• No results found

2.2 Decision Making in Computational Design

2.2.2 Performance Simulation

During early design stages, designers and architects contribute in significant ways to the determination of building performance, while at later stages those activities are frequently carried out by engineers. Therefore, the integration of simulation and analysis of building performance during the early design stages is crucial for the design of high-performance buildings. Nowadays, digital models of building designs and components are frequently used to evaluate performance parameters and predict the efficiency of proposed designs in building performance simulations [Blocken et al. 2011], [Hensen and Lamberts 2011]. Even if the literature presents the progress made in performance simulation as a success story, the reality in architectural practice looks somewhat different 24. Even if we consider the technology used in performance simulation as advanced enough to deal with the complexity of architectural design, the use of optimisation methods in architectural design is still not very popular today.

22

An adaptive optimisation setup, which adjusts the criteria and weighting during the optimisation process, leads to solutions that are hard to compare numerically as a direct derivation of the fitness value. Therefore, measuring the convergence of the optimisation process is complex.

23

Constraints are usually defined by parameter domains or functions that evaluate the validity of solutions in respect to the relationship between different variables.

24

George Stiny points out that current optimisation tools lack mechanisms for choosing desired designs from the vast number of generated design solutions [Stiny and Gun 2012].

In general, only a small number of developed design solutions in architectural practice are based on performance analysis. As performance analyses are not implemented in most CAD software packages, building simulation is frequently introduced into the planning process after finishing the initial design. The consequences of a lack of integration of simulation processes include expensive design changes undertaken to optimise building designs later on in the design process [Schlueter and Thesseling 2009]. As a result, building designs are adjusted in an inefficient manner simply to meet certain performance standards [Azhar et al. 2011]. Parametric design and associated simulation capacities 25 represent an enormous step forwards in performance-based design26.

The application of architectural optimisation in real-life architectural projects re- vealed that the computation time for simulation of building performance takes up the majority of computational resources, and suggests a parallel implementation to evaluate larger numbers of solutions, as discussed by Robert Vierlinger [Vierlinger 2015]27.

With regard to optimisation criteria in architecture, Jane Burry [Burry 2007] stated that a wide range of research in architectural design has been undertaken to explore optimisation potential in relation to directional sunlight. The present literature review showed that a breadth of research has recently been conducted in the field of structural design in architecture (e.g. [Leary et al. 2014], [Marler and Arora 2004], [Seifi et al. 2016], [Crolla et al. 2017]). As designers are interested in developing continuous computational processes to constantly adapt their own thinking in parallel to the evolution of the model, real-time evaluation is critical for the engagement of designers in the optimisation process. There are a variety of computational methods for reducing computation time for building performance simulation. These approaches reduce the amount of simulation needed by reducing the number of simulation cycles initiated by approximating the fitness of similar designs. In optimisation, these methods include surrogate models [Wortmann et al. 2015], [Jin 2011], multi-variate analysis [Sileryte et al. 2016], pareto front clustering [Veerappa and Letier 2011] and artificial neural networks [Wilkinson and Hanna 2014] to name just a few. The success of those methods is proportional to the complexity of the simulation needed to describe a certain process. One particularly heavy computation that seems to be promising in architectural design - especially in the context of tall building design - is Computer Fluid Dynamics (CFD).

25

The major plugins for performance simulation in Grasshopper visual programming editor are Ladybug for the evaluation of environmental parameters [Roudsari et al. 2013] and Karamba for evaluation of building structures [Preisinger and Heimrath 2014].

26

George Stiny remarks the loss of visual interactivity in parametric design as dominant mode of computational design practice [Stiny and Gun 2012].

27

First attempts to generate an infrastructure for cloud access the Grasshopper visual programming editor to outsource these processes to high performance computing were proposed at the ACADIA 2015 Hackathon with new web-based tools. An implementation in a more generic environment (like Python) will allow the whole optimisation process to run in the cloud. Data-exchange and programming interfaces for the interaction of various software packages are the key challenges in programming design tools using cloud computing.

Based on the demand for improvement of building performance and reduction of environmental impact caused by the built environment, the understanding of air flow inside and outside buildings is growing in significance in the context of architectural dis- courses [Kaijima et al. 2013]. While CFD is widely used in industrial design as an efficient evaluation tool, it is just starting to enter the domain of architectural design discourse [Kaijima et al. 2013]. Besides the great potential of CFD simulation in architectural de- sign applications, the limitations around the description of turbulent flows will prevent the development of a computational wind tunnel in the close future [Stathopoulos and Baniotopoulos 2007]. A hybrid approach using physical simulation in a mini wind tunnel (conceptually similar to [Moya Castro 2015]) and CFD simulation as tested by myself in collaboration with a strong team of SIAL experts [Muehlbauer et al. 2016] for the design of origami-folded facades is highly promising, and deserving of further research, though it is beyond the scope of the present research and discussion.