A Neural Networks Approach to Measure Residual Stresses Using Spherical Indentation

In the present study an Artificial Neural Network (ANN) approach is proposed for residual stresses estimation in engineering components using indentation technique. First of all, load-penetration curves of indentation tests for tensile and compressive residual stresses are studied using Finite Element Method (FEM) for materials with different yield stresses and work-hardening exponents. Then, experimental tests are carried out on samples made of 316L steel without residual stresses. In the next step, multi-layer feed forward ANNs are created and trained based on 80% of obtained numerical data using Back-Error Propagation (BEP) algorithm. Then the trained ANNs are tested against the remaining data. The obtained results show that the predicted residual stresses are in good agreement with the actual data.