Prediction of load–displacement curves of flow drill screw and RIVTAC joints between dissimilar materials using artificial neural networks

Abstract

Abstract The load–displacement curves of flow drill screw (FDS) and high-speed bolt joining process (hereafter referred to as RIVTAC) joints between dissimilar materials are predicted via development of an artificial neural network (ANN) model. The predicted load–displacement curves accurately describe the joint strength and failure mode of joints. From a lap shear test with 14 material combinations of aluminum alloys and steels for FDS joints, it was found that the load–displacement curves of FDS joints could be classified as a pull-out of fastener, plate failure, and fastener failure. From a lap shear test with 10 material combinations of aluminum alloys and steels for RIVTAC joints, it was found that the failure modes of RIVTAC can be classified as a plate failure and fastener failure. With the obtained experimental results, the ANNs were trained to predict the load–displacement curves that include the failure modes and lap shear strengths of FDS and RIVTAC joints with the material properties and plate thicknesses. The coefficients of determination between the measured and predicted loads were 0.84 and 0.96 for the FDS and RIVTAC joints, respectively. This indicates that the trained ANNs exhibit a strong correlation between the measured and predicted loads. The errors of the predicted lap shear strength were within 15.2 % and 11.1 % for the FDS and RIVTAC joints, respectively. This study provides a systematic analysis of the characteristics of FDS and RIVTAC joints between dissimilar materials and an efficient and accurate tool for predicting the load–displacement curves of FDS and RIVTAC joints between dissimilar materials.