Machine learning-based two-stage damage prediction method for RC slabs under blast loads

Abstract

Reinforced concrete (RC) slabs are extremely vulnerable to damage in explosions and terrorist attacks as the force members of building structures. It is necessary to evaluate and predict the damage of the RC slabs to improve the explosion protection of building structures. In this study, a two-stage damage prediction method for RC slabs under blast loads is developed using machine learning method. In the first stage, the parameters related to the RC slab and the explosion are used as input feature variables, and a machine learning algorithm is adopted to establish a displacement prediction model for the RC slab under explosion loading. In the second stage, the prediction of the maximum displacement of the RC slab under blast loads is carried out using the proposed model, and the damage of the RC slab is evaluated following the damage assessment criteria. Finally, the accuracy and reliability of the two-stage prediction method is validated by the present empirical methods. The results show that the two-stage prediction method under the damage assessment criterion of the support rotation has the best damage identification results with an accuracy of 93.1 %. Furthermore, the two-stage prediction method has better generalization performance with an accuracy of 90 % compared with the present empirical prediction methods.