Neural Network Prediction of Stress Concentration Factors for r/d Ratios for Various Cavity Dimensions on Round Sleeve Automotive Terminals

Round sleeve automotive terminals are used in automotive connectors in safety and infotainment systems. During terminal assembly in connector, the plastic cavity is subjected to stress concentration at the sharp edges. To overcome this, edge blends were considered to be included. Stress Concentration Factor curves and associated data corresponding to double and single edge blend design of cavity at the terminal interface were considered for the training and testing of a feed-forward Neural Network with backpropagation developed using TensorFlow open source library. Predictions for the Stress Concentration Factor (Kr) were done and the results were tabulated for values within and beyond the dataset. A study of variation of Mean Absolute Error against number Epochs demonstrated a decreasing correlation and the percentage error showed an overall decreasing trend. The maximum stress values from the Curve generated, finite element analysis and neural network predictions were found to be in close agreement with each other, thereby validating the process. The neural network can thus be used to predict the stress concentration factor for any size and edge blend to cavity diameter ratio with as well as beyond the dataset for the configuration under consideration with deviation of mere 0.14%, which is the novelty.