Data‐driven approach for blast damage prediction of precast reinforced concrete slabs

Abstract Precast building structures have been extensively used in civil engineering due to their advantages of energy conservation and environmental protection, controllable quality, and rapid construction. Precast reinforced concrete (PRC) slabs, as the main load‐bearing components in precast structures, are vulnerable to gas explosions and terrorist attacks. Consequently, it is essential to study the dynamic response of PRC under explosions with the aim of improving the blast resistance. In this study, firstly, an explosion database for PRC slabs was established. Eight parameters related to size, material, and explosion load were selected as input features to predict the maximum displacement of PRC slabs subjected to blast loads. Then, a displacement prediction model was established using two single models and two ensemble machine learning (ML) algorithms and verified by performance evaluation indexes and experimental results in literature. Finally, the causes of errors are analyzed, followed by a discussion on the improvement and application of the model. The results show that the ML model can predict the maximum displacement of PRC slabs under blast load. It is indicated that the extreme gradient boosting algorithm has higher accuracy and computational efficiency with an R 2 of 0.975 compared with existing methods. Furthermore, the ML algorithms have potential applications in predicting damage to the PRC slabs for blast engineering.