Multiobjective Optimization Of Laser Cladding Process Parameters Based On GRNN And NSGA-?

With the continuous development of aerospace,biomedical,automobile manufacturing,shipbuilding and other fields,there are more and more demands for complex structural parts.Rapid prototyping demand and structural personalization demand are also increasing.But the traditional manufacturing technology for complex structural parts has the disadvantages of processing difficulties,waste of materials,low production efficiency and so on.Laser cladding technology is a kind of metal additive manufacturing technology with laser as heat source,which is used in the manufacturing of complex parts,the surface coating of various functional parts,the repair of parts and other fields.However,the performance parameters of laser cladding layer are directly affected by the process parameters.How to optimize the process parameters(laser power,scanning rate,powder feeding rate,etc.)to optimize the performance parameters of multiple cladding layers has become a research hotspot.The influence of technological parameters of laser cladding on the macro geometric dimension(height,width,cross-sectional area)of cladding layer is studied theoretically in this paper.The temperature distribution equation of single cladding layer is established by using three-dimensional heat conduction equation,and the temperature distribution of single cladding layer is obtained by solving the equation.Then the temperature distribution law of cladding layer along y-axis is obtained.The law of the influence of process parameters on the geometry of cladding layer is obtained according to mass conservation of laser cladding powder.This paper is based on the experimental data of laser cladding process parameters.The influence trend and secondary order of the process parameters on the performance parameters of cladding layer(cross-sectional area,microhardness,dilution ratio,depth of heat affected zone)are obtained through the visual analysis of the experimental data.Then the significance of influence of the process parameters on the performance parameters is analyzed by variance analysis.The regression analysis of the functional relationship between the performance parameters and the process parameters of cladding is carried out by using the ternary quadratic polynomial model,BP neural network model,RBF neural network model and GRNN neural network model based on the experimental data of laser cladding.And the errors of different models in the training set and the test set are studied.It is found that the error of polynomial model in training set and test set is large.The error of the training set of BP and RBF model is very small,while the error of the test set is large.But the comprehensive error of GRNN model is the smallest.So GRNN model is used as the regression model between the process parameters and performance parameters of laser cladding.The multi-objective optimization of process parameters is carried out based on GRNN regression model of laser cladding process parameters and performance parameters,and three objective functions of microhardness,dilution ratio and heat affected zone depth are established.Firstly,the unified objective method is used to optimize the multi-objective function,and four different unified objective evaluation functions are established to obtain the optimization results.Secondly,the NSGA-II multi-objective evolutionary algorithm is used to optimize the multi-objective function,and the effects of population size,crossover rate and mutation rate on the optimization process and results are studied respectively.The final optimization results are better than the unified objective method.Results compared with that before optimization,microhardness increased by 27.4%,dilution ratio decreased by 82.4%,and the depth of heat affected zone decreased by 3.78%.