Corestudio Two Documentation
Core studio 2, Term 2 , Emtech 2011 , AA Mohammad Nabil Suleiman,Jens Pedersen,Shanyun Huang, Norman Hack
Introduction Site The analysis of the site has shown that the weekly saturday marked (see fig.01) had an big influence on the development of the site. Many new shops and cafes have opened their doors on broadway marked since its revitalization in the about 10 years ago. The area east to the marked though is predominantly marked by large scale industrial use , and the influence of the canal (see. fig 02) on the site is relatively week. Our approach to the site is built upon the idea of having a positive micro climatic effect on the site and programmatically activate the waterfront . It looks at strategies that enhance the micro climate and organize the site by applying geometric rules.
market activating the area (once a week)
canal unused but with potential (polluted)
fig. 01: Site with market in red fig. 02: Site with market in blue
Core studio 2, Term 2 , Emtech 2011 , AA Mohammad Nabil Suleiman,Jens Pedersen,Shanyun Huang, Norman Hack
Introduction Micro/macro climate Based on our observation of the site, wesurface picked a lotas of climate moderator Water 53000 square meters as our experiment area(see fig xxx). which has influence in city scale The low-quality office buildings and under-conditioned bus garages will be demolished for new urban design.
Maximise the water surface London
Water surfaces can reduce ambient air temperatures and increase humidity levels in the surrounding areas, whereas the artificial, man-made surfaces greatly raise them. Evaporation results in a cooling effect, which positively affected local microclimate and reduce the urban heat island effect in macro scale as well. However, the canal lying through the site is seperated from the local residents and highly underused. Thus, our initial idea emerged as to introduce water into the site and maximize the water surfaces. And by doing this, a variety of both public and private programs could be implanted into the area.
Increase Humidity Moderate Temperature
Heat/Moisture Exchange fig. 04 fig. 03: Experiment area fig. 04: Microclimate effect
Core studio 2, Term 2 , Emtech 2011 , AA Mohammad Nabil Suleiman,Jens Pedersen,Shanyun Huang, Norman Hack
Fractal geometry 2d on site: In correlation to the initial concept of maximizing the surface area of water in the site, it was chosen to work with fractals due to its ability to maximize the length of a path and create patterns that then later on could be abstracted into different functions.
fig. 06 fig. 05: Parametric definition in grasshopper fig. 06: 2d Parametric fractal applied to site
Fractal geometry Program related to fractal depth In order to explain the implementation of programmatic sensitivity, the term “fractal depth” would have to be defined: “Fractal depth” was defined as; “the number of runs that the system performed, in order to reach a certain size for the fractal geometry”. Based on this simple understanding of the system, different program was assigned based on the size of the fractal geometry i.e. the “fractal depth”. (See fig xxx.) The orientation sensitivity is being applied by controlling the offset distance of the generated fractal pattern, so south facing parts get a certain offset, northern another etc. This adaptation strategy was combined with the program sensitivity to size, so a feedback loop was created, where both systems would re-evaluate each other.
fig. 07: 3-D fractal of depth 3 fig. 08: Matrix showing relationships of program, orientation, height and fractal depth.
Fractal geometry Conclusion fractal geometry: Different ways of viewing the output have been explored, especially in regards to the water land relationship. (see fig. to the right) . In order to get an idea of scale, the most plausible layout has been applied to the site.
Land - water
The results mark the end of the one week workshop with Cristina Diaz Moreno and Efran Garcia Grinda, which has primarily served to generate Ideas of how to organize the site. The approach of using fractal geometry has shown some useful results in terms of borderline and water surface maximization, nevertheless the rules are rather rigid and self referential. It seemed to be hard to incorporate external factors like site, climate or program. This inflexibility led us to undertake further research into other systems as described as follows.
