An ANN-based instrumented indentation method for evaluating biaxial residual stress in metallic materials

Residual stresses in metallic components can significantly impact their mechanical performance. Accurately evaluating biaxial residual stresses is crucial for analysing mechanical behaviour and assessing service safety. This study proposes a novel biaxial residual stress evaluation method that integrates spherical indentation with neural network tools. Dimensional analysis and correlation analysis were first employed to select the optimal analytical parameters (such as the plasticity parameters, the dimensionless loading work, and the ellipticity of the residual imprint) for evaluating biaxial residual stresses. Then, an artificial neural network (ANN) was applied to map the relationships between the analytical parameters and the biaxial residual stresses based on a database constructed by finite element simulations. With these relationships, an ANN-based instrumented indentation method to evaluate biaxial residual stresses was established. Finally, numerical and experimental validations were conducted to ensure the accuracy and reliability of this newly proposed method. The advantage of this method is to measure arbitrary biaxial residual stresses in a non-destructive way without using stress-free specimens which are difficult to obtain in engineering.