the use of the formula

Vn ,acc (t ) = Vg ρst

pavg Tst pstTZ

where Vg is the geometrical volume of the pipeline system, T absolute temperature. and pavg, average pressure in the pipeline section can be calculated as

pavg =

pi2+1  2   pi + 3 pi + pi +1 

(8)

Calculation of the t(t) can give important information about the leak intensity, which can be used as input parameter in some procedures of leak localization. This paArticles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

rameter fluctuates about some medium value m, mainly due to the instruments drift, gas meters systematic errors (changing with the flow rate) or uncontrolled temperature changes. These fluctuations can be characterized with the variance s2. Let us denote the momentary deviation from the mean value m, as Dm. Than, the cumulative sum a(t) given by the formula (8) can constitute one of the criterions of alarm generation when it excess some level, an adequate procedure has been presented in [6]. This level can vary depending of the state of the pipeline, in steady state it can be rather low, and during the technological operations generating instabilities, transients etc. it can be set to the higher value.

α (t ) = α (t − 1) +

∆m  ∆m  τ (t ) − m −  2  2  σ 

N° 2

2012

The simulation of the gas temperature distribution with the use of Computational Fluid Dynamics (CFD) is presented in Fig. 5.

(9)

During the research authors have encountered difficulties trying to make the use of the formulae (6) and (9). Both the differences between the incoming and outgoing flow rate and the cumulative sum exceeded significantly the expected values (fig. 4).

Fig. 5. CFD simulation of the gas temperature in the pipeline heated in the upper part to t = 40 °C (313 K), by solar radiation, temperature of the incoming gas is 10 °C (283 K) The temperature difference between gas flowing under ground can differ from that measured above the soil up to several degrees. Underground installation of the temperature sensors may be a solution, as shown in Fig. 6.

Fig. 4. Cumulative sum (upper diagram) and difference between incoming and outgoing flow rate of gas (lower diagram) The excessive values were attributed to the improper temperature measurement method, because they were strongly correlated with the day/night cycle. Probably the inventory volume of the gas in the pipeline was therefore calculated not correctly.

7. Temperature measurement

Temperature influences the density and viscosity, so Reynolds number and in consequence friction coefficient l. It also influences the inventory volume of the fluid in the pipeline and can be the source of problems described in previous chapter. The problem is the right localization of temperature sensor. The natural place is the valve system, usually situated above the ground level. Especially for gas it can be source of significant errors. Because of low gas heat transfer coefficient measured temperature in winter is lower than the actual temperature of gas, and in summer – higher. 40

Articles

Fig. 6. Superficial temperature sensor and the method of its installation The new temperature sensors have been installed during upgrading the measurements instruments of a gas pipeline to make it possible installation of the leak detection system. Because it would be very costly to install the thermowell in the pipe under pressure (gas transport can not be interrupted), the surface temperature sensors were chosen. The sensors must be isolated, both against the moisture and thermally. Probably the excessive changes of the difference between incoming and outgoing gas presented in Fig. 4 were due to the improper temperature sensor installation.

8. Conclusions

This paper can seem discouraging for the potential user or designer of the leak detection and localization systems, but most of the problems can be kept to a minimum or even eliminated during implementing such systems. In most of cases the existing, dedicated for the routine maintenance, measurement and telemetry system have to be retrofitted to comply with the needs of the leak detection and localization system.

Journal of Automation, Mobile Robotics & Intelligent Systems

The short discontinuities of the signal can be eliminated by extrapolation of the previous data, the longer discontinuities, however, demand the exclusion of the measured parameter from the system. The bad calibration of the measuring transmitters can be corrected in some extent, with the use of model data. The operator of the Leak Detection System must however close cooperate with pipeline operator, it concerns mainly the procedures of calibration of field transmitters, because the currently calibrated transmitters should be ignored by the system. The slowly changing parameters, as instrumentation drift, changes of pipe roughness etc. can be compensated by the tunning of the model. The manner of the installation of temperature sensors may have important influence on the system performance. It is preferable to install the temperature sensors underground, because the temperature measured above ground may be not representative, it concerns mainly gas pipelines. The use of the superficial sensors makes it possible the retrofitting easily and without disturbances of gas supply.

VOLUME 6,

N° 2

2012

the leak detection and localization systems”. IN: ACADEMIC NETWORK EVENT at EGATEC2011, The European Gas Research Group, Copenhagen, Denmark, 12th-13th May 2011. [6] R. Beuhausen et al., “Transient leak detection in crude oil pipelines”. In: Proceedings of International Pipeline Conference, 4th–8th October, 2004 Calgary, Alberta, Canada.

Acknowledgements

This work was partially supported by Polish Ministry of Science and Higher Education within research project nr O R00 0013 06.

Authors:

Mateusz Turkowski* – professor in the Institute of Metrology and Biomedical Engineering, Warsaw University of Technology, ul. św. A. Boboli 8, 02-525 Warsaw, Poland, m.turkowski@mchtr.pw.edu.pl Andrzej Bratek – Industrial Research Institute for Automation and Measurements, Al. Jerozolimskie 202, 02486 Warsaw, Poland, abratek@piap.pl *Corresponding author

References:

[1] M. Turkowski, A. Bratek, M. Słowikowski, “Methods and Systems for leakage detection in long range pipelines”, Journal of Automation, Mobile Robotics and Intelligent Systems, vol. 1, no. 3, 2007, pp. 39‑46. [2] M. Turkowski, A. Bratek, M. Słowikowski, “The improvement of pipeline mathematical model for the purposes of leak detection”. In: Recent Advances in Mechatronics – 7th International Conference Mechatronics, Warsaw, Poland, 2007, pp. 573-577. [3] R. Sobczak, M. Turkowski, A. Bratek, M. Słowikowski, “Mathematical modeling of liquid flow dynamics in long range transfer pipelines”. In: Problemy eksploatacji, Part 1, ITE, Radom, 2007. [4] J.M.Kościelny, M. Syfert, Ł. Tabor, “Sequential Residual Design Method for Linear Systems”. In: Conference on Control and Fault-Tolerant Systems SysTol’10, Nice, France, 2010. [5] A. Bogucki, M. Turkowski, “The influence of the measuring instruments accuracy on the reliability of Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Modelling of Mechatronic Devices Supported by 3D Engineering Software Submitted 27th June 2011; accepted 3rd September 2011

Jakub Wierciak, Ksawery Szykiedans, Aleksandra Binder-Czajka

Abstract:

Modelling is considered to be an inherent part of design process of mechatronic devices and systems, in particular when solving problems of dynamics and accuracy of actuators and sensors. Mathematical software packages such as Matlab/Simulink are commonly used for this purpose. Engineers and designers who integrate mechanical components of devices under design, usually employ special computer software known as 3D CAD for creating three dimensional images. Typically it is used to generate technical drawings and layouts but its usability is more extensive. The paper presents ability of such software to support simulation where a linear stepping actuator is an example of a device under design.

2. Determination of user’s needs and analysis of a main function of the system

Modelling is recognized as one of characteristic features of mechatronic systems and devices designing. However the term “modelling” is a general one and it covers various techniques referring to various domains and parts of a system under design. The authors attempt to address those techniques to particular phases of design process as they were previously defined. The analysis has shown that the following phases of the design process could be identified [7]: • developing of a system structure, • identifying sensors and actuators,

Designing should be started from the definition of client’s requirements. It has to consists of following parts [7]: – system functions, commonly presented as operating algorithms or lists of operations that device has to do with numerically defined characteristics, – description of a system structure including mechanical, electronic and software subsystems, – description of system environment with physical and legal conditions, also possible presence of other systems and user. In this phase, when user’s needs are reviewed, 2D or 3D geometrical models of the device under design can be generated in order to obtain the client’s approval for either its look or layout (Fig. 2). It can be extended by modelling some activities when using the device or system – sometimes technical models are used. When analysing a main function of a system, time relations between particular operations are crucial.

