Prediction And Optimization Of Machine Learning Based RSM-CCD For Low Carbon And Energy Saving Hobbing

Computer Numerical Control(CNC)gear hobbing machine tool is the working machine tool of gear,which is the key part of transmission.Its total energy consumption and carbon dioxide emissions are huge in the gear hobbing process,and it has great potential of energy saving and CO₂ emission reduction.Accurate prediction of energy consumption and carbon emissions is an important prerequisite for the study of optimization.However,prediction models were constructed from the perspective of experiment in most existing studies.With the development of artificial intelligence,scholars find the machine learning models showing better prediction performance.However,most of Chinese enterprises have not realized the workshop networking and automatic collection of manufacturing data,which hinders the application of machine learning methods.It's a worthing problem of deep reflection that how to build a better prediction model to optimize the hobbing process from the perspective of machine learning.Based on this,this paper took the gear hobbing process in front of the heat-treatment in a selected gear company as the research object,in addition to,studied the prediction of energy consumption and carbon emissions with the multi-objective optimization of process parameters.Besides,this paper proposes a machine learning prediction and optimization method based on Central Composite Design(CCD)in Response Surface Method(RSM)to realize CNC gear hobbing high-efficiency,energy-saving and low-carbon operation of processing.Firstly,the energy consumption characteristics and carbon emission characteristics of CNC hobbing machine tool under axial hobbing process(up cutting)were analyzed,besides,the energy consumption model based on time characteristics and the carbon emission model on generalized boundary of machining system were constructed respectively.Further more,the research objectives and data acquisition methods of sensor collection and numerical simulation were proposed.Secondly,based on the principle of RSM and the Design-expert v12,the CCD scheme of hobbing process in front of the heat-treatment for the second gear of a gearbox's shaft in the selected company was designed.On this basis,the RSM model of CNC gear hobbing was constructed by using the simulation results of CCD and the Design-expert v12,and the accuracy and effectiveness of the RSM model was verified by analysis of variance(ANOVA)and residual analysis.Then,a small sample machine learning prediction method based on CCD in RSM is proposed.Three popular machine learning algorithms(BP Neural Network,Support Vector Regression and Random Forest)were selected to construct the multi-objective prediction model by using the data set of the CCD in the previous chapter.Compared with the response surface model,it is found that the prediction effect of machine learning model is better,and the prediction effect of BP neural network in machine learning model is better than that of support vector regression and random forest.Finally,a multi-objective optimization model with CNC gear hobbing parameters as optimization variables was constructed,and a multi-objective optimization method of BP neural network-MOGWO was proposed to solve the model.The EM-TOPSIS comprehensive evaluation technology was used to select the optimal process parameters.The results show that the optimized parameters can realize the high efficiency,energy saving and low carbon operation of CNC gear hobbing.On this basis,the superiority of MOGWO algorithm was verified by NSGA-II algorithm comparison.