Study On Surface Roughness Prediction Based On Copula Function In Milling

As the key technology of intelligent manufacturing,intelligent machining needs to monitor the online machining process and collect detection signals including the cutting force,vibration,sound emission and temperature.By extracting related features,it can judge the state of the cutting process and predict the processing quality,tool durability and the system stability,which can adjust cutting parameters to realize intelligent control.Using milling 45 steel with micro lubrication as the experiment subject,this paper proposes a prediction method of surface roughness based on Copula function and neural network.The main research work are as follows:(1)The prediction method of surface roughness and the application of Copula function at home and abroad are reviewed.The basic concepts and analysis methods of the correlation are expounded.The characteristics of the Copula function method and the traditional correlation analysis method are compared.The feasibility of applying the Copula function to the correlation of cutting processing is analyzed.(2)By establishing the experiment platform of MQL milling,effects of cutting parameters on all component cutting force,cutting force,vibration and surface roughness were analyzed.It was found that apart from the axial force,influences of the cutting velocity,feed rate per tooth and the cutting depth on the cutting force,radial force,tangent force,vibration and surface roughness were basically the same.(3)Five kind of the Copula functions were selected respectively to fit the sample data of cutting force and surface roughness,vibration and surface roughness.Kendall rank correlation coefficient is derived to describe the global correlation by comparing above-mentioned Copula functions.Tail correlation coefficient of mixed Copula function based on Clayton & Gumbel was used to describe the tail correlation.The result shows that the upper correlation of the cutting force and surface roughness is relatively larger.The upper correlation of the vibration and surface roughness is also larger,while the tail correlation is nearly zero.(4)Three kind of the Copula functions(single two-dimensional,single three-dimensional and mixed)and two kind of neural networks(BP algorithm and Copula EDA-BP hybrid algorithm)are proposed for surface roughness prediction.Through the experimental verification,the result shows that the prediction method based on mixed Copula function and the neural network prediction based on the Copula EDA-BP hybrid algorithm have the best effect.