Fig. 1. Classification of models with respect to type of state variables by [1]

Fig. 2. Virtual three dimensional model of an actuator developed with Autodesk Inventor. Use of such a model allows presenting desired dimensions, movement ranges or displacement of mounting holes

Keywords: mechatronic drive, linear actuator, modelling

1. Introduction

• designing of sensors and actuators, • designing of subsystems: electromechanical, electronic and software, • supervising of making a prototype, • trials with modifications of a system in any layer. Mathematical models used in design process (Fig.1) usually are expressed in either balance, constitutive or phenomenological equations. Their complexity depends upon the described phenomena, the goal of experiments, available data as well as simulation software capability.

Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

At the phase of main function analysis, the functional concept of a device is being developed. Upon this concept a list of necessary actuators and sensors is formulated. User’s requirements are translated into technical requirements for each unit, either actuating or measuring. 3D software is not useful for this stage of project. Block diagrams are the most common form to illustrate these works.

3. Developing of actuators and sensors

A result of this stage of a design should be a proposal of technical solutions of particular actuators and sensors identified at previous stage. In this phase typical modelling and simulation of dynamical systems is performed in order to obtain time responses, which supports decisions upon technical solutions of actuators and sensors [7]. 3.1. Modelling with CAD 3D software This stage of design can be aided to some extent by 3D CAD software. The software gives engineers an opportunity to create in virtual reality various versions of mechanical parts of sensors and actuators in order to test them at very low cost. Definition of physical parameters of 3D models such as material properties, factor of friction, stiffness or elasticity are so close to reality, that in majority of cases the models are very reliable. Simulations of object dynamics can be carried on final model of an assembly or on demonstrative, draft one. In the second option parts will be made in simplified way to illustrate only their main functions, essential for results of simulation (Fig. 3). Simplified models allow to proceed more tests and simulations with bigger number of possible configurations. This leads up to choosing optimal solution. The following example presents ability of 3D CAD software to perform such simulation. Autodesk Inventor

Fig. 3. Screw gear – virtual model prepared for preliminary simulation software package was used for experiments upon screw gear (Fig.4) used in linear actuator driven by stepping motor. When mechanical properties of the gear are declared, actuating forces and loads can be applied to the model. Results of such simulation showing time series of pusher travel and velocity are presented in Fig. 5. Due to some limitations of the program the motor torque which drives the nut was modelled as a step function. Though it differs much from real form of torque of stepping motor, experiments generated time responses of pusher position and velocity being a basement for some practical conclusion. When loading force is opposite to the pusher movement (0 – 1 s) then velocity grows slowly, with acceleration 19,5 mm/s2. When motor torque drops

Fig. 4. Selection of components of screw gear to 0 (1 – 2 s) the pusher stops after 0,16 s. As the screw gear used in actuator is self-locking one, no movement occurs under axial force applied to the pusher. When motor torque acts in consistency to the loading force vector (2 – 3 s) acceleration is more than twice bigger (43 mm/ s2) and the pusher travel is also bigger, so it overcomes the starting position. Repeating of the same operation cycle makes position of the pusher even more distant from the starting point. It proves the essential effect of load upon performance of the actuator. In fact when stepping motor is applied, positioning is controlled by the number of timing pulses. The 3D software used for simulation described above has many limitations referring to shape of input signals applied to the model. Therefore other tools should be used in order to obtain more reliable results.

Fig. 5. Simulation of a helical joint dynamic with applied load of 30 N; actuating torque M, rod velocity V, rod position s 3.2. Modelling with software for multidomain simulation of dynamic systems When modelling actuators, the procedure starts with either rotational or linear movement model, depending upon the kind of movement realised by a drive used. If rotational movement takes place, then the classical equation of torque equilibrium makes the model 2

dγ d γ  + M lf sgn   + M l = M m dt  dt  (1) where: Jl – moment of inertia of rotating elements reduced to the motor shaft [kg/m2], Jm – moment of inertia of motor rotor [kg/m2], KD. – coefficient of viscous damp-

(J m + J l ) d

dt2

+ KD

Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

ing [Nm/rad/s], Mm – motor torque [Nm], Ml – active load torque reduced to the motor shaft [Nm], Mlf – frictional load torque reduced to the motor shaft [Nm], γ – angle of rotation of rotor [rad]. It is a base for the whole model, which is created by developing models of all components included into the above equation. Model of mechanism

Model of control unit

Mechanical load

Control signals

Model of transmission 1

Electrical load

Model of motor

Reduced load

Motor torque

Model of rotational movement

Feedback signals

Angle of rotation of output shaft

(6)

where: γ – angle of rotation of rotor [rad], γu – instantaneous stable balance position of rotor [rad], Zr – number of teeth of rotor. Application of the selected motor model is dependent not only upon its credibility but also upon availability of its coefficients. An example of difficulties of such kind utilizing of the “idealized” stepping motor model is. Even renown manufacturers in their catalogues do not refer to this simplest model – do not publish values of coefficient of electromagnetic damping Dm, necessary for proper simulation.

Fig. 6. Concept of modelling Modelling of driving motor The driving motor is represented in the model as its torque Mm as well as mass moment of inertia of rotor Jm. Torque Mm is an electromagnetic torque Me [Nm] diminished by internal losses of motor M m = M e − ∑ M si

2012

Modelling of control unit and feedback loops A typical way of modelling control units is developing adequate functions of control voltage or voltages being responsible for generating of electromagnetic torque. In the simplest case it can be an undisguised function of time

Angle of rotation of rotor

Model of transmission 2

δ = Z r [γ − γ u (t )]

N° 2

(2)

u = u (t ) .

(7)

Drive systems of mechatronic devices usually operate with closed feedback loops either of position or velocity type depending upon main function of analysed assembly. Usually it is then sufficient to apply surrogate models of control units available in various publications e.g. [4] and frequently expressed as transmittances. It also has to be emphasised that modern controllers for stepping motors operate with very sophisticated procedures, usually not sufficiently described and therefore it is far difficult to prepare a credible model of such device.

where Ms denotes torque of losses inside the motor [Nm]. Electromagnetic torque is a function of control voltage or voltages. In case of direct current motor it is given by simple linear relation [5] M e = KT ⋅ i

(3)

where: KT – torque constant [Nm/A], i – motor current [A]. Motor current is determined using voltage equilibrium equation [5]

u = Rt ⋅ i + K E ⋅ ω + L

di dt

(4) in which: u – control voltage [V], Rt – terminal resistance [Ω], L – winding inductance [H], KE – back EMF constant [V•s/rad], ω – angular velocity of rotor [rad/s]. In case of stepping motor the so called “idealized” model is usually used. Its electromagnetic torque is a function of discrepancy angle between magnetic axis of rotor and stator and is given by the equation.

M e = − M max sin (δ ) − Dm

dγ dt

(5)

where: Mmax – maximum torque of motor [Nm], Dm – coefficient of electromagnetic damping [Nm•s/rad], δ – discrepancy angle of position of the rotor in relation to the axis of electromagnetic field of stator [rad]. In hybrid stepping motors the discrepancy angle is calculated as 44

Articles

Modelling of transmission From the point of view of modelling transmission system plays two separate functions: – reduction of mechanical loads related to the driven mechanism to the motor shaft, – reduction of velocity of motor shaft to mechanism. Reduction of torques and reduction of moments of inertia is performed using classical relations

Ml =

J M mech , J l = mech , η p ip i 2p

(10)

in which: Mmech – load torque resulting from mechanism [Nm], ip – gear ratio, Jmech – moment of inertia of rotating parts of mechanism [kgm2], ηp – efficiency of a gear. Reduction of velocity is given by formula coming straight from definition of gear ratio

ω mech =

ω ip

(11)

where ωmech denotes angular velocity of output shaft of a gear [rad/s]. Depending upon type and quality of gear its ratio as well as efficiency may vary, particularly cyclically in function of angle of rotation. In this case it is justified to use models created by expansion into Fourier series experimentally registered relations between torque as well as kinematic accuracy and angle of rotation [4].

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Modelling of a mechanism Because of adopted concept of modelling the driven mechanism is represented by torques and moment of inertia, which can be functions of time, angle of rotation, velocity or other quantities M mech = f (t , γ ,ω , ...) , J mech = g (t , γ ,ω , ...) . (12) In Fig. 7 simulation model of the actuator developed in Matlab-Simulink software environment is presented.

