Rapid optimization of iron-based alloy laser cladding process based on orthogonal experiment and machine learning for Q345

The properties of laser cladding coatings are closely correlated with process parameters. The laser power (P), scanning speed (V_S), and powder feeding rate(V_F) were optimized to enhance the comprehensive properties of laser cladding iron-based alloy single coating on Q345, with a focus on achieving targets such as dilution rate (η), aspect ratio (W/H), and hardness (H_D) of single coating. Firstly, the impact of process parameters on coating properties were analyzed through orthogonal experiments. Subsequently, prediction models between process parameters and coating properties were established by using back propagation neural network(BPNN), which optimized by particle swarm optimization(PSO) and Genetic Algorithm(GA). Finally, non-dominated sorting genetic algorithm II(NSGA-II) was employed to optimize the process parameters, and the optimized results were verified and analyzed. The findings indicate that laser power is the primary factor influencing dilution rate and hardness, while powder feeding rate primarily affects aspect ratio. The optimized process parameters include a laser power of 934 W, scanning speed of 352 mm/min, powder feeding rate of 0.64r/min, corresponding aspect ratio of 3.06, hardness of 613HV, dilution rate of 0.33. There were no obvious defects in the coating and the coating hardness was three times higher than that of the substrate. The optimization results of orthogonal experiments are constrained by the process range, fortunately, the utilization of machine learning optimization methods can effectively address these limitations and avoid the re-experiment due to improper settings in orthogonal experiments. It offers a rapid and efficient method for optimizing laser cladding process parameters.