Optimizing WAAM process parameters using statistical tools and prediction of input parameters using machine learning models for depositing ER70S

Wire arc additive manufacturing (WAAM) is a type of fusion manufacturing technique that uses the heat energy of an electric arc to melt the electrodes and deposit material(s) layer-by-layer to build a wall or to concurrently clad two materials to produce a composite structure. In recent past, a significant interest is shown by various industrial sectors—aerospace, automotive, nuclear, and mould and die making—for the use of WAAM, which indicates compatibility and comprehensiveness. The successful printing of WAAM depends on parameters like bead width, bead height, tensile strength, and hardness, for shape, size, and quality. Optimization of these process parameters can lead to optimal deposition and high-quality products. The current work explores analysis of variance (ANOVA) and response surface methodology (RSM) methods for arriving at optimized input parameters to deposit the copper coated AWS ER70S-6. The determined optimized input parameters—travel speed, wire feed rate, and welding current, are 30 cm/min, 5 m/min, and 90 A respectively. In addition, prediction of the input parameters via machine leaning based techniques has also been carried out. Predictive models for bead width, bead height, tensile strength, and hardness have proven effective in optimizing process parameters for specific applications. Amongst the three models used, Extra Trees and XGBoost can be considered as good candidates to solve a regression problem, whereas multilayer perception (MLP) performs poorly on new data as applicable to the current study.