Fig. 9. Results of simulation of a linear actuator driven by stepping motor

4. Designing of mechanical subsystem

Fig. 7. Simulation model of the actuator driven by a hybrid stepping motor

When designing mechanical subsystem, it becomes a common practice to develop its 3D model aimed at avoiding geometrical collisions between parts and used for calculation of distribution of various space dependent quantities such as: temperature, stress or strength (Fig. 10).

The role of CAD 3D programs can be quite extensive at this stage for it is a source of reliable mechanical data to be included in the model, such as: masses, moments of inertia, location of gravity centres and geometrical parameters. Special modules of 3D software are used for this purpose (Fig. 8).

Fig. 10. FEM strength analysis can be based on forces calculated during dynamic simulations

Fig. 8. Dialog window of part properties. In physical properties tab one can describe material the part will be made of. On this basis parameters such as mass, gravity centre, mass moment of inertia are calculated Typical software for simulation of dynamic systems produces time responses of the system output quantities as well as quantities that are not visible or not measurable in a real device, for instance discrepancy angle in a stepping motor (Fig. 9).

Fig. 11. Three dimensional models of parts and assemblies are used to create technical drawings Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

Design process has to be ended with technical drawings of the whole device. When 3D CAD software is used virtual prototype of the device is being created. This prototype represents geometry, material and physical properties that real device will have. It assures elimination of mechanical collisions between parts and is a base for developing set of standard 2D drawings (Fig.11).

5. Summary and conclusions

The software for various kinds of modelling becomes still more and more compatible, giving opportunities to exchange data between different packages. Analysis of conducted modelling and simulations allows authors to end with some conclusion upon reliable use of such tools in order to keep engineers conscious of actual role of modelling. First of all, it is worthful that possibility of data exchange between engineers software constantly spreads. That makes modelling easier and simulation results more accurate. During designing mechanical subsystems of mechatronic devices, significant role plays three dimensional modelling bounded both with geometry of assemblies under design and with other quantities related to their location in space. Development of actuators and sensors requires in most cases time domain simulations. Models of lumped parameters described by ordinary differential equations are commonly used.

Authors

Jakub Wierciak* – Institute of Micromechanics and Photonics, Warsaw University of Technology, św. A. Boboli 8 Warszawa, 02-525, Poland, j.wierciak@mchtr.pw.edu.pl Ksawery Szykiedans – Institute of Micromechanics and Photonics, Warsaw University of Technology, św. A. Boboli 8 Warszawa, 02-525, Poland, k.szykiedans@mchtr.pw.edu.pl Aleksandra Binder-Czajka – Graduate of Micromechanics on Faculty of Mechatronics at Warsaw University of Technology. aleksandra.binder@gmail.com *Corresponding author

References:

[1] M. Gawrysiak, „Mechatronika i projektowanie mechatroniczne”, Politechnika Białostocka. Rozprawy naukowe, no. 44, 1997, Białystok, Poland. (in Polish) [2] B. Heimann, W. Gerth, K. Popp, „Mechatronika. Komponenty, metody, przykłady”, PWN: Warsaw, 2001. (in Polish) [3] R. Isermann, „Mechatronic Systems – Fundamentals”, Springer, 2005. [4] T. Kenjo, C. Nagamori, Dvigateli postojannogo toka 46

Articles

VOLUME 6,

N° 2

2012

s postojannymi magnitami. Énergoatomizdat, Moskva 1989. (in Russian) [5] B. Mrozek, Z. Mrozek, Matlab i Simulink. Poradnik użytkownika. Wyd. Helion. Gliwice 2004. (in Polish) [6] G. Pelz, “Mechatronic systems. Modelling and simulation with HDLs”, John Wiley and Sons: Chichester, 2003. [7] J. Wierciak, “An algorithm for designing mechatronic systems”. In: Rohatyński R. (ed.), Design methods for industrial practice, Oficyna Wydawnicza Uniwersytetu Zielonogórskiego: Zielona Góra 2008, pp. 255-260. [8] J. Wierciak, „Modelowanie elektrycznych układów napędowych urządzeń precyzyjnych”. In: XV Sympozjum Modelowanie i Symulacja Systemów Pomiarowych, Krynica, Poland, 18th-22th September 2005, pp. 239-247. (in Polish).

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

The Process of Aluminium Moduls Warming in the Car Industry Submitted 26th June 2011; accepted 23rd September 2011

Jaroslav Mlýnek, Radek Srb

Abstract:

This paper concerns the heating of aluminium moulds in the car industry. The moulds are intended for the production of artificial leathers. The mould is sprinkled with a special powder and is subsequently warmed by infra heaters located above the mould. It is necessary to ensure approximately the same heat intensity radiation on the surface mould, and in this way, the same material structure and colour of the artificial leather. The mould surface is described by its elementary surfaces. A producer uses moulds of different sizes and they are often very rugged. We must keep track of possible collisions of heater locations. We have used a genetic algorithm to optimize heater locations for a given mould. In this article, we will focus on the calculation procedure of thermal radiation intensity on the mould surface and on the determination of the average aberration of radiation intensity for particular heater locations, which is an important part of our genetic algorithm. Keywords: intensity of heat radiation, experimental measurement of radiation intensity, interpolation in multidimensional space, software implementation

1. Introduction

This article is focused on the technical problem of aluminium mould warming in the car industry. A mould weighs approximately 300 kg and is used for the production of artificial leathers in the car industry (e.g. the bonded artificial leather on a car dashboard and the surfaces of plastic parts in the car interior). The mould is sprinkled with a special powder and is subsequently warmed by infra heaters located above the mould to a temperature of

250°C. It is necessary to ensure approximately the same heat intensity radiation on the whole surface of mould. In this way, the same material structure and colour of artificial leather is ensured. The mould surface is described by its elementary surfaces. The moulds used by a producer may be of different sizes and they are often very rugged. We assume that the same type of heaters (in practical problems usually from 100 to 150 heaters) are used for heat radiation. An infra heater has a tubular form (see Fig. 1) and its length is usually between 15 and 25 cm. The heater is equipped with a mirror located above the radiation tube, which reflects back heat radiation in the adjusted direction. The suitable positioning of heaters above the mould has been done by a technician on the basis of experience and is very labour-intensive and time-consuming. We have used a genetic algorithm to optimize the position settings of heaters. In this article, we will focus on the calculation of thermal radiation intensity on the mould surface and the determination of average aberration radiation intensity for particular heater locations, which is an important part of our genetic algorithm. The experimentally measured values of heat radiation intensity in the surroundings of an infra heater were used during the calculation of radiation intensity on the surface of the mould. A model of the radiated mould will be described in more detail in the following chapter.

2. Model of heat radiation

We will describe the model of heat radiation on the mould surface when using infra heaters. We will assume Euclidean space E3 with a coordinate system (O; x1 , x 2 , x 3 ) . The mould surface is described by the elementary surface p j , where 1 ≤ j ≤ N (i.e. mould surface P is described by N elementary surfaces). We assume that ∪ p j = P is true and int p i ∩ int p j = Ø for 1≤ j ≤ N

i ≠ j . Every elementary surface is presented by the following parameters: T T T – its centre of gravity T j =  x1 j , x 2 j , x 3 j  ;  

Fig. 1. Philips infra heater with 1000W capacity

– the unit outer normal vector v j = ( β j , ω j ) , where positively oriented angle β j ( 0 ≤ β j < 2π ) determines the angle size of the positive part of axis x1 and the vertical projection of the outer normal vector to a plane given by axes x1 and x 2 (ground plane), the angle ω j ( 0 ≤ ω j ≤ π / 2 ) determines the angle size of the outer normal vector v j and axis x 3 (see Fig. 2); – the area of elementary surface s j . Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

2012

r – vector of heater axis

N° 2

d – heater length

o – heater axis

u – vector of radiation direction

x1 v – outer normal vector of elementary surface at gravity centre Tj = (xTj1, xTj2, xTj3)

ωj

Fig. 3. Schematic representation of the heater βj

 x T j 1 

T x2 j

T , x 3 j  

Fig. 2. Parameters determining the elementary surface pj Every elementary surface p j is then defined by 6 parameters: T T T p j :  x1 j , x 2 j , x 3 j , β j , ω j , s j  , 1 ≤ j ≤ N .  

