Parameter optimization of the dissimilar welding between AISI304 and S690QL steels using Box–Behnken design and artificial neural networks

Abstract

This study investigates the effects of voltage, wire feeding speed, and root gap on the tensile strength and toughness of welded AISI 304/S690QL joint by GMAW. This is a novel study that optimizes the welding parameters for these materials by models using Box–Behnken design and artificial neural networks. Welding parameters were optimized using prediction models that were developed with the Box–Behnken design and tested by ANOVA. Artificial neural network models were developed based on the same data and the findings were correlated with Box–Behnken. Two-way partial dependence plots were obtained from the ANN results and the joint effect of the welding parameters was observed. The suitability of the wire and welding parameters was confirmed by microstructure examinations. The tensile strength of the joint was highly affected by the voltage and wire feeding parameters due to their effects on heat input. While the maximum tensile strength was achieved at the lowest values of voltage and wire feeding speed, the highest toughness was obtained at the highest voltage. The parameter search enabled gains of up to 8% and 17% in ultimate tensile strength and toughness, respectively. Box–Behnken Design and ANN were useful in the modeling of welding process.