Evaluation of plastic properties and equi-biaxial residual stress via indentation and ANN

• Published in: Materials & design 2024-03, Vol.239, p.112745, Article 112745
• Main Authors: Han, Giyeol, Lee, Bohyun, Lee, Sihyung, Jeong, Chanyoung, Lee, Hyungyil
• Format: Article
• Language: English
• Subjects: Artificial neural network (ANN); Digital image correlation (DIC); Finite element analysis (FEA); Indentation; Plastic properties; Residual stresses
• ISSN: 0264-1275; 1873-4197
• DOI: 10.1016/j.matdes.2024.112745

[Display omitted] •An indentation technique is proposed for the simultaneous evaluation of material properties and residual stresses.•The indentation horizontal displacement field, indentation depth, and imprint size are used as inverse parameters.•The range for in-plane displacement field ensuring the identifiability of inverse solution is suggested via identifiability analysis.•The ANN model evaluating the material properties is constructed using data from indentation FEA.•The evaluated material properties from ANN are verified by comparing them with those from a tensile test. This study presents a method for evaluating plastic properties and equi-biaxial residual stress (RS) simultaneously using the images of generated imprints in indentation tests and an artificial neural network (ANN). The ANN model is trained with data from finite element analysis (FEA) for establishing the relationship between material properties and the radial (ur) and vertical (uz) displacements resulting from indentation tests. To deal with the inherent noise in experimental data, an artificial random noise was added to the FEA data used for training the ANN model. By inputting parameters with artificial errors into the ANN model, the robustness of predicted values against potential testing errors was examined. Also, the accuracy of the predicted residual stresses and material properties from ANN are validated using tensile tests and indentation tests on the stress-induced specimens.