(1)

We assume all the heaters used have the same capacity and are the same sort of heaters. Every heater is represented by abscissa d[m] in length. The location of the heater is described by the following 6 parameters: – coordinates of the heater centre S = [ x1S , x 2S , x 3S ];

(

)

– radiation direction vector u = x1u , x 2u , x 3u , we assume the component x3u is negative (i.e. the heater radiates “down”), then the coordinate x3u of unit vector u is explicitly allocated; – the unit vector r of the heater axis is given by the angle ϕ ( 0 ≤ ϕ < π ), we define the vertical projection of vector r to a plane given by the axes x1 and x 2 vectors, the angle size ϕ is given by this projection and positive part of axis x1 , vectors u and r are orthogonal. The location of every heater Z is described by the following 6 parameters:

(

)

Z: x1S , x 2S , x 3S , x1u , x 2u , ϕ .

(2)

Then the location of M heaters is described by 6M parameters. An infra heater is schematically demonstrated in Fig. 3.

3. Radiation intensity determination in heater surroundings

We will need to know the radiation intensity in heater surroundings to calculate the radiation intensity on the mould surface. We can’t use radiation point source properties for this determination. A mirror is placed above the heater and reflects the radiation back in the adjusted direction (see Fig. 1). The distribution function of radiation intensity in the heater surroundings is also not known. Thus an experimental measurement was taken of the radiation intensity of a given type of heater for selected points in the vicinity 48

Articles

of the heater with the aid of a checking member. We accomplish the determination of the radiation intensity at the defined point with the help of interpolation in multidimensional Euclidean space. 3.1. Experimental measurement of radiation intensity We will assume the heater location Z defined by relation (2) is given by parameters Z: (0, 0, x 3S , 0, 0, 0) , i.e. the centre of a heater S lies on the positive part of axis x 3 , union radiation direction vector u = (0, 0, − 1) and vector of heater axis r = (1, 0, 0) (in Euclidean coordinates). The experimental measuring of the radiation intensity in the vicinity of heater Z was accomplished at node points a = [a1 , a 2 , a 3 , a 4 , a 5 ] of the regular net in accordance with the setting of the elementary surface p j given by relation (1) (here, points [a1 , a 2 , a 3 ] lie on a few planes parallel with the plane given by axes x1 and x 2 ). Coordinates a 4 , a 5 determine the location of the unit outer normal vector (on an imaginary elementary surface with the centre of gravity at point [a1, a2 , a3 ] ). The heater radiation intensity on the elementary surface depends not only on the distance of the elementary surface from the heater, but also on the direction of the outer normal vector in the centre of gravity of the elementary surface. The measurements were only performed in selected points a = [a1 , a 2 , a 3 , a 4 , a 5 ] for which a1 ≥ 0 and a 2 ≥ 0 are true (with regard to the heater location and symmetry of the radiation) and 0 ≤ a 3 < x 3S , 0 ≤ a 4 < 2π , 0 ≤ a 5 ≤ π / 2 , where x3S is a component of heater centre S. We will assume that the radiation intensity I (b) for point b lies within the “hyperrectangle” net points determined by heater Z. We will use the linear interpolation of the function of 5 variables. We will assume that the point b = [ x1b , x 2b , x 3b , x 4b , x 5b ] holds a j , i j ≤ x bj ≤ a j , i j +1 for 1 ≤ j ≤ 5 . Let us denote m j =

x bj − a j , i j

for

a j , i j +1 − a j , i j

1 ≤ j ≤ 5 . Then it holds for the interpolation value at point b of the radiation intensity I (b) of heater Z I (b) = I ( x1b , x 2b , x 3b , x 4b , x 5b ) =

i 1+1

i 5 +1



∑ ... ∑ I (a

k1 = i 1

k5 = i 5

(3) 

∏ H (l, k − i )

1, k1 , a2, k 2 , a3, k 3 , a4, k 4 , a5, k5 )

l =1

Journal of Automation, Mobile Robotics & Intelligent Systems

and where H (l , 0) = 1 − ml and detail e.g. in [1]).

VOLUME 6,

H (l , 1) = m l (in more

3.2. Interpolation of radiation intensity in general Now we assume a general case of the heater location (we only assume that it holds true for the value of component x3u of the radiation direction vector x 3u < 0 , i.e. the heater radiates “down”). In this case we accomplish a transformation of coordinate axes x1 , x 2 , x 3 , x 4 , x 5 and we transfer this problem to the problem described in paragraph 3.1. Here, at a transformation origin of coordinates O of the Euclidean space E3 lies in the intersection of line k , which contains the centre of heater S and the direction line is given by the radiation direction vector u and plain ρ which contains point [ x1b , x 2b , x 3b ] and is orthogonal to line k. The transformation of axis x1 is defined by the origin O and the vector of heater axis r , axis x 3 is identical with line k and axis x 2 is given by the origin O and vector n , where n is determined by the vector product of the vectors r and −u (in more detail e.g. in [2]) and is given by the relation  xu n = (−u ) × r =  − 2r  x 2 

x 3u x 3r

x1u

x 3u

x1r

x 3r

,−

x1u x1r

x 2u x 2r

 .   (4)

The transformed orthogonal system of coordinates is positively oriented. We establish 1 2 , x3 new coordinates for the heater centre S and point [ x1b , x 2b , x 3b ] in the transformed system. Consequently, we transform the coordinates of component x 4b and x 5b (which describe the location of the outer normal vector at the point [ x1b , x 2b , x 3b ] ) with regards to the new coordinate system O; x 1 , x 2 , x 3 . In this way, the general location of a heater is transferred to the problem described in paragraph 3.1.

(O; x , x

)

(

)

4. Calculation of radiation intensity on the surface mould

In this chapter we will describe a procedure radiation intensity calculation on a particular elementary surface of a mould for given location of heaters. Then we can express the average difference of radiation intensity on the mould surface. We denote L j a set for all the heaters radiating on the j-th elementary surface for the defined locations of the heaters, 1 ≤ j ≤ N . We will further denote Ij l [W/m2] radiation intensity of the l-th heater on the j-th elementary surface. Then the total radiation intensity I j on the j-th elementary surface is defined by the relation (in more detail e.g. in [3])

I j = ∑ I jl .

(5)

l ∈L j

We will denote I opt the recommended radiation intensity on the mould surface by the producer. We can determine the difference F j of the radiation intensity on the j-th elementary surface I j ( 1 ≤ j ≤ N ) from the recommended intensity I opt upon the basis of the relation F j = I j − I opt

(6)

N° 2

2012

and the average aberration of radiation intensity F given the relation N

∑F s j

j =1

∑s

(7)

j =1

where we recall that s j denotes the area of the elementary surface p j . The genetic algorithm was used to optimize the locations of heaters. We searched for the minimum of function F in relation to the heater locations. Function F was used as an evaluation function (fitness) for this algorithm.

5. Practical example

We require the results of the radiation intensity calculation on the mould surface and the determination of the average aberration of radiation intensity on the mould surface. A software application was programmed in the Matlab language. The parameters of the heaters were as follows: producer Philips, capacity 1600 W, length 15 cm, width 4 cm. The heater characteristics were measured in 5 levels (planes) located below the heater at distances 5, 10, 15, 25 and 30 cm. Only one quadrant was measured (where values x1 and x 2 were non-negative) at every level. 13 measurements were accomplished in axis direction x1 at 20,1 mm intervals, 24 measurements were accomplished in direction x 2 with 10 mm intervals. The radiation intensity was measured for the small horizontal area (the fictive vertical outer normal vector) and for the small inclined areas (the fictive inclined outer normal vectors). The measurements were accomplished for the declination of the corresponding values β j = 0, π / 2, π , 3π / 2 and ω j = 0, π / 6, π / 3, π / 2 in relation (1). The calculation was accomplished for the aluminium mould (0,6´0,4´0,12 m3) displayed in Fig. 4. and 20 heaters.

