Cutting Parameter Correlation Analysis And Surface Roughness Prediction Research

The "Made in China 2025" action plan has determined that smart manufacturing is the development direction of the manufacturing industry.Intelligent processing technology adopts advanced sensors,processing equipment and data processing methods to realize the monitoring and adjustment of the process system through the collection of information on the machining process.This article use turning 45 steel with minimum quantity lubrication as the experimental object.Useing sensors to collect process parameter signals with sensors analyze the correlation between cutting parameters,process parameters and result parameters by Copula function.A Method is proposed that Copula estimation of distribution algorithm and wavelet packet transform optimize neural network.to predict surface roughness.The main research work is as follows:(1)Research progress about correlation,surface roughness prediction and Copula function is reviewed at home and abroad,The definition,type,parameter estimation method and optimization method of Copula function was elaborated,The method is determined about prediction of surface roughness with Copula estimation of distribution algorithm and wavelet packet transform.(2)An experiment plan was designed and conducted that 45 steel was turned with minimum quantity lubrication by controling variable.The influence of cutting parameters(cutting speed,feed rate and cutting depth)on process parameters(cutting force,vibration,cutting temperature)and result parameters(surface roughness,surface hardness,chip deformation)was analyzed.The correlation between the parameters is revealed qualitatively(3)The marginal distribution function of the process parameter and the result parameter is established.C vine Copula model is constructed and the best Copula function is determined based on AIC and BIC.Through the Kendall rank correlation coefficient,the correlation between the variables is systematically analyzed.(4)The relevant cutting force is used as a single input and cutting force and vibration is used as multiple inputs.Four surface roughness prediction models are constructed with Copula estimation of distribution algorithm and wavelet packet transform.The experiment shows that the predictive model that wavelet packet transform optimizates BP neural network with the single-input has the best effect.