Optimization of submerged arc welding parameters to improve corrosion resistance and hardness in API 5L X70 steel joins using Support Vector Regression and Multi-Objective Genetic Algorithm

Mathematical techniques such as Support Vector Regression (SVR) and Multi-Objective Genetic Algorithms (MOGA) were used for a multi-objective optimization of corrosion rate (Rcorr) and hardness in an API 5L X70 steel welded by submerged arc welding (SAW) process with a double-V bevel shape. The inner and outer bevels (IB and OB, respectively) were joined at different conditions of voltage (V ), amperage (A) and travel speed in inches per minute (TS (ipm)), with a range in heat input (Q) of 1278–1693 J/mm. As Q is responsible for the microstructural behavior of the welds, their particular characteristics are defined by welding parameters giving as response variables the hardness and Rcorr in the fusion zone (FZ). For the experimental corrosion evaluation, the samples were tested in the FZ and base metal (BM) with potentiodynamic polarization test by three-electrode cell in an H2O + 3.5 wt.% NaCl electrolyte, and the Vickers microhardness (HV ) profiles were measured with a 500 g force. The experimental results (HV and Rcorr) were used for the corresponding prediction and optimization by SVR and MOGA. The main results show that Rcorr using optimized parameters decreases significantly from 2.356 mils per year (mpy) to 0.577 mpy in the FZ with a predominant microstructure of acicular ferrite (AF) and small regions of ferrite at the grain boundary (FGB). For the hardness, the predicted results were 217.36 HV (IB) and 225.63 HV (OB) against the 224.58 HV and 215.75 HV recorded in the validation sample revealing the great effectiveness of the applied method for prediction and optimization.