Fig. 4. Aluminium mould In the first step of optimization procedure of genetic algorithm we set the initial locations of heaters. The parallel plane with axis x1 and x 2 contains the centres of all the heaters and is in the distance g =10 cm over the T maximum x 3, max of values x 3 j of all centres of gravity T j of the elements of mould surface. The locations of heaters defined by relation (2) is expressed in the form Z : x1S , x 2S , x 3, max + g ,0 ,0 ,0 , i.e. the heaters radiate down and their vectors of the heater axis are parallel with

(

)

Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

axis x1 . We apply a procedure of genetic algorithm and we obtain optimized locations of heaters. The calculated radiation intensity is described for 5 randomly chosen elementary surfaces in Tab. 1, where elementary surfaces are defined in accordance with relation (1). Tab. 1. Calculated radiation intensity Elementary surface ( p j )

Total radiation intensity ( I j ) [kW/m2]

(-0.2781, -0.1513, 0.0031, 3.1010, 0.7546, 0.00019)

38.97

(0.2326, -0.0563, 0.0700, 4.7372, 0.1886, 0.00017)

45.76

(-0.0008, 0.0189, 0.0838, 5.1672, 0.0714, 0.00025)

35.92

(-0.1881, 0.1934, 0.0581, 1.3879, 0.6898, 0.00018)

39.92

( 0.0966, 0.0280, 0.0665, 5.9614, 0.0364, 0.00017)

43.58

VOLUME 6,

N° 2

2012

The calculated radiation intensity for individual elementary surfaces is demonstrated in Fig. 5 (the lighter shade of grey color denoting a higher heat radiation intensity). The example described is illustrative. It is obvious that it is necessary to increase the number of heaters and to accomplish optimization of their locations.

Acknowledgements This work was supported by MPO project No. FRTI1/266.

Authors Jaroslav Mlýnek* – Department of Mathematics and Didactics of Mathematics, Faculty of Sciences, Humanities and Education – Technical University of Liberec, Studentská 2, Liberec 1, 461 17, Czech Republic, jaroslav.mlynek@tul.cz Radek Srb – Institute of Mechatronics and Computer Engineering, Faculty of Mechatronics, Informatics and Interdisciplinary Studies – Technical University of Liberec, Studentská 2, Liberec 1, 461 17, Czech Republic, radek.srb@tul.cz *Corresponding author

References

Fig. 5. Graphical representation of the calculation of radiation intensity The recommended intensity is I opt =44[kW/m2], the average aberration of radiation intensity F is equal to F = 11,21[kW/m2].

Articles

[1] Antia H. M., Numerical Methods for Scientists and Engineers, Birkhäuser Verlag: Berlin, 2002, pp. 114153. [2] Budinský B., Analytical and Differential Geometry, SNTL: Prague, 1983, pp. 57-75. (in Czech) [3] Linhard IV J. H., Linhard V J. H., A Heat Transfer Textbook, http://web.mit.edu/lienhard/www/ ahtt.html.

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Analyses Of Micro Moulding Process Of The Microelements From Ceramic Powders Submitted 27th June 2011; accepted 16th September 2011

Andrzej Skalski, Dionizy Biało, Waldemar Wisniewski, Lech Paszkowski

Abstract: The article discusses the issues related to moulding of micro-elements from powder materials and it covers the first stage of the process – injecting to the mould cavity and filling the micro-channels. The material, which constituted the composition of a special thermoplastic binder and ceramic powder, was injected. The binder consisted of paraffin, polyethylene, wax and stearic acid. The nanometer powders from Al2O3 ceramic material with granularity of 660 nm and 135 nm and irregularly shaped particles were used. Different loading of composition by powder were used. The analyses of the impact of injection parameters, such as the mould’s temperature and the temperature of the material, on the quality of filling the micro-mould cavity with the material, were presented. The special moulding insert with micro-channels was made to perform the analyses to that effect with width from 50 to 1000 mm. The presented results of filling the micro-channels indicate considerable influence of the mould’s temperature and the cross-section of microchannels. Slight impact of the material’s temperature was observed; however, this factor does not have a considerable influence on filling the channel. The obtained information was used in the experiments of injection the samples for bending tests and tensile tests, and the shapes in the form of toothed wheels. Keywords: injection moulding, micro-elements, ceramic powders

1. Introduction

In the recent decade we have observed intensive development of processes involving production of microproducts and micro-elements related to development of Micro Technologies Systems (MTS). The Micro Technologies Systems combine micro-electronics with many other micro technologies e.g. mechanical, optical, chemical technologies, etc. The objective is to fully concentrate various functions in a single miniaturized product. It should be noted that the dimensions of the produced structures or their components are expressed in micrometres. Previously used methods of production of micro-elements are limited to selected groups of materials. This pertains to the following processes: LIGA, laser processing, erosion, etching, etc [1, 2, 4, 5]. A large problem involves adaptation of well-known technologies from the macro scale to the micro scale. It turns out that this is not always possible or does not give the desired effects because the dimensions of the elements are less than 1 mm. One of the more promising processes is the

process of moulding micro-elements through injecting. Such a moulding method allows the production of microelements with complex shapes, with high accuracy, in large series, and in a manner that is efficient and competitive compared to other production methods. The micro injection moulding is based on the currently used method of making products from thermoplastic polymers through injection. It should be emphasized that the first analyses and applications focused on microelements manufactured from such materials [3, 5, 6, 7]. The micro-elements injected with metal and ceramic powders are designated for operation in harsher thermal and mechanical conditions than elements made from plastics. In this technology, the problem involves not only making the injection micro-mould but also selecting the parameters for the entire injection process. No tests have been developed which would allow us to unequivocally determine the material’s suitability for production of micro-elements by the injection method. Such a test, which is known as the spiral test, exists for macro-elements. The presented article concerns the analyses of the first stage of micro-element production from ceramic powders. The entire process of micro-element production includes preparation of the mass consisting of powder and special thermoplastic binder, injection, binder removal (debinding) and sintering [8, 9, 10]. The results of filling the channels and the micro-cavity of moulds were presented along with the examples of the obtained shapes. In this part of the article, the impact of temperature of the injected mass and the mould’s temperature was discussed.

2. Materials

The injection moulding uses the masses which are the compositions consisting of the thermoplastic binder and the specified micro-powder. In the presented analyses, Al2O3 ceramic micro-powders with granularities of 0.66 μm and 135 nm were used. The description of the individual powders is included in Table 1 and the photographs of the selected powders are presented in Figure 1. Table 1. Powder description Powder type

Symbol

Manufacturer

Average granularity

Particle shape

Al2O3

Martoxide

0.66 μm

Irregular

Al2O3

Tamei Chemicals Co, Ltd

135 nm

Irregular

Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

mould has a built-in heater with a temperature regulation system as well as a cooling system. The moulding insert is presented in Figure 2. The mould’s injection nozzle directed the mass centrally to the moulding insert, from where it continued to flow to micro-channels located along the radii. Dimensions of the channels are presented in Table 2. Table 2. Micro-channel dimensions Width D, mm

Fig. 1. Al2O3 ceramic powders: M, 660 nm (a) and TM, 135 nm (b) The composition of the binder is as follows: LD polyethylene 20% paraffin 69% Carnauba wax 10% stearic acid 1% Preparation of the injection mass consisting of powder and binder was carried out in a type 2Z mixer with the heating mantle at a temperature of 125°C for a time of 1 hour. Such a period of time was sufficient to prepare a homogeneous mass. For tests involving injections and filling the microchannels, the mass with the powder content of Vp = 50 and 55% by vol. was used. Such high Vp values are necessary to produce micro-elements with complicated shapes and very small structural details. • • • •

0.2

0.3

0.4

0.5

0.8

Depth h, mm 0.17

0.27

0.35

0.44

0.71

0.9

CrossS, mm2 0.033 0.077 section

0.13

0.21

0.54

0.84

The mould, which was designed in such a way, allows us to simultaneously obtain a series of data from one injection cycle because it gives the information on the process of filling several micro-channels with different cross-sections at the same time. The following technological parameters of the injection process were used: Temperature of the mass Tw – 125, 150 and 170°C Temperature of the mould Tf – 25, 40, 50, 60, 70 and 80°C Pressure p – 60 MPa Powder content in the mass Vp – 50 and 55% by vol.

