Prediction of bolted joint preload based on FEM-VMD-ISSA-LSTM under cyclic transverse load

Abstract

Under the action of cyclic transverse loading, the bolt preload undergoes periodic variations. Analyzing the changes in bolt preload will contribute to a better understanding of the effect of cyclic transverse loads on bolted joints. In this paper, a precise finite element model of the bolted joint is established, and the accuracy of the finite element results is verified by experiments. The bolt stress and contact conditions are analyzed in the preload local peak and non-peak stage. Based on the finite element results, a FEM-VMD-ISSA-LSTM method is proposed to predict the change of bolt preload. The preload data is divided into multiple sub-sequences by variational mode decomposition (VMD). The improved sparrow search algorithm (ISSA) is used to optimize the hyperparameters of long short-term memory (LSTM) neural networks to improve prediction accuracy. The optimized LSTM is used to predict the changes in each sub-sequence data. Finally, the predicted sub-sequences are superimposed to obtain the final loose preload prediction result. The results show that when the preload reaches the local peak, the distribution of the bolted joint stress is close to the symmetry on both sides, and the thread contact of the upper part of the bolt is partially slipped. The former 40% of the preload data is used as the training set, and the latter 60% is used as the test set. The proposed FEM-VMD-ISSA-LSTM can predict the change of bolt preload under different parameter conditions.