Research And Application Of Key Process Parameters Prediction Model Based On Machine Learning

In recent years,with the vigorous development of China’s social economy,industrial production will generate massive data.The high complexity of massive data will seriously affect the parameters of subsequent processes,and the deman d for prediction of process parameters will become increasingly high.However,i n the period of underdeveloped intelligent manufacturing process,the prediction o f subsequent process related parameters requires experienced teachers to operate t hrough observation and perception.The emergence of machine learning provides a turning point for this problem.As one of the core technologies of machine learning,BP neural network’s p owerful parallel distributed processing ability and self-organization ability play a vital role in the production process.Because the accuracy of the traditional BP n eural network prediction model is not very high and the initial weights and thres holds are random,it requires constant error reverse iteration to determine the opt imal solution.The number of iteration steps will be large,which will lead to th e problem of high delay.Therefore,how to reduce the number of BP neural net work iteration steps and improve its accuracy has become a key issue to conside r.To solve these problems,this paper proposes two improved combination forec asting methods.(1)By combining the combined prediction model of GA genetic algorithm a nd BP neural network,the weight and threshold of BP neural network are optim ized directly,which effectively reduces the number of iteration steps and the res ponse time of the prediction model.(2)Aiming at the low accuracy of BP neural network prediction model,the grey model is introduced on the basis of the above methods.Due to the high co mplexity of industrial process data,the prediction model needs to have a more c omprehensive feature capture capability.This model uses the ability of gray mod el and BP neural network to capture linear and nonlinear features respectively to improve the accuracy of prediction.The experimental results show that the accur acy of the prediction model can be greatly improved by using this model.