4. Results

Figure 3 presents the chart describing the dependencies, which are typical of the discussed analyses, between the micro-channel filling process (mass inflow distance, L) and the temperature of the mould Tf. The chart pertains to the mass containing the ceramic powder with granularity of 0.135 µm with powder content in the mass of Vp = 55%. The obtained results show that as the temperature Tf increases, the flow distance of the material increases.

3. Tests

A special injection mould with micro-channels with cross-sections of 0.033; 0.077; 0.13; 0.21; 0.3; 0.54 and 0.84 mm2 was designed for the analyses. This was necessary because there are no standardized tests for analyzing the micro-moulding, and the spiral test, which is commonly used in moulding the macro-elements from thermoplastics, cannot be used in the case in question. The Fig. 3. Dependency among the inflow distance L, the temperature of the mould Tf and the cross-sections of microchannels for mass with TM powder (135 nm) with Vp = 50% and Tw =125°C

Fig. 2. The moulding insert for evaluating the inflow distance of the mass in the mould. 1 – micro-channel, 2 – injection point, 3 – cross-section view of the channel 52

Articles

The chart shows the temperature threshold above which it is easier for the material to fill the channels. For larger micro-channels, this temperature is approx. 45°C; for smaller ones it is slightly higher. The cross-section of the channel is an important factor. As it increases, the

Journal of Automation, Mobile Robotics & Intelligent Systems

inflow distance considerably increases. In such case, the stream of material flowing in the channel has a relatively higher volume and therefore a higher weight. Consequently, it cools off at a slower pace when in contact with the mould, which is colder. According to literature [6], under extreme conditions, when the length of the feeding channels is large and the micro-moulds have complicated shapes, in order to facilitate the flow of mass the so-called “Variotherm” process is used during the injection, in which the temperature of the mould reaches up to 100°C. Figures 4 presents dependencies between the inflow distance L (mm) and the temperature of the mould Tf for various cross-sections of micro-channels and both injected materials.

VOLUME 6,

N° 2

2012

efficacy of filling the micro-channel and also filling the mould. In such a situation, it is possible to reduce the temperature of the mould while preserving the satisfactory efficiency of filling it – Figure 6.

Fig. 5. Dependency of the inflow distance L of the mass on the cross-section of the micro-channel for mass with powder TM with Vp = 50% for mould temperature of Tf = 25°C and mass temperature of Tw =125, 150 and 170°C.

Fig. 4. Dependency between the inflow distance L of mass and the temperature of the mould Tf at Vp = 50% and Tw =125°C a – in the micro-channel with cross-section of S = 0.033 mm2 , b – in the micro-channel with crosssection of S = 0.21 mm2 If the temperature is lower than 50°C, the inflow distance in the micro-channel with cross-section of 0.033 mm2 is similar for the individual materials, and it amounts to 1 and 2.5 mm. After this value is exceeded, the impact of the type of material is observed. The mass with finer powder fills the micro-channels more efficiently, and as the temperature increases, the inflow distance rapidly increases. The mass with powder M with granularity of 0.66 µm fills the micro-channel less efficiently. As the mould temperature grows, the inflow distance increases to the specified level, after which it has approximately the same value despite increase of temperature Tf. This dependency is similar for the micro-channel with cross-section of 0.21 mm2. Increasing the cross-section of the micro-channel resulted in extension of the inflow distance, and the curves for the individual granularities show the same trend. After increasing the cross-section of the channel almost six times, the mass additionally increases the inflow distance and it begins to fill the channel at a lower mould temperature (room temperature). The next technological parameter, which may result in improvement of parameters of the produced elements, is the temperature of the injected material (Tw). Figure 5 presents the inflow distance figures as a function of micro-channels’ cross-sections for the temperature of the injected material of 125, 150 and 170°C. The injected mass contained the ceramic powder with granularity of 135 nm. It may be observed that, as it was predicted, the increase of the material’s temperature will improve the

Simultaneous impact of temperature of the mass and temperature of the mould on the inflow distance in channels with various cross-sections is presented in Figure 6. The obtained curves confirm that the conclusion is right. An increase in mass temperature Tw results in the inflow distance increasing, especially for larger cross-sections of micro-channels. In addition, increasing the mould temperature facilitates the flow of mass, which is very important for small cross-sections of micro-channels when the increase of Tw itself does not have a sufficient impact on filling the micro-channels or micro-moulds.

Fig. 6. Dependency of the inflow distance L of the mass on the cross-section of the micro-channel for mass with powder TM with Vp = 50% for mould temperature of Tf = 25 and 50°C and mass temperature of Tw = 125 and 150°C It should be remembered that the temperature of the material and the mould should not be excessively increased due to the possibility of degradation of the mass before its injection into the mould cavity. The presented results of the analyses were verified through production of micro-elements in the form of bars for bending tests and samples for tensile tests, as well as miniature toothed wheels. The examples of such elements are presented in Figures 7 and 8.

Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Authors

Fig. 7. Micro-samples for tensile tests moulded at various temperatures of the mould and with the following injection conditions: Tw =115 °C, p = 60 MPa and Vp = 60% It is visible that filling of the micro-mould in not complete as the mould temperature is too low.

Fig. 8. Examples of moulded micro-elements before sintering a – bars for bending tests: 1x1x10, b – bar for bending tests: 2x2x12 mm c – sample for tensile tests 0.5x0.5x5 mm d – toothed wheel

5. Conclusions

The following conclusions may be drawn on the basis of conducted analyses of filling the micro-channels through the process of injection moulding of micro-elements from ceramic powders: • The proposed analysis method, which uses a special moulding insert with many channels with different cross-sections, has turned out to be very useful in evaluating the behaviour of the masses consisting of ceramic powder and binder during the process of micro-injecting. • The greater the cross-section of the micro-channel, the longer the inflow distance of the mass with the set Tw, Tf and p. • The most important parameter, which determines the course of filling the micro-channels, is the mould temperature. It has to be considerably higher than in the case of moulding of macro-elements with ceramic macro-powders. • The finer the powder particles, the easier it is for the mass to flow to micro-channels. An increase in temperature of the injected mass results in improvement of filling the micro-channels and micromould cavity.

Articles

Andrzej Skalski*, Dionizy Biało, Lech Paszkowski, Waldemar Wiśniewski – Institute of Metrology and Biomedical Engineering, Warsaw University of Technology, ul. sw. A. Boboli 8, 02–525 Warsaw, Poland *Corresponding author, askalski@mchtr.pw.edu.pl

References

[1] Oczoś K. E., “Kształtowanie mikroczęści – charakterystyka sposobów mikroobróbki i ich zastosowanie”, Mechanik, no. 5-6, 1999, pp. 309-327. (in Polish) [2] Mrugalski Z., Rymuza Z.: “Mikromechanizmy”, Pomiary, Automatyka i Kontrola, no. 6, 1998, pp. 4-9. (in Polish) [3] Gad-el-Hak M., The MEMS Handbook, 2nd Edition, Virginia Com. University, USA, 2005. [4] Poiter V. et al., “Micro Powder Injection Moulding”. In: EURO PM2000, 18th-20 th Oct. 2000, Munich, Germany, vol. PIM, pp. 259-264. [5] Hasselbach J. et al., “Investigation on the International State of Art of Micro Production Technology”. In: Euspen Int. Topical Conf., Aachen, Germany, May 19-20, 2003. pp. 11-18 [6] Piotter V. et al., “Micro Injection Molding of Components for Microsystems. In: 1st Euspen Topical Conf. on Fabrication and Metrology in Nanotechnology, Copenhagen, May 28-30, 2000, vol. 1, pp. 182-189. [7] D. Biało, A. Skalski, L. Paszkowski, “Selected Problems of Micro Injection Moulding of Microelements”. In: Recent Advances in Mechatronics, Joint publ. ed. by J. Jabłoński, Springer-Verlag, Berlin Heidelberg 2007, pp. 370-374. [8] Benzler T. et al., “Fabrication of Microstructures by MIM and CIM”. In: PM World Congress PIM. Granada, Spain, 1998, vol. 3, pp. 9-14. [9] Zauner R., Korb G., “Micro Powder Injection Molding for Microstructured Components”. In: PM Plansee Seminar, Reute, Austia, 2005, PL 5, pp. 59-68. [10] D. Biało, “Metoda formowania wyrobów z proszków poprzez wtrysk”. In: N-T Conference. Postępy w Elektrotechnologii, Szklarska Poręba, Poland, 14th-16th Sept. 1994, vol. 1, pp. 235-240. (in Polish)

