Surface Roughness Prediction And Optimization Method For CNC Milling Based On Multi-source Heterogeneous Data

Manufacturing industry is extremely competitive in the world,production quality and production efficiency are the key to the development of manufacturing enterprises.Therefore,how to achieve high quality and production efficiency has become a key research problem for many enterprises and universities.Surface roughness seriously affects the product surface hardness,wear resistance and fatigue strength,and it is often used as a key evaluation index to weigh the quality of mechanical products.So,it is significant to study the prediction and optimization of surface roughness for CNC milling.Sensor technology,communication technology and cloud platform manufacturing technology have made many breakthroughs,which provide strong support for the acquisition of multi-source heterogeneous data of CNC milling.An accurate surface roughness prediction model can be fitted by fusing the multi-source heterogeneous data of CNC milling.And then the optimization of surface roughness can be effectively realized.Therefore,this thesis carried out the research on the modeling method of surface roughness prediction for CNC milling driven by multi-source heterogeneous data and the multi-objective optimization method for CNC milling considering surface roughness and machining time.Firstly,the static and dynamic data of CNC milling are collected under variable process conditions,the static data including process parameters,workpiece material and tool diameter,while the dynamic data include vibration,force and power signals.The original multi-source heterogeneous data are pre-processed to prepare for the creation of prediction model.Secondly,the features of dynamic data are automatically extracted by convolutional neural network(CNN),while the features of static data are extracted by manual extraction.Gaussian process regression is applied to fuse the dynamic and static data features to fit the surface roughness prediction model of CNC milling under variable process conditions.A particle swarm algorithm is adopted to optimize the CNN to improve the prediction accuracy of the model,and the prediction performance of the model is evaluated by the four indexes of determination coefficient,root mean square error,mean absolute percentage error and mean absolute error.Then,a multi-objective optimization model is established with the spindle speed,feed rate,depth of cut and cutting width as the optimization variables,and the surface roughness and machining time of the CNC milling as the optimization objectives;the Tabu Search algorithm is applied to solve the optimization model to obtain the process parameters that minimize the surface roughness and machining time.Finally,the surface roughness prediction and optimization method for CNC milling based on multi-source heterogeneous data proposed in this thesis is analyzed in a case study.The proposed method can be effectively applied to CNC milling surface roughness prediction and optimization and has strong advantages through the verification of CNC milling experiments.