Bead geometry plays a significant role in the fabrication of artefacts via the wire arc additive manufacturing (WAAM) route especially with respect to the dimensional characteristics, surface properties, and part density. It also substantially affects the material penetration, part accuracy, and thickness of subsequent layers. The present study focuses on the effect of processing parameters on single-bead geometry for fabricating 316LSi stainless steel in pulsed-mode metal transfer WAAM. Models were developed by utilizing multiple regression modelling, machine learning (ML) based on back propagation neural network (BPNN), and finite element modelling (FEM) with four process parameters, wire feed rate (WFR), torch travel speed (TTS), voltage (V), and contact tip-to-substrate distance (CTSD), each considered at three levels. Response surface methodology was utilized for design of experiments to obtain three-bead geometry responses, namely width of weld bead (WWB), height of weld bead (HWB), and depth of penetration (DOP). The results confirm that all three proposed models predict bead geometry reasonably. However, accuracy of ML-based BPNN method is better as compared to the regression and FEM-based approaches. For multi-objective optimization, a desirability function approach is used, which predicts the optimal conditions at WFR of 4.0 m/min, TTS of 284.545 mm/min, voltage of 18.0 V, and the CTSD of 12.0 mm for minimum WWB, maximum HWB, and minimum DOP. A multi-layered wall of SS316LSi is fabricated with optimized process parameters, and it is found that the ultimate tensile strength, yield strength, and percentage elongation are 546.3 MPa, 440.1 MPa, and 43.7%, respectively. WAAM-deposited structure reveals stable deposition, no debonding/delamination, and complete fusion between layers. The results of the present article will go a long way in aiding the researchers/industrial personnel in arriving on the optimal modelling approach and deposition conditions for WAAM-fabricated SS316LSi parts with enhanced mechanical properties and productivity.
Read full abstract