Artificial neural network and YUKI algorithm for notch depth prediction in X70 steel specimens

The presence of notches directly influences the material's degree of resistance to fracture, as each notch is defined by its size and location in the structure. In this study, we present a methodology using an optimization technique that allows us to predict the notch depths based on one of the most important mechanical parameters, namely the maximum stress for each notch depth. This study starts with the experiments of standard pipe specimens, which have notches with different depths. A numerical model, using GTN parameters, is created based on the experimental model and its results are presented in form of stress and strain curves of API X70 steel specimens under uniaxial tensile tests. The high-performance numerical models used to simulate the experimental model consider a considerable number of different cases. The study is concluded by choosing the best optimization technique among different proposed algorithms, namely NN-YUKI, NN-JAYA, and NN-EJAYA. The validation of models and their accuracy is determined by comparing the outputs of NN-YUKI, NN-JAYA, and NN-EJAYA with experimental and numerical data. The obtained results using NN-YUKI determine the best optimization parameters and provide a good prediction of notch depth. This methodology gives an insight into the notch size under different loading conditions and is applicable to large structures like pipes under various loadings.