基于神经网络和遗传算法的激光熔覆工艺参数多目标优化

Multi-objective Optimization of Laser Cladding Process Parameters Based on Neural Network and Genetic Algorithm

为了提高再制造工件的激光熔覆层的综合质量,选取激光功率,送粉量,扫描速度为优化变量,熔覆层的宽高比、稀释率、粉末收集率作为优化目标,基于综合加权法与层次分析法将3个优化目标转化为综合质量目标,设计全因子试验,利用MATLAB软件基于试验结果建立BP神经网络预测模型,通过遗传算法确定使综合质量达到最佳的工艺参数组合。研究结果证明,装备工件再制造激光熔覆的最优工艺参数组合为:激光功率3.0 kW,送粉量47 g/min,扫描速度5.5 mm/s。

In order to improve the comprehensive quality of the laser cladding layer of the remanufactured workpiece, the laser power, the powder feeding amount and the scanning speed are selected as the optimization variables, and the aspect ratio, dilution rate and powder collection rate of the cladding layer are selected as optimization targets, based on the comprehensive weighting method. The analytic hierarchy process is used to transform the three optimization objectives into comprehensive quality objectives, design the whole factor experiment, and use the MATLAB software to establish the BP neural network prediction model based on the experimental results, and determine the combination of the best process parameters by genetic algorithm. The optimal process parameters for re-manufacturing laser cladding of equipment parts are: laser power is 3.0 kW, powder feeding is 47 g/min, and scanning speed is 5.5 mm/s.