Parameter identification method of the semi-coupled fracture model for 6061 aluminium alloy sheet based on machine learning assistance

[Display omitted] The ability to describe the fracture behaviour of materials and the difficulty of parameter calibration are important factors for evaluating the ductile fracture criterion. To find a ductile fracture criterion that can balance both these factors, in this study, the fracture behaviour of 6061 aluminium alloy was characterized based on a semi-coupled damage mechanics framework, and a parameter identification method based on machine learning assistance was proposed. Using this method, the calibration method of the damage initiation model was improved to fully consider the non-linearity of the loading path. Moreover, the calibration process of the fracture and damage evolution models was replaced by the neural network model. The results indicated that the semi-coupled fracture model can accurately predict the damage initiation, accumulation, and fracture of the 6061 aluminium alloy during the forming process. The parameter identification method based on machine learning assistance can circumvent the process of extracting the stress state evolution from the numerical models of calibrated specimens; a set of reliable parameters can be extracted from the force–displacement response using only one specimen. The combination of the semi-coupled fracture model and parameter identification method provides a new approach to balance the difficulty of parameter calibration and the description ability of the ductile fracture criterion.