Introduction to Architectural Design Optimization

This chapter presents black-box (or derivative-free) optimization from the perspective of architectural design optimization. We introduce and compare single- and multi-objective optimization, discuss applications from architectural design and related fields, and survey the three main classes of black-box optimization algorithms: metaheuristics, direct search, and model-based methods. We also give an overview over optimization tools available to architectural designers and discuss criteria for choosing between different optimization algorithms. Finally, we survey recent benchmark results from both mathematical test problems and simulation-based problems from structural, building energy, and daylighting design. Based on these empirical results, we recommend the use of global direct search and model-based methods over metaheuristics such as genetic algorithms, especially when the budget of function evaluations is limited, for example, in the case of time-intensive simulations. When it is more important to understand the trade-off between performance criteria than to find good solutions and the budget of function evaluations is sufficient to approximate the Pareto front accurately, we recommend multi-objective, Pareto-based optimization algorithms.