Research On Optimization Of Surface Roughness Of Laser Selective Melting Forming Part Based On BP Neural Network

The special forming principle of laser selective melting makes the surface quality of molded parts unsatisfactory,which will restrict the development of laser selective melting.The optimization of the surface quality of the molded parts is inseparable from the research on the surface roughness,which is affected by the forming parameters and other external conditions.Generally,the study of surface quality requires repeated experiments to obtain molded parts,which not only has a long research period,but also causes waste of materials.For this reason,this paper selects the surface roughness as the characterization of the surface quality of the molded parts.For the purpose of optimizing the surface quality,different parameter combinations are designed,and the BP neural network is used as the medi?m to provide a new way to optimize the surface roughness of the laser-selected melting molded parts.In view of the highly nonlinear characteristics of the surface roughness of the laser melting forming parts and the process parameters,the BP neural network surface roughness prediction model was first established to predict the laser power,scanning speed,powder layer thickness,and powder spacing.The surface roughness of the molded part under the combination of different process parameters.The results show that the average relative error of the sample prediction is 10.55 percent,indicating that the BP neural network can express the nonlinear coupling relationship between the laser selective melting forming process and the surface roughness.In order to solve the problem that BP neural network is too sensitive to initial weights and thresholds,this paper uses genetic algorithm to optimize the ability to provide appropriate initial weights and thresholds to BP neural network.After genetic algorithm optimization,the average relative error of BP neural network sample prediction is reduced to 6.91%,indicating that genetic algorithm is effective in optimizing the BP neural network surface roughness prediction model.In order to further verify the validity of the model,the orthogonal method was used to design the experiment,and then through the model prediction,the predicted surface roughness value was analyzed,showing that under the orthogonal method design experiment,the influence law of each parameter in the predicted data on the surface roughness realistic.The neural network prediction model optimized by the genetic algorithm is used as the fitness function in the genetic algorithm,and the surface roughness value is used as the fitness value to establish an optimization model for the surface roughness of the laser-selected melting forming part.Select the combination of process parameters obtained by optimization(a)P=120W,V=500mm/s,H=0.12 mm,T=20?m(b)P=120W,V=1300mm/s,H=0.12 mm,T= 20?m(c)P=60W,V=500mm/s,H=0.09 mm,T=20?m(d)P=110W,V=1200mm/s,H=0.07 mm,T=30?m,etc.to verify and find the molded part The surface is smooth,the texture is clear,the melt channel is completely melted,and the measured value of surface roughness has little fluctuation,and all are lower than the sample value.The results show that the surface roughness obtained by the optimization has reached the expected goal,and the corresponding parameter combination can also meet the actual situation.It further shows that the combination of the optimization algorithm and the BP neural network can be used to optimize the surface quality of the laser-selected melting forming part.