Multi- objective modeling and optimization of dissimilar laser welding by integrating an artificial intelligence predictive model with NSGA-II algorithm

Abstract

This study presents a pioneering approach in laser welding of two dissimilar materials by integrating an Artificial Intelligence (AI)-based on predictive model with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) for optimizing key performance characteristics of Nickel-based alloy and duplex 2205 stainless steel. Focusing on four primary input variables including laser power, welding speed, focal distance and deviation, the research aims to predict and optimize the responses including temperature field adjacent to the melt pool, penetration depth, and tensile strength of the joint according to the experimental results. The developed AI model first accurately forecasts these characteristics based on the inputs. This predictive accuracy is critical in defining the optimal target values. Considering the multidimensional nature of the problem, where enhancing one characteristic could compromise another, the study employs a Multi-Objective Optimization (MOO) strategy. This is where the innovative integration with NSGA-II becomes pivotal. Renowned for its efficiency in navigating multiple, potentially conflicting objectives, NSGA-II assists in achieving a balanced optimization of all target parameters. This method is adept at considering the complex interdependencies among various characteristics. The novelty of this work lies in its unique combination of AI for prediction and MOO for optimization and it is for the first time that machine learning is applied on this kind of alloys with four features and four targets.