Fundamental study for Multi-Purpose Optimal Design of wooden buildings using interpretable AI

Abstract

In recent years, the construction field faces the challenge of balancing sustainability and disaster resilience. Because wood is a renewable resource that contributes to reducing environmental impact, designs that simultaneously meet seismic performance, environmental performance, and economic viability are essential. Although multi-objective optimization using simplified analysis models has been attempted, few examples target detailed analysis models. This study introduces explainable AI for a multi-objective optimization method with seismic response, carbon dioxide emissions, and volume for timber buildings as objective variables. This approach enables the explicit identification of design variable contributions while quantitatively assessing trade-off relationships. We reported the results of evaluating its effectiveness through a case study targeting a two-story wooden building.