fig. 09: Different water land relationships fig. 10: Diagrammatic sketch on site
Water - land
Organizational patterns Gray - Scott model of reaction diffusion We explored two general algorithms that exhibit more complexity than a fractal system and control cell interactions through various pattern forming techniques: cellular automata (CA) which is similar to the game of life, and reaction diffusion processes (RD). In a CA system, cells could adhere to one of two states: on or off, the choice of which depends on the combination of the states of surrounding cells. An RD system is based on mathematical equations whereupon there is a continuous gradient of states for the cell. However computationally the “method for controlling local interactions” (Tsamis, 2010) within an RD system is similar to that of a CA system where a single cell is dependent on the combination of the states of surrounding cells. Recognizing the potentials of RD systems with the embedded logic of CA’s within it we decided to further explore the properties and possibilities of reaction-diffusion algorithms.
fig. 11: Reaction diffusion fig. 12: Cellular automata, Game of Life
Organizational patterns Gray - Scott model of reaction diffusion There are many interesting phenomena in nature varying in scale, which can be described and understood by reactiondiffusion systems: Seashell patterns, zebra stripes, cave stalactites, limb development, fingerprints etc. In general, reaction-diffusion systems are “mathematical models that describe the spatial and temporal variations of concentrations of chemical substances involved in a given process” (Molnar, 2010). First proposed by Alan Turing in his seminal paper “The Chemical Basis of Morphogenesis”, the evolution of RD systems is governed by two essential components, namely the reaction component, and the diffusion component. The combinational logic and competition between reaction and diffusion often leads to the emergence of intricate spatial and/or temporal patterns. Reaction: The reaction component describes the local production or consumption of chemical elements, where the rate of reaction is proportional to concentration of reacting substances. There are many possible reaction equations, which people are still inventing and modify. Some are based on real or hypothetical chemical reactions. Any system of chemical reaction can be turned into a differential equation. The most popular and thoroughly researched is the Gray-Scott model of reaction, which we adopt within our algorithm for its variety of control and pattern forming techniques. Diffusion: Diffusion is a passive spreading or averaging process. The diffusion component within an RD system describes the diffusive transport of these elements due to concentration gradients; it follows the ordinary laws of diffusion, a chemical moves from a region of greater to regions of less concentration at a rate proportional to the gradient of concentration. Mathematically, the model uses Partial Differential Equations (PDEs) to describe the dynamics of continuous quantities through space. “This gives us a relation between how quantities change relative to their neighborhood…Over time in a closed system it averages out the concentration.
fig. 13: Reaction diffusion pattern in nature
Turing patterns Gray - Scott model of reaction diffusion This section outlines the equations and parameters that define a reaction-diffusion system based on a Gray-Scott model. Although an RD system could accommodate for more than two elements, we will look at the system involving only two generic chemical species U and V, whose concentration at a given point in space is referred to by variables u and v. They react with each other, and they diffuse through the medium. Therefore the concentration of U and V at any given location changes with time and can differ from that at other locations. There are two reactions which occur at different rates throughout the space according to the relative concentrations at each point: U + 2V â†’ 3V Vâ†’P P is an inert product and since V appears on both sides of the first reaction it acts as a catalyst for its own production. The overall behavior of the system is described by the following formula comprised of two equations which describe three sources of increase and decrease for each of the two chemicals:
fig. 14: Allen Turing and a Set of Turing patterns
Reaction diffusion Gray - Scott model of reaction diffusion Each of these equations has two chemicals “u” and “v”. K represents the rate of decay of a chemical (the rate at which the reaction V→P takes place) and F the rate that a chemical is constantly being produced (feed/flow rate of U). One of the key elements to utilize RD systems is to understand how changing various parameters within the system change the resulting pattern and overall behaviour of the system. There is a narrow range in which RD does anything at all. “Changing parameters moves between different types of pattern. There are three basic types of patterns that most RD equations produce: dots, lines and an inverted dots pattern or polygons.”(Nervous website) There are also three states of an RD system: stable, oscillatory and degenerate.
