An ANN-GA optimization method for building clusters based on wind-induced vulnerability

Abstract

Optimizing the building parameters of low/high-rise building clusters using wind-induced vulnerability (WIV) is valuable for minimizing building envelope damage losses due to strong winds and windborne debris. In general, the building parameters with the minimum WIV from typical scenarios are selected as the optimal ones. However, due to the multitude of parameter variables influencing the WIV of building clusters, such as roof slope, building spacing, building orientation, aerodynamic mitigations, etc., the vast number of parameter scenarios across various wind directions required makes it difficult to determine the optimal building parameters (OBPs) by enumerating the WIV for each scenario. To address this issue, this study proposes a novel artificial neural network (ANN)-genetic algorithm (GA) optimization method based on WIV to efficiently obtain the OBPs of building clusters. This ANN-GA method involves three steps: first, establish an ANN surrogate model to predict extreme wind loads for building clusters with different building parameters; second, perform WIV analysis after incorporating the models for wind-induced roof panel damage and windborne debris damage to windows and considering the uncertainties in wind speed and direction; and third, optimize the building parameters of the building clusters using the GA. To validate the effectiveness of the proposed method, an optimization case study is conducted for the orientation and spacing of a typical row of three gable-roof buildings. The findings demonstrate that the proposed ANN-GA optimization method can efficiently solve complex multivariable optimization problems of building clusters to find the OBPs and achieve a lower wind-induced economic loss for building clusters. For scenarios with fixed building orientations, the wind-induced loss with the optimal spacing determined using the ANN-GA method is reduced by 59.95 %–88.87 % compared to that with the most unfavorable spacing. Compared to selecting the OBPs from typical scenarios, the ANN-GA optimization method can achieve a further reduction in wind-induced loss by 2.21 %–3.39 %.