Advantages of surrogate models for architectural design optimization

Abstract

Climate change, resource depletion, and worldwide urbanization feed the demand for more energy and resource-efficient buildings. Increasingly, architectural designers and consultants analyze building designs with easy-to-use simulation tools. To identify design alternatives with good performance, designers often turn to optimization methods. Randomized, metaheuristic methods such as genetic algorithms are popular in the architectural design field. However, are metaheuristics the best approach for architectural design problems that often are complex and ill defined? Metaheuristics may find solutions for well-defined problems, but they do not contribute to a better understanding of a complex design problem. This paper proposes surrogate-based optimization as a method that promotes understanding of the design problem. The surrogate method interpolates a mathematical model from data that relate design parameters to performance criteria. Designers can interact with this model to explore the approximate impact of changing design variables. We apply the radial basis function method, a specific type of surrogate model, to two architectural daylight optimization problems. These case studies, along with results from computational experiments, serve to discuss several advantages of surrogate models. First, surrogate models not only propose good solutions but also allow designers to address issues outside of the formulation of the optimization problem. Instead of accepting a solution presented by the optimization process, designers can improve their understanding of the design problem by interacting with the model. Second, a related advantage is that designers can quickly construct surrogate models from existing simulation results and other knowledge they might possess about the design problem. Designers can thus explore the impact of different evaluation criteria by constructing several models from the same set of data. They also can create models from approximate data and later refine them with more precise simulations. Third, surrogate-based methods typically find global optima orders of magnitude faster than genetic algorithms, especially when the evaluation of design variants requires time-intensive simulations.

Machine–Learning in Optimization of Expensive Black–Box Functions

“…Common metamodel variants include artificial neural networks, Kriging, polynomials, and radial basis functions (RBFs) (Wortmann et al, 2015;Forrester and Keane, 2008;Queipo et al, 2005;Regis, 2014;Viana et al, 2013;Tenne, 2013). Metamodel-assisted frameworks typically operate by first training a metamodel and then seeking its optimum.…”

“…Baseline metamodel-assisted optimization. sample an initial set of vectors; while stopping criterion not met do train a metamodel with the vectors evaluated so far; search for an optimum of the metamodel; evaluate the solution found with the true expensive function; possibly sample additional vectors, to update the metamodel; return the best solution found; Recent studies have explored more involved frameworks which include using multiple metamodels concurrently in an ensemble setup (see, e.g., Viana et al, 2013;Tenne, 2015;Muller and Shoemaker, 2014), selecting an optimal metamodel type dynamically during the search (Gorissen et al, 2008;Tenne, 2015), or 107 using more involved optimization procedures (Wortmann et al, 2015;Regis, 2014).…”

Enhancing Simulation-Driven Optimization by Machine-Learning

“…Simulation-driven optimization problems are increasingly common in engineering and accordingly several approaches have been studied which are tailored for such problems. One such established approach is that of using approximation models, also termed in the literature as metamodels or surrogates, which approximate the true expensive function and provide predicted objective values at a lower computational cost [1][2][3]. Metamodels variants include radial basis functions (RBF) and Kriging which originated in geostatistics, artificial neural networks from the domain of machine learning, and polynomial approximations from applied mathematics.…”

Otimização de projeto orientado ao desempenho em arquitetura

“…A "Consciência do problema" da presente pesquisa aponta que a clareza sobre o método de otimização empregado no processo de projeto em arquitetura contribui com a melhora dos resultados. A falta de clareza sobre o método de otimização aplicado pode ser problemática ao resultado do projeto, pois as especificidades de um método de otimização deveriam ser baseadas nas necessidades e na natureza de um problema específico WAGEMANS 2005;KOZIEL;YANG 2011;MACHAIRAS et al, 2013;WORTMANN et al 2015;NANCINNI 2016).…”

“…Tal condição impossibilita conclusões seguras sobre as características das otimizações produzidas, dificulta a reprodução dos métodos e desampara discussões sobre soluções alternativas. Desta maneira, torna-se reduzida a reflexão na arquitetura sobre quais métodos são mais apropriados para a otimização de determinados problemas do projeto arquitetônico (WORTMANN et al, 2015).…”