Optimization of process parameters of SMAW process using NN-FGRA from the sustainability view point

Welding process does not possess a good environmental image due to fumes, noise and other health hazards. But, it has been widely applied and is irreplaceable as far as structural industries are concerned. In welding process, many factors are responsible for environmental burdens, of which this study investigates and optimizes process parameters. Based on the pilot study, operating levels for parameters like current, voltage and welding speed, and various responses were identified. Further, the responses considered in the study include spatter, slag, fume generation rate, particulates and power consumption. A full factorial design of three factors at five levels requires 125 experiments for a complete analysis; however the conduct of a full factorial design experiment is expensive and consumes much time. Hence, $$\hbox {L}_{27}$$L27 orthogonal design was adopted for three inputs and five levels and 27 experiments were conducted. Further, five back propagation neural network model with network structure of 3-12-1 was individually developed to predict responses. $$\hbox {R}^{2}$$R2 values obtained for back propagation neural network model pertaining to slag, spatter, power consumption, fume generation rate and particulate formation are 0.99, 0.99, 0.989, 0.965 and 0.97 respectively. The responses for full factorial interaction were simulated. Fuzzy based grey relational analysis was used to optimize process parameters. Finally, life cycle assessment methodology outlined by ISO 14040 and 14044 was adopted to quantify the improvement in environmental performance due to parameter optimization. Thus, the proposed methodology helps to improve environmental performance of the manufacturing process with the experiments conducted using partial factorial design. Apart from environmental benefits, economic benefits and man hours saving are attained without compromising accuracy.