Currently architects use parametric modelling and they could generate thousands of variations of the same design concept – filling “the pool of alternatives”. They are not able to assess all of them manually, especially if the assessment criteria involve the daylight or energetic efficiency analysis. Even engaging a computer to calculate all alternatives and indicate the best performing ones, the process most of the time is infinite. Even a small optimization problem, with 5 parameters and 100 possible values of each parameter, has 1005 = 10000000000 possible options. Therefore, if one calculation might take just a minute for a computer (and the simulations can take up to 30 minutes per alternative!) it gives 6944444,(4) days = 19 025 years of
Therefore we have to implement Artificial Intelligence to the parametric design – in order to have the design answers within the finite time. Parametric Support implements both mathematical methods and algorithms from the field of Artificial Intelligence to find the wisest ways of the search in the pool of design alternatives and indicate the best performing options within hours and not days.
Architectural Design Optimization
Architectural Design Optimization (ADO) is an emerging subfield of architectural engineering. It uses optimization techniques to solve architectural design problems, such as “daylight availability, circulation, optimal floor plan layouts, air circulation, adaptation of projects to law restrictions, cost-effectiveness, energy consumption […] and many others“ (Cichocka & Musikhina, 2014).
In ADO, in order to assess the quality of an alternative, an architect or a designer (or Parametric Support on their behalf) formulate a fitness function, which assess how „fit” the solution is.
Fitness function is a formula used to assess how a given design solution is close to achieving the set aims of optimization. For example, in the optimization of the earth works, we formulate the fitness function which calculates the total amount of the earth that has to be added or removed. The aim of the optimization is to minimize the landscape change, therefore the lower the fitness function value is, the “fitter” (better, more optimal) solution is. The fitness function value for the design solution 1 is 3500m³ and for design option 2 is 4200m³. Obviously the first solution is more optimal, as it require less digging and less earth to be added to build this design option. In short, fitness function tells us how good is the design solution.
Optimization goal is to minimize or maximize one of the futures. Benefits coming from the process could be defined as direct and indirect: from raw material reduction to amount of energy used by a building for lighting. For instance robust optimization processes on building massing can indirectly reduce carbon footprint by savings in the material use and in the operational energy of the building. Modern optimization techniques could aid architects in balancing environmental, structural and geometrical aspects and discovering energy-efficient, cost-effective and sensible design solutions that have the minimal environmental impact.
Accordingly to our latest research (Cichocka et al. 2017) in the robust optimization process by reducing the maximum nodal displacement we could reduce the thickness of the reinforced concrete shell from 20 to 12cm. It gives a direct benefit in the 40% reduction of the raw material used. However the profits of optimization could go far beyond that. The implementation of the optimization processes in the design workflow can reduce the time of the design and construction phase. It could help to increase the property value and maximize the Return On Investment by incrementation of the usable area, views enhancement, reduction of the landscape change or by reducing the energetic demand of the building. Buildings use 80% of the total energy during their life cycle for the operations of the building.
it gives the direct savings on the air conditioning.
By changing the global shape and optimized distribution of the windows, we could guarantee perfect daylight conditions in the office buildings. Taking into account that office buildings use 25% of the operational energy for the lightning , thanks to optimization we can not only create the occupant-friendly working conditions, but again save a lot of energy used for lightning (our studies demonstrated around 20% on savings in this aspect). These are just several examples of the optimization benefits. The optimization service is mostly project-based and we could help to achieve the personal aims of the real estate developer by parameterizing and optimizing the design task.
The design optimization problems could be classified in multiple ways. For instance we could distinguish them accordingly to:
Number of criteria:
Single-objective ( the problem that has just one criterion to be maximized or minimized; although the objective is just single, a multiple solutions could be found as the answers)
Multi-objective ( involves two or more objective functions; there is no single solution exists that simultaneously optimizes each objective to the multi-objective problem, there exists a number of Pareto optimal solutions (also called nondominated, Pareto optimal), if none of the objective functions can be improved in value without degrading some of the other objective value)
Or a number of design domains involved in the problem formulation:
Single -modal (just one design domain in the optimization problem e.g. structural optimization; a single modal problem could have both single or multi-criteria)
Multi-modal (a few design domains engaged in the optimization process e.g. at the same time the design is going to improves its structural and environmental performance)
We could distinguish more categories depending on the problem formulation itself, however specifying it in the architectural design might be a tricky task and Architectural Design problems are usually considered as the black box problems.
Methods: Dedicated vs Generic
Optimization techniques can be divided into two major groups: dedicated and generic methods.
Generic Methods – approaches that can be applied to any design problem which is formulated in the form of a function. With generic methods we can search for the input values that minimize or maximize this function, we can also search for such input values that the function attains some wanted value.
Dedicated Methods – these techniques are designed to solve some specific problems. For instance, the most of Discrete Surface Optimization techniques are dedicated. Another example are form finding approaches such as Dynamic Relaxation which aims to find the geometrical configuration where all the forces are in the equilibrium.
Generic Optimization Methods have the advantage that we, as the designers and decision makers, can design our own fitness functions according to what we would like to achieve. On the other hand, there are
generic, meaning that that they are able to solve the most of problems, but we have to sacrifice performance (Rutten 2014).
Dedicated and generic methods can be compared to two sorts of knives.
like with the set of knives, obviously it is going to take longer time than with the dedicated tool.
None of these categories dominates the other one. Depending on available time and on the characteristics of a problem we use either generic or dedicated methods, or a combination of both.
What do we use?
We have developed our in-house generic solver based on Swarm Intelligence that can solve multicriteria and multi-domain problems. The developed by us tool is a generic type and could handle most of the problems that could be encountered in the practice. In order to present the results of the multi-criteria optimization process we create Digital Catalogues of design solutions with the Pareto Front solutions. The clients could assess them and compare the parameters for each solutions themselves, use the catalogue as the communication media with other parties engaged in the design process.
However at the moment we are developing a set of dedicated optimization tools that aim at the Incrementation of the Return On Investment. They are going to be soon available on our website and are going to be wrapped up as a stand-alone, web-based tool package.
Cichocka, J. M. et al. (2017) ‘SILVEREYE – the implementation of Particle Swarm Optimization algorithm in a design optimization tool .’, In: Çağdaş G., Özkar M., Gül L., Gürer E. (eds) Computer-Aided Architectural Design. Future Trajectories. CAADFutures 2017. Communications in Computer and Information Science, vol 724. Springer, Singapore
Cichocka, J. M. and Musikhina, E. A. (2014) ‘METHODS OF OPTIMIZATION IN ARCHITECTURE’, Research Journal of International Studies, 3, pp. 109–111.
Rutten, D. (2014) ‘Navigating Multi-Dimensional Landscapes in Foggy Weather as an Analogy for Generic Problem Solving’, 16th International Conference on Geometry and Graphics.