Optimization and prediction of resistance seam weld quality in secondary coated steels using machine learning

With the use of ethanol blended fuel, secondary coating on steel becomes crucial to enhance the corrosion resistance in fuel tanks. Resistance seam welding is extensively used in the manufacturing of leakproof joints. An additional layer over the hot dip galvannealed steel increases electrical resistance. This requires the selection of a suitable welding parameter to achieve the desired nugget diameter and tensile strength. However, assessing the quality of resistance seam weld for a new coating is time-consuming and results in significant material wastage. Consequently, there is a need to optimize the welding process to enhance weld quality while saving both time and materials. An artificial neural network was utilized to predict weld quality by establishing a correlation between critical welding parameters such as welding current, welding speed, welding time, and cooling time on the nugget diameter and the failure type.Additionally, a mathematical model developed through multivariate regression analysis establishes a correlation between welding parameters and weld quality indicators. Optimization is also done using the genetic algorithm to achieve consistent and predictable weld quality with an accuracy greater than 98%. The outcomes of the optimization were subsequently validated through repeatability trials.