The Gray-Scott model is considered to be “activator-substrate models or depletion models.” (Sanderson, 2008) As chemical U activates, chemical V is depleted and vice versa. This leads to patters that are inverses of one another as shown in the graph.
fig. 15: Parameter space of Gray - Scott Reaction diffusion
Reaction diffusion 2d reaction diffusion In this section we describe and illustrate some of the patterns that wereproduced with numerical experiments of F and k on a simple Reaction-Diffusion model set up on a square flat plane, in two dimensions. One of the difficulties of working with RD systems is finding the right parameters for a set of reaction equations, wherein diffusion rates and reaction parameters become essential. Changes to these parameters lead to certain behaviours and a range of patterns being formed. It is difficult to control the process and the exact output is often unpredictable. A small change in parameters could lead to the formation of an entirely different pattern, or it can prohibit a pattern from forming at all. As a result we conducted a series of experiments to test out various values of F and k which resulted in the illustrated patterns. These simulations of a simple reaction-diffusion model revealed a surprising variety of irregular and regular spatiotemporal patterns. For example some patterns were spots that grew until they reached a critical size, at which time they divide in two. Some other values showedthat when spots overcrowd a region, all of the spots in that region decay into the background. These experiments would later inform our control of the RD algorithm in self-organizing pattern formation.
fig.16: 2d experiments with differing F and K values
3d Reaction diffusion Point cloud representation In order to explore the spatial qualities of the Reaction Diffusion system the next experiments were taken into the 3rd dimension. Instead of merely checking the horizontal neighbors, also the neighbors above and below are being considered. With regard to a better understanding of the system a threshold dependent display mode is introduced.
fig. 17: display of different thresholds fig. 18: Stills from animation, showing the dynamic behaviour of Reaction Diffusion
Site specific Reaction Diffusion Boundary sensitive RD As a first step of applying the reaction-diffusion model on site we modified the 3D voxel set up to accommodatefor an irregular boundary condition to the system. As a result the new set up could take various border forms and it would run accordingly within the new set perimeter. Moreover the trigger points and thus the initial state of the system are informed by the various programmatic functions/buildings that currently exist on site. As a result these add a layer of relevance and association between the model and current urban conditions. The screen captures illustrate the patterns forming during time within a certain boundary, existing buildings were set as trigger points for the initial state, and the whole site was given an F value = 0.05 and k value = 0.058.
fig. 19: Stills form animations
Site specific Reaction Diffusion Taxonomy To get a sense of what parameters produce which patterns in 3D on site and the resulting behaviour of the system, we carried out a similar experiment to the one set up earlier in 2D. Due to the sensitive nature of an RD system, the scaling up of the model and the extra dimension added resulted in a slight shift in the previous parameters. A taxonomy of the generated patterns (by varying F & k) is illustrated and was evaluated. This would later be used as a basis for further controlling the system and producing specific local pattern behaviour more accustomed to the current urban condition, rendering further flexibility within the system.
Semi - public
fig. 21 fig. 20: Diagram showing the privacy gradient fig. 21:. Experiments on site with variing F and K values
Site specific Reaction Diffusion Taxonomy The evaluation was carried out based on a programmatic gradient that was cross-referenced to the colour gradient output from the system; where green denotes public spaces (open urban spaces), and blue denotes private spaces (built-up). Ratios of the colours and patterns formed were evaluated, where some F/k values would yield more built up space, other values would give more public open spaces and some values produce mixed and balanced ratios of public to private spaces. Ratios and periodic patterns were evaluated to give us better understanding of input parameters and their effect on the system in general and site in particular. This would later help us better regulate urban program and pattern formations.