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Using Reverse Engineering In Archaeology: Ceramic Pottery Reconstruction Submitted 10th October 201 ; accepted 12th November 2011

Calin Neamtu, Sorin Popescu, Daniela Popescu, Razvan Mateescu

Abstract: The present paper presents a model for using CAD software and reverse engineering methods for the reconstruction of ceramic vessels, as part of the larger field of digital archeology. The case study focuses on the reconstruction of a specific Dacian pottery, namely the “chiup”, used for storing food ad liquids, and the proposed reconstruction method is a graphical one. Keywords: CAD, reverse engineering, digital archeology, ceramic pottery, reconstruction

1. Introduction

In the UNESCO vision, “cultural heritage is our legacy from the past, what we live with today, and what we pass on to future generations”, and in our opinion digital archeology is a way to contribute to this desiderate. Knowledge and understanding of history is an important part of the education of any modern citizen and it has been so even since ancient times. It is believed that the lessons of the past should be passed on to the new generations in order to avoid making the same mistakes and to increase the speed of progress. Engineers can help to save and reconstruct the past with specifics tools and techniques from their domain used in archeology and in connected domains, in an interdisciplinary approach. Historians have been preoccupied for a long time with the reconstruction of historical artifacts, events or Table 1. Reconstruction methods used by historians Studied object

Reconstruction method

Partial artifacts

Laboratory reconstruction with specific methods and tools

Lost artifacts

Drawings, 3D models based on existing information

Events

Reenactments, drawings, computer simulations

Figures

Drawings, 3D models, animations

Media/Information

Interpretation of specialists

Society/Relationships

Reenactments of social situations

Architecture

Renovation of buildings, monuments, etc.

Fig.1. Dolium, left (Museo Archeologico Nazionale della Sibaritide) and Chiup, right (National History Museum of Transylvania Cluj-Napoca) figures, a few of the methods used in this direction are presented in Table 1. In this paper there are presented an algorithm and an experimental method for ceramic pottery reconstruction (starting with fragments) based on reverse engineering techniques and CAD modeling. The paper focuses on a special pottery type: the chiup from the Dacian civilization (Romanian civilization ancestor). The Dacian chiup imitates the Greek Dolia or Pythos vessels, retaining most of their features with small modifications [1], [2]. The chiup is an arched, oval shaped vessel, with its maximum diameter in the superior part of the body, used for grain storage. Some specimens have a 2 m diameter and in order to be filled and used properly had to be buried in the ground or hold on special support, due to the instability resulted from its shape. In proper conditions a great quantity could be stocked, but the degree of instability proportionally increased with the amount of filled grain. These types of vessels were obtained using the wheel, with a special technique necessary for managing the high weight of the raw material, ceramics. The vessels’ walls were shaped on segments being added successively as the previous one dried. The joints were made perfectly, without any visible marks, and the bottom part was attached to the corresponding walls at the final stage of the process. Regarding the dimensions, the vessels diameters move mainly on a scale of 0.50 m to 1.50 m, but there also exist samples which reach a maximum of 2 m. It is a remarkable fact that among the several elements of these vessels there can be identified a clear correlation. It can almost be stated that their development followed some kind of standardization [1], [2]. Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

2. Methodology and set-up

In the authors’ opinion, the results of digital archeology are an electronic file (or application) with a complete description of an historical artifact (3D and other information). In this sense, restoration and digitization should envision at least two possible uses of the virtual artifact: · virtual museums – digitization is focused on the aspect of the 3D artifact which should closely resemble the real one; · research – digitization is focused on the shape and material of the artifact, so that it could be used for comparison with other similar vessels or vessel fragments, from the same or from other studied cultures. Performing artifact reconstruction should be a structured process aimed at achieving the most accurate soluTable 2. Step for the reconstruction of ceramic artifacts No Operation

Artifact analysis – determining the need for reconstruction, time period characteristics, physical characteristics, manipulation constraints, legal issues etc.

Set-up and observations Assembling the reconstruction team and regular team meetings for establishing a common road map and time frame.

Available reverse engineering techniques and equipment compared Establishment of to reconstruction costs and the best reverse engineering techniques accuracy requirements. It is possible that a combination (scanning device, software, environmental of scanning techniques is necessary (regular laser conditions etc.) scanning, texture laser scanning etc.) Scanning device constraints Establishment of the 3D should be observed and the best approach should be scanning strategy determined. Defining and building Fixtures, supports, other the necessary protection means of protection. measures 3D scanning device, scanning techniques, Digitization of the scanning program, CAD artifact fragment (3D software able to process scanning and creation of clouds of points and to the 3D model) communicate with the scanning device. Method and geometrical algorithm for reconstruction, Implementing the choosing reconstruction detailing and finishing of the CAD model in the solution reconstructed state. Additional information will be added to the 3D model: set-up visualization mode Digital artifact and alternative, protect the digital artifact-copyright, etc. Articles

VOLUME 6,

N° 2

2012

tion, i.e. the reconstructed product reproduces faithfully the characteristics of the original. Also, the reconstruction effort should be a team one, with at least the following important functions being represented: · historians and archaeologists: provide essential knowledge about manipulation permissions and restrictions, ensure the historical accuracy of the resulting reconstruction, as well as to put forward solid hypotheses about missing parts based on knowledge of the given historical age; · reverse engineering / CAD specialist: designs and implements the technical solutions for digitizing the partial existing artifact and for “filling in” the missing pieces; · artists / art historians: artists that are specialized in certain time periods can bring the aesthetic perspective in correctly defining the missing parts of different artifacts; · IT specialist: develop virtual artifact (3D model + additional information) for electronic libraries or virtual museums. The authors of this paper propose the following methodology that the interdisciplinary team should follow in order to obtain a 3D model (including reconstructed model) for a ceramic pottery (see Table 2).

3. Case Study

In the following we will present a case study of the proposed methodology for reconstruction in the digital environment of the chiup pottery type, used by the Dacian civilization around the 1st century BCE and 1st century CE

1.1. Artifact analysis

Archeologists have discovered that these types of vessels are mostly found around the capital of the kingdom, Sarmizegetusa Regia [1], [2]. One of the explanations could be provided by the capital region’s geography, a mountainous region, where the usual way to store grain, that of digging holes and reinforce them with clay, like it used to be practiced in the plain areas, could not be applied, so they used pottery vessels instead [1], [2]. Regarding the discovery places, the vessels were discovered in houses, as well as inside the fortress. The presence of a large number of vessels inside the fortress can be considered natural, as the quantity of water and grain needed was higher than for a usual house. Considering the vessels’ dimensions, the physical reconstruction requires a high amount of effort especially if the pieces found are small regarding their size. The existing fragments reconstitute approximately 1/4 of the entire chiup. There are some particularities of these vessels that facilitate the reconstruction using a graphical method: · the archaeologists have realized that these vessels are sort of standardized, therefore establishing a series of relations between the proportions of certain parts of the vessel such as: that between the vessel’s height and the diameter of the entrance is between 1/2 and 1/4, between the height and the maximum diameter the ratio is between 2/1 and 3/2, and between the low end diameter and the height it is between 1/4 and 1/9 [1], [2].

Journal of Automation, Mobile Robotics & Intelligent Systems

another particularity of these vessels is the fact that they are modestly decorated, which makes the reconstruction process difficult, because it does not give the restaurateur important indications about the actual position of the pieces; the decoration elements were placed on the upper side due to the fact that the lower side was in the ground [1], [2]. the vessels were created using the potter’s wheel, which means they have a well-defined symmetry axis.

1.2. Choosing the reverse engineering technique

Even if the vessel has been baked in special ovens, the use of contact scanning can be harmfull for the remaining fragments. Laser scanning will the method be used for digitization, by using a precise laser scanner (Kreon Zephyr) and a portable CMM for obtaining the research model, and a hand-held texture laser scanner (VIUScan from Creaform) for obtaining the model destined for a virtual exhibition.

