Parameters optimization for laser-cladding CoCrFeNiMn high-entropy alloy layer based on GRNN and NSGA-II

Abstract

In order to improve the bonding strength of FeCoCrNiMn cladding layer, laser power, scanning speed and powder feeding rate were selected as the optimization variables; the aspect ratio, dilution rate and heat affected zone depth of the cladding layer were selected as the optimization indicators; a three-factor three-level full-factor experiment was designed; a nonlinear model of cladding parameters and cladding performance was established by generalized regression neural network algorithm (GRNN), while the error of prediction value and experimental results was less than 10%. A set of optimal Pareto frontiers were obtained by the Non-Dominance Ordering Genetic Algorithm (NSGA-II), and the optimal individuals were obtained based on the equal weights, and the optimal cladding parameters were as follows: laser power 2.33 kW, scanning speed 2.90 m/min, and powder feeding rate 12.46 g/min. The cladding layer prepared under the optimal parameters has a dense cross-section with no defects such as cracks and porosity. The joint optimization using the GRNN neural network and the NSGA-II genetic algorithm can obtain excellent results, and the relative error between the experimental and predicted values can be controlled within 10%. The application of optimized laser process parameters can significantly improve defects such as poor adhesion, cracks and porosity on the surface of the cladding layer, and also improve the surface hardness and wear resistance. The experimental results show that the wear rate of the cladding layer prepared by the optimized parameters is 3.21×10−5 mm3/N·m, which is 1.5 times higher than that of the cladding layer prepared by P=2.1 kW, V=3 m/min and C=9 g/min. The wear mechanism mainly includes abrasive wear, plastic deformation and oxidative wear.