Semi - public
fig. 23 fig. 22.: Diagram showing the privacy gradient fig. 23: Experiments on site with varying F and K values
Site specific Reaction Diffusion Boundary sensitive RD The Reaction Diffusion system is refined with four operators, respectively responding to four architectural intentions. First of all, Proximity Operator takes existing site condition into account. We give different K and F values to different regions based on the proximity to the Broadway Market, canal and railway, which allow the differentiation of the pattern. Secondly, the solar environment is considered with the Solar Exposure Operator. When the system detects a south-facing cube, it increases its concentration in order to make the south-facing parts grow higher than the north-facing ones. Whatâ€™s more, Open Space Orientation Operator creates uneven offset of the borders according to the orientation. The offsets in the south are larger than the east and west, while the ones in the north are the smallest. This rule set generates larger south-facing open spaces. Lastly, we introduced Transitional Space Operator as to manipulate the threshold of each type of space.
- Take site condition into account in the proximity aspect
ne ig hb o
n th ow gr
- Increase the concentration on south facing cubes, therefore let them grow higher than e.g. north facing facades
Solar Exposure Operator
- Uneven offset of the borderline e.g. decreasing offset distance
- Increase south facing public spaces
Open Space Orientation Operator
- Alternate display threshold
o ts es
- Increase solar exposure
ti o n
- Apply different k and f values according to the proximity to certain objects on the site
- Semi-public spaces e.g. water as buffer zones in-between public and private, preferably south
Transitional Space Operator
Reaction Diffusion Internal parameters F and K
fig. 24: Diagram of External and Internal Parameters fig. 25: Diagram of operators
Site specific Reaction Diffusion Locally differentiated concentration Besides the boundary and the initial condition of the model, the current RD algorithm exhibits a closed system which does not respond to the differentiated conditions on site. Trying to employ greater control and greater differentiation within the system we gave each cell of the grid different F & k parameters depending on their proximity to certain identifiable elements on site, thus changing the pattern through space. By cross referencing the taxonomy created of patterns and their corresponding parameters with the new proximity model set up, we were able to control which pattern or gradient of spaces are formed close to the canal, market, or train tracks for example. Moreover by manipulating the different curves we were also able to control the degree of influence each element could impose. This resulted ina more interactive RD model that produces locally controlled and differentiated patterns and spatial organization across the urban tissue which is responsive to diverse and dynamic inputs from the site.
KcFc KbFb 0.085 0.03
fig. 26: F and K values and associated patterns fig. 27: Proximity related concentration change
Architectural Interpretation Metaball Interpretation Besides the pattern forming qualaties of reaction diffusion and its capacity to distribute Volumes and Voids rather evently within the assigned borders, other parameters for evaluation were needed. The computational logic of the algorithm restricted us to the use of a grid with well defined set of neightboring conditions. In the first experiments this grid was represented by dots , later experiments used a voxel representation, but for further evaluations of the system a â€œmore architecturalâ€? representation or interpretation was needed. If e.g. light blue represented a buffer zone inbtween the private and the public , its architectural expression was of great importens .In order to make a meaningful evaluation, by e.g. sections or solar exposur , it has to be defined, if a light blue box will be interpreted as a balcony, an elevated street or a water surface. A whole set of experiments have been conducted and discussed. The most promising interpretation which offered the possibility for an architectural evaluation was a 2d metaball representation in horizontal planes. The legend below shows its architectural meaning .
fig. 28: Legend of architectural interpretation fig. 29: Metaball representation
Architectural Interpretation Internal and external parameters on site Stills from the animation show the development on the site over a course of 1400 iterations. In this experiments the external parameters were taken into account. The proximity operator and the solar exposure operator ( see matrix p. 16) overwrite the F and K values returned by the â€œpureâ€? reaction diffusion model.
fig. 30 fig. 30: Stills from the animation showing Reaction Diffusion applied on the site.
Methods of evaluation Solar exposure
Insolation Analysis Total Sunlight Hours
Value Range: 0.00 - 5.00 Hrs ÂŠ ECOTECT v5
fig. 31: Daylight exposure analysis
Total Sunlight Hours
Value Range: 0.00 - 5.00 Hrs
ÂŠ ECOTECT v5
The daylight exposure analysis could be a way to evaluate the form generated by the system. By calculating the daylight hours on 21th December, the shortest day of the year in London, we could evaluate the solar exposure level of the buildings.
