Optimisation of convolutional neural network architecture using genetic algorithm for the prediction of adhesively bonded joint strength

Abstract The classical method of optimising structures for strength is computationally expensive due to the requirement performing complex non-linear finite element analysis (FEA). This study aims optimise an artificial neural network (ANN) architecture perform task predicting adhesively bonded joints in place FEA. A manual multi-objective optimisation was performed find a suitable ANN design space. Then genetic algorithm reduced space conducted optimum architecture. generated predicts efficiently high degree accuracy comparison with legacy using FEA 93% savings computational cost.