A practical decision process for building facade performance optimization by integrating machine learning and evolutionary algorithms

With the development of building design parameterization, more and more designers want to explore design options that optimize performance. However, the enormous time and costs that accompany optimization exploration are often beyond the reach of design practices. The application of machine learning in the construction field in recent years has offered potential solutions. Training predictive models through machine learning enables the rapid assessment of built environment performance and thus brings the optimization process closer to reality. In this paper, we mainly developed a process that integrates machine learning predictive model and multi-objective algorithms to achieve rapid evaluation and obtain optimal solutions. Using facade design as a case study, we demonstrate the design decision process with regard to the optimized solutions. The results showed that, through the proposed visual design decision process, designers can easily compare the performance and design appearance of different solutions and make informed decisions. In addition to saving 87% of time compared to the traditional simulation process, the integrated process also introduced the predictive model, which can achieve optimization exploration in one day. These results all demonstrate that the use of an integrated approach boasts considerable time advantages and potential feasibility in design practice.