Research On Prediction Of Tool Remaining Useful Life Based On BILSTM And ARIMA Model

As the main body of cutting,the unreasonable and standardized use of cutting tools will lead to the sharp increase of manufacturing cost and the sharp decrease of production quality and efficiency.Therefore,whether the cutting tools can be fully used in the service life of the cutting tools to reduce the production cost is an urgent problem to be solved.In addition to the tool flank wear,the vibration signal of the tool also contains rich information.The current wear state of the tool can be obtained through the vibration characteristics of the tool.It is of great significance to analyze the wear and vibration characteristics of the tool flank,and predict the remaining service life of the tool according to the corresponding change trend,so as to make full use of the tool and reduce the production cost.The main research work is as follows:1.Based on the influencing factors and wear mechanism of tool wear,an experimental system of tool life is built,and the tool life experiments of cutting speed,feed speed and back feed are designed.On this basis,nine groups of orthogonal experiments and one group of control experiments are carried out,and the data of flank wear and vibration are obtained.2.The feature of “fractal dimension” is introduced into the feature of tool vibration signal.Through correlation analysis,8 time-domain features,4 frequency-domain features of tool vibration signal and the trend of fractal box dimension changing with tool number in tool life experiment are compared.The results show that the correlation coefficient between the fractal box dimension and the number of tool runs is the largest,and the fractal box dimension is selected as the sensitive feature of tool wear.3.In order to solve the problem that the failure process of tool wear is dependent on time,a model based on BI-Directional Long and Short-Term Memory Network(BILSTM)is proposed to predict the tool remaining useful life.Bayesian Optimization(BO)algorithm is used to optimize the number of hidden layer units and initial learning rate of BILSTM model,and BO-BILSTM model is used to predict the tool remaining useful life,which is compared with BO-LSTM model and Gray model contrast.The results show that the prediction performance of BO-BILSTM model is the best,but there are still large errors in the prediction.4.In view of the large error of prediction of BILSTM model,combined with the advantages of high short-term prediction accuracy of Autoregressive Integrated Moving Average model(ARIMA),a new hybrid model of tool remaining useful life prediction based on the BILSTM-ARIMA is proposed,and the tool remaining useful life is predicted by using the hybrid model.The results show that the performance of the hybrid model is better than that of BO-BILSTM model,the prediction error is reduced and the prediction accuracy is higher and the hybrid model can effectively improve the prediction accuracy of BILSTM.