Methods of evaluation Dimensions and distances The light blue represents the water surfaces, which are the buffer zone between the buildings (dark blue zone) and the open spaces (green zone). The three types of spaces are distributed and mixed in a dynamic manner, providing various opportunities for different programs. The water system moderates the temperature by taking the heat away and increasing the humidity, thus optimizes the micro-climite in this district. An amount of water fronts allow a wide range of activities. The evaluation of the dimensions and distances of the possible outcome could feedback as an architectural view to the system. Distribution of densities could also be an effective evaluator.
E fig. 32
fig. XXXX. Diagrammatic section through Site and Topview of Site
Methods of evaluation Conclusion The reaction diffusion model has given us an interesting insight into complex dynamic systems and their potential application to architecture on an urban scale. We have tried to develop methods of manipulating the system by introducing architecturally meaningful external parameters, which would overwrite and broaden the internal logic of the Reaction diffusion model. However the experiments remained in a rather diagrammatic state, which only allows us to speculate on its architectural potential. Deeper research has to be undertaken in order to overcome the rigidity of the 3 dimensional dominant grid. The metaball strategy has served as a valuable diagram of the gradient space generated by the system, nevertheless it is not yet to be seen as architecture. the spatial qualifies seem promising but a taxonomy has to be extended and evaluated entirely, preferably in an automated manner.
Appendix Jury Comments The 31st of March was the date of the final review where the current state of the project was presented, and the following text will be a brief summary of the critique given: - The height of the buildings was undifferentiated and that different scaleâ€™s was not explored. The height of the proposed architecture at the site was defined by the system, which means that there were no control of where and how high the buildings would be. After the system had run for a given time period all the heights would seem the same. This specific issue will be further developed later on in this paper. - Why didnâ€™t we explore the system further? By this critique point there was referred to why it wasnâ€™t tried to let the system take over every aspect of organization. This implies that the system should posses a recursive- logic, which means that the systems starting condition would be redefined over time, so what is defined as building space would then be the new starting condition for another run of the system. This would be a very interesting investigation, but unfortunately a very time consuming one, because the computational model would have to completely reconfigured, but it will be a strategy that will be kept in mind for future use. Beside the deserved critique their where were some positive comments which focused on the research and exploration of the reaction diffusion system and the way in which it was tried to control the system. Of the two summarized critique points it was chosen to incorporate the one that focused on the height differentiation. Alongside this experiment it will tried to refine the affect of the local sun exposure to the pieces of architecture. These two explorations will be further explained on the following pages. Core studio 2, Term 2 , Emtech 2011 , AA Mohammad Nabil Suleiman,Jens Pedersen,Shanyun Huang, Norman Hack
Appendix Annotations and Adjustments The incorporation of height into the computational model was inspired by the way in which the variation of the “F” & “k” values were applied, so the height of the proposed building envelopes is based on their proximity to the canal, the rail tracks or the Broadway market. The logic of this implementation is that if the point value is within a given proximity to one of the triggers and above a certain height range, the concentration of that point will be lowered.(H1) H2 is a description of the certain height values for varying proximity.
Solar: The previous logic(fig xxx) was to find a cluster of blue boxes, find the most south facing box’s and raise the concentration above these boxes, this would then make the proposed architecture taller. The problem with simply making the structures taller is; the contradiction that the added evaluation of the height based on its proximity poses. The two parameters simply cancel each other out. Because of this conflict of parameters it was chosen to re-adjust the solar evaluation, so that it would affect the curves that were generated as boundary conditions for the build envelope. In affect this means that the reaction diffusion system is not being directly affected by this approach, this is viewed as a post evaluation of the output of the reaction diffusion system. The way in which the approach is applied is by moving the control points of the curves upwards in the z direction. The distance that they are moved is based on their orientation (north, south, east or west).
market prox up to 2 stories
water prox up to 3 stories
rail prox up to 8 stories
new solar S