Forming process prediction of a self-piercing riveted joint in carbon fibre reinforced composites and aluminium alloy based on deep learning

As a mechanical internal locking structure, the forming process of a self-piercing riveted (SPR) joint is currently mainly investigated through riveting interruption experiments and finite element simulation. To solve the problems of complex simulation modelling processes, high experimental costs, and low efficiency, an SPR process prediction method based on deep learning was proposed to predict the deformation process and damage of SPR joints in carbon fibre-reinforced composites and aluminium alloys. The original images of the model dataset were obtained based on the simulation results, and image segmentation technology was used to classify the cross-section and damage morphology of the joint. A deep learning model based on a convolutional neural network and conditional generation antagonism model architecture was established, and the section shape and damage morphology of the joint were predicted by inputting the percentage of punch displacement. The k-fold cross validation method was used for model training, and the untrained data were used as the test set to verify the predictive ability of the model by comparing the forming section parameters of the SPR joint. The results show that the deep learning model used can accurately predict the deformation state and damage evolution of riveted materials at different joining stages. The average prediction accuracies of the height of the riveted head, residual thickness, and rivet spread are 95.80 %, 95.68 %, and 92.40 %, respectively.