An improved artificial neural network using gradient-based optimizer for precise ultimate tensile strength prediction in GTAW of Inconel 825

Abstract This study presents an innovative approach to improve prediction accuracy for gas tungsten arc welding (GTAW) of Inconel 825 alloy. A comprehensive comparison of seven advanced metaheuristic algorithms, spider wasp optimizer (SWO), weighted mean of vectors (INFO), gradient-based optimizer (GBO), artificial rabbits optimization (ARO), blood-sucking leech optimizer (BSLO), RUN beyond the metaphor (RUN), and successive history adaptive differential evolution (SHADE) was conducted for training artificial neural networks (ANNs) to predict ultimate tensile strength. The ANN-GBO model demonstrated superior performance with excellent generalization capability, achieving significant improvements in prediction accuracy with coefficient of determination (R2) increasing from 0.6844 to 0.8669 (26.7% improvement) and root mean square error (RMSE) decreasing from 51.89 MPa to 33.71 MPa (35.0% reduction) compared to conventional ANN models in previous study. The proposed methodology effectively addresses limitations of traditional ANN training methods while providing practical insights for industrial welding applications of high performance nickel-based alloys.