Numerical failure load analysis and prediction (ANN technique) of AA2014 adhesively bonded single lap joint: An experimental validation

The present investigation focuses on the failure load analysis of adhesively bonded single lap joints (SLJs) and the prediction of failure load using machine learning artificial neural network (ANN) tool. Initially, a numerical model was developed in the ABAQUS simulation software, and the correctness of the numerical model in computing the failure load was verified with the available data in the literature. The obtained numerical results are experimentally verified by conducting experiments on aerospace-graded aluminium alloy (AA2014) adherends joined with epoxy based adhesive bonded joints. The research also examines the effect of hybrid surface treatment (chemical and mechanical) on the failure load and failure mechanism of adhesively bonded joints. The influence of adhesive thickness and overlap length on bonded joints failure load and shear strength are studied experimentally. The verified numerical model is extended to generate the 54 datasets by varying the geometrical parameters (i.e., overlapping length, adhesive thickness, and adherend thickness) of the bonded joint. The ANN architecture has been confirmed by executing the model with the experimental data available from the literature, and its functions are verified by performing an Ablation study. The generated data set is trained to the verified ANN model and predicts the failure load for unknown geometrical parameters. Lastly, the ANN-predicted failure load is experimentally validated with the desired geometrical parameters of adhesively bonded SLJ.