基于“SC_ISSA-BP网络”驱动的激光熔覆表面粗糙度优化

Laser Cladding Surface Roughness Optimisation Driven by SC_ISSA-BP Network

在激光熔覆过程中,工艺参数多样化,导致结果控制表现出非线性关系。通过深入分析各参数对熔覆层的影响,可以快速获得最优工艺,提高熔覆层的性能,推动激光熔覆技术的应用。基于正交实验设计,对实验数据进行极差分析,研究工艺参数对激光熔覆CoCrFeNiMo_(0.2)涂层表面粗糙度的影响。通过超景深扫描仪对实验样本进行微观分析,获得其表面粗糙度值。根据算法对工艺参数进行优化,探讨激光功率、送粉速度、扫描速率和搭接率多因素耦合作用下对激光熔覆多道搭接涂层表面粗糙度的影响,以优化最佳工艺参数组合,同时获得最优表面粗糙度。以拟合程度为指标对比优化模型,BP神经网络为94.79%,SSA-BP神经网络为96.981%,SC_ISSABP神经为98.528%。传统BP神经网络的误差指标均方误差根RMSE为58.3858μm,而SSA-BP神经网络的RMSE为51.2974μm,SC_ISSABP神经网络为43.9408μm。SC_ISSABP神经网络的优化能力最为明显。

In the process of laser cladding,the process parameters are diversified,resulting in the resultant control showing a nonlinear relationship.By deeply analysing the influence of each parameter on the cladding layer,the optimal process can be obtained quickly to improve the performance of the cladding layer and promote the application of laser cladding technology.In this paper,based on the orthogonal experimental design,the experimental data were analysed in terms of extreme deviation to study the influence of process parameters on the surface roughness of laser melted CoCrFeNiMo_(0.2)coatings.The experimental samples were analysed microscopically by means of an ultra-depth-of-field scanner to obtain their surface roughness values.The process parameters were optimised according to the algorithm to investigate the effects of laser power,powder feeding speed,scanning rate and lap rate on the surface roughness of the laser melted multi-lap coatings,in order to optimise the best combination of process parameters and obtain the optimal surface roughness at the same time.Comparing the optimised models in terms of the degree of fit,the BP neural network is 94.79%,the SSA-BP neural network is 96.981%,and the SC_ISSABP neural is 98.528%.The Root Mean Square Error RMSE,an error metric for the conventional BP neural network,is 58.3858μm,whereas the RMSE for the SSA-BP neural network is 51.2974μm,and for the SC_ISSABP neural network is 43.9408μm.The optimisation ability of the SC_ISSABP neural network is the most significant.