Input–Output Modeling and Multi-objective Optimization of Weld Attributes in EBW

Abstract

SS 201 had been reported as a good substitute for SS304 without any significant compromise in performance. However, modeling EBW process using an efficient tool like adaptive neuro-fuzzy inference system (ANFIS) and use of multi-objective optimization to optimize its performance are not reported yet. Thus, the present study employed ANFIS models tuned by genetic algorithm, particle swarm optimization, gray wolf optimizer, and bonobo optimizer (BO) to predict weld attributes during EBW of SS201 as a function of input process parameters. Among the developed models, ANFIS tuned by BO was seen to yield the best prediction accuracy. In multi-objective optimization (MOO), the two conflicting goals were to minimize secondary dendritic arm spacing and maximize Vicker’s hardness number simultaneously. In MOO, some interesting facts were observed, such as the fixed input parameter of power (P) as 3200-W and squeezed experimental range for the welding speed (S).