Prediction of failure stages for double lap joints using finite element analysis and artificial neural networks

Abstract

Abstract Prediction of failure stages of double lap bolted joints is an important task for design. Finite element (FE) and artificial neural networks (ANNs) methods are widely used in such tasks to reduce time, money, and efforts consumed in laboratory experiments. Three dimension elastic-plastic FE analysis was adopted to simulate the failure stages of double lap joint in the present work. After completing mesh sensitivity analysis for the present FE model, only nine experiments were conducted to verify the prediction of the present FE model. Taguchi technique was applied to reduce the number of loading cases (different joint geometry and tightening torque) from one hundred twenty five cases of loadings to only twenty seven cases of loadings. The FE results of twenty seven cases of loadings were accomplished to provide the training set of ANNs. Finally, a comparison between the results obtained from FE and ANNs was made. Results indicated that, either FE or ANNs can highly predict the failure stages of double lap bolted joints for different e/D, w/D, and tightening torque values. Therefore, FE and ANNs methods may be considered good candidates to predict the mechanical behavior and failure mode of such joint.