Research On Laser Polishing Technology Based On Neural Network Expert System

As a new laser processing technology,laser polishing(LP)has the advantages of less pollution,good flexibility,high polishing precision and wide application range compared with other polishing technologies.It is a promising polishing technology.LP is divided into hot polishing and cold polishing.This paper mainly discusses hot polishing.There are many factors that affect the effect of laser thermal polishing,among which the average power density and spot overlap are the main factors,but these two parameters are closely related to other parameters.In order to improve the precision of hot polishing and verify its feasibility,the LP theoretical model is built,and the hot polishing experiment is designed reasonably.According to the design,the main parameter experiments are carried out to verify the influence of each parameter on the polishing results.In view of the complexity and uniqueness of LP parameters and the dependence of LP parameter selection on experience,in order to improve polishing efficiency,an expert system model based on BP neural network is constructed to recommend polishing parameters efficiently.The main contents of this paper are as follows:First of all,the physical basis of LP is studied,two polishing mechanisms are explained,the law of physical change in LP working process is explored,and the heat source model of LP is built.Abstract and model the surface morphology of the material in the polishing process,creatively put forward the concept of over calculated surface,which extends the analysis basis of the surface modeling from point to surface,improves the accuracy of the theoretical model,and more in line with the actual situation.Using ANSYS to simulate the temperature change in the polishing process,the accuracy of calculating the excessive surface is verified.Secondly,according to the characteristics of LP parameter selection,an expert system based on neural network is studied and built to make the expert system have the learning function and retain the reasoning function of the expert system.The combination of the two can get the optimal parameters of LP more quickly and accurately.The optimal structure of neural network is obtained by training,and the feature extraction method of workpiece is designed.Because of the excellent nonlinear fitting function of neural network,LP parameters can be accurately provided by a large number of data training.Thirdly,based on the existing experimental conditions,the LP experiment is designed,and an improved single factor experiment method is proposed,which can reduce a large number of experimental processes and improve the accuracy and efficiency of the experiment.Finally,based on the improved single factor experiment method,the LP experiment is carried out for seven parameters,such as power,frequency,pulse width,spot diameter,power density,overlap rate and scanning speed,to explore the influence of each parameter or comprehensive parameter on the polishing effect.Through the observation of the surface of the sample,it is concluded that LP plays an important role in improving the surface morphology of the material,verifying the influence of various parameters on the polishing effect;through the component analysis,we can know the oxidation degree of the material surface and the cause of microcracks in the polishing process;through the roughness measurement,we can characterize that laser polishing can effectively reduce the surface roughness of the material,aiming at the roughness of individual material surface The degree of reduction can reach 95.66%,which strongly proves the feasibility of laser polishing.