심층신경망을 사용하여 모드 I 균열의 응집 법칙을 구하는 방법의 개발

In this study, we propose an inverse analysis technique to obtain crack tip cohesive laws using deep neural networks (DNN) based on displacement and reaction force data from crack propagation simulations of a double cantilever beam using cohesive elements. A novel method is developed to effectively perform DNN learning for obtaining bilinear cohesive laws by organizing the learning data into polynomials. In addition, we present an efficient method to reduce DNN training time using orthogonal array tables. The performance of the proposed method is verified through numerical examples.