1.3. Scanning strategy

VOLUME 6,

N° 2

2012

1.4. Protection measures

The following issues have been considered: the object was manipulated with gloves, for fixing fragments plasticine was used, a material softer than the artifact, with no additional help (clamps, glues or other means).

1.5. Digitization of the fragments

The digitization of the vessel’s pieces is done using two laser scanning devices: Kreon Zephyr and Creaform ViuScan. In this phase, a preliminary grouping of the fragments can be done, for the fragments whose position in the vessel structure is obvious. The digitization step ends once the surface or the virtual solid object is generated and can include a step in which the textures are applied [8], [9]. Using the mentioned laser scanning technique, the resulted cloud point (Fig. 3a) has been transformed into a surface (Fig. 3b) on which later on the texture has been applied (Fig. 3c) from the original fragment (Fig. 3e) in case of scaning with Kreon Zephyr, and the fragment scaned using ViuScan with the scaned texture (Fig. 3d).

When using the portable CMM + laser scanner solution, the fixing of the part must expose its surface within the working zone of the CMM. When using the VIUScan texture scanner, which is a self-positioning scanner, positioning targets must be applied on the surface of the fragments, but in order to avoid this, the work team has manufactured the necessary target grid by using transparent foils cut out in the shape of the actual fragments.

Fig. 3. Digitized vessel fragment: a-point cloud; b-surface;, c-textured surface; d-fragment scanned with ViuScan; e-fragment discovered at Gradistea de Munte

1.6. Reconstruction solution

Fig. 2. Target grid and ViuScan

In the chiup case, the reconstruction method is based on rebuilding the profile of the vessel, which is used to generate the entire artifact through an operation that implies rotating the rebuilt profile around the theoretical axis of the vessel [3], [4], [6]. After obtaining the surface for each piece, a series of transversal and longitudinal curves can be determined, this will be used at the next step for grouping the fragments. In figure 4 is represented a fragment with the adjacent curves obtained from the intersection of the surface with the perpendicular plains on the theoretical axis of the vessel, and with the planes that contain this axis. For each of the fragments, there are generated circles with Articles

Journal of Automation, Mobile Robotics & Intelligent Systems

VOLUME 6,

N° 2

2012

Fig. 6. Intermediary stage (left) and final reconstruction

Fig. 4. Generation of transversal and longitudinal curves a maximum and a minimum diameter which will be used in the next phase when grouping the fragments on diameter intervals. The next step in the reconstruction of the chiup is to group the fragments, in this case six segmenting intervals have been established. For each interval, the inferior and superior limits were set. The boundary curves which separate the surfaces are generated for each fragment of the vessel [7]. These are used later on for matching of the components of a certain group of pieces.

Fig.7. Digital artifact in web browser

1.7. Digital artifact

For visualizing and manipulating the virtual artifact, the wrl format is used, containing in a single file both the 3D model and additional information (e.g. short description, links etc.). This format can be accessed with the help of a web browser and it can be easily integrated into a virtual museum.

3. Conclusion

Fig. 5. Matching fragments using boundary curves Generation of the vessel’s profile has to be performed as much as possible in the most complete section. The profile can be generated using more pieces if it is necessary and then, through a projection into a plane, the obtained profiles can be joined in order to achieve the complete profile. Using a simple geometrical operation (revolve) the entire vessel is generated. 58

Articles

The reconstruction of ancient ceramic pottery is a subject of interest approached by many re-searchers [3], [4], [5], [10], [11] and [12] in some cases attempting an automated reconstitution by using special algorithms from the field of image processing. A graphical reconstruction (based on the geometry of the fragments) involves a high work load, but in the situation described in this paper, reconstruction of a chiup, it may be the only method that leads to a satisfactory result, due to the particularities of these vessels, such as the absence of decorations, which makes image processing reconstruction nearly impossible. An interdisciplinary team is necessary in order to carry out such a project, and there exist a series of limitations which, in some conditions, cannot be overcome. For example, when only smaller fragments exist, it may be

Journal of Automation, Mobile Robotics & Intelligent Systems

difficult to distinguish the symmetry axis of the vessel, making it impossible to establish the position of the fragment. Yet this problem may occur in any re-construction method, even in the conventional one when the vessel is complete. Using reverse engineering as an instrument for archaeology may lead to obtaining virtual artifacts that can be used for experimental research, if their digitization is precise, as well as for virtual exhibitions, if the artifact is digitized together with its texture. Acknowledgements This paper was supported by the project “Progress and development through post-doctoral research and innovation in engineering and applied sciences – PRiDE – Contract no. POSDRU/89/1.5/S/57083”, project cofounded from European Social Fund through Sectorial Operational Program Human Resources 2007-2013.

AUTORS CălinNeamțu*, Department of Design Engineering and Robotics, Technical University of Cluj-Napoca, bd. Muncii 103-105,Cluj-Napoca, 400641, Cluj, Romania, calin.neamtu@muri.utcluj.ro Sorin Popescu, Department of Design Engineering and Robotics, Technical University of Cluj-Napoca, bd. Muncii 103-105,Cluj-Napoca, 400641, Cluj, Romania, sorin.popescu@muri.utcluj.ro Daniela Popescu, Department of Design Engineering and Robotics, Technical University of Cluj-Napoca, bd. Muncii 103-105,Cluj-Napoca, 400641, Cluj, Romania, daniela.popescu@muri.utcluj.ro *Răzvan Mateescu, National History Museum of Transylvania, Cluj, Str. C. Daicoviciu, no. 2, Cluj-Napoca, 400020, Cluj, Romania, razvanmateescu@yahoo.com

VOLUME 6,

N° 2

2012

[6] X. Tang, “A sampling framework for accurate curvature estimation in discrete surfaces”, IEEE Trans. Visual. Comput. Graphics, vol. 11, no. 5, 2005, pp. 573-583. [7] J. McBride, “Archaeological fragment re-assembly using curve matching,” M.S.thesis, Brown Univ., Div. Eng., Providence, RI, Sept. 2002. [8] M. Sağiroğlu, A. Erçil, “A texture based approach to reconstruction of archaeological finds”. In: Proc. of 6th Int. Symp. Virtual Reality, Archaeology, and Cultural Heritage, 2005, pp. 137-142. [9] P. Besl, N. McKay, “A method for registration of 3-d shapes”, IEEE Trans. Pattern Anal. Mach. Intell., vol. 14, no. 2, 1992, pp. 239-256. [10] G. Uçoluk, I.H. Toroslu, “Automatic reconstruction of broken 3-D surface objects,” Comput. Graph., vol. 23, no. 4, 1999, pp. 573-582. [11] A. Gilboa, A. Karasik, I. Sharon, U. Smilansky, “Towards computerized typology and classification of ceramics,” J. Archaeological Sci., vol. 31, no. 6, 2004, pp. 681-694. [12] W. Kong, On solving 2-D and 3-D puzzles using curve matching”, M.S. thesis, Brown Univ., Div. Eng., Providence, RI, May 2002.

*Corresponding author

References

[1] Ioan Glodariu, “Vase de provizii de inspiraţie elenistică”, Arheologia Moldovei, XVIII, 1995, pp. 45-50. (in Romanian) [2] E. Iaroslavschi, “Vase tradiţionale de provizii în epoca Latène tardive”, Arheologia Moldovei, XVIII, 1995, pp. 51-58. (in Romanian) [3] G. Papaioannou, E.-A. Karabassi, T. Theoharis, “Reconstruction of three dimensional objects through matching of their parts”, IEEE Trans. Pattern Anal. Mach. Intell., vol. 24, no. 1, 2002, pp. 114-121. [4] Q.-X. Huang, S. Flöy, N. Gelfand, M. Hofer, and H. Pottmann, “Reassembling fractured objects by geometric matching,” ACM Trans. Graph., vol. 25, no. 3, 2006, pp. 569-578. [5] Andrew R. Willis, David B. Cooper, “Computational Reconstruction of Ancient Artifacts”, Signal Processing Magazine, IEE, ISSN: 1053-5888, 2008, pp. 65-83. Articles