Research On Remaining Useful Life Prediction Technology Of Tool

As the most critical executing element in machining,tool is the most sensitive and vulnerable part in machining production.The quality of the tool directly not only affects the quality of the workpiece,but also affects the production efficiency of mechanical equipment.If we can correctly identify the wear state of the tool and accurately predict its remaining service life,it can not only timely replace the damaged tool to ensure the processing quality of the workpiece,but also improve the processing efficiency of the whole equipment production.Therefore,it is of great significance to study the identification of tool wear state and prediction of tool remaining service life.Taking milling cutter as the research object,this paper proposes a tool wear state recognition and residual service life prediction model based on neural network and short and long time memory neural network,and verifies the effectiveness of the proposed method through experiments.The main research contents are as follows:(1)Firstly,the research status of tool wear status identification and residual service life prediction at home and abroad are introduced.The PHM2010 high-speed milling data set used in this paper is introduced.The equipment used in the experiment and the cutting conditions of the tool are given.The tool wear mechanism,wear form,wear classification and wear stage are analyzed,which makes the necessary theoretical basis for the research of this paper.(2)According to the cutting force signal,vibration signal and acoustic emission signal collected in the experiment,the invalid data at the beginning and end were removed and the outlier processing was carried out to obtain a clean and tidy experimental signal.Feature extraction was carried out on the pre-processed signals from the three aspects of time domain,frequency domain and time-frequency domain,and the features strongly related to tool wear were screened.Principal component analysis(PCA)was used to reduce the dimension of the screened features,and the final feature set highly related to tool wear was obtained.As the input of subsequent wear state identification and remaining service life prediction model.(3)In order to better divide the tool wear stage,the neural network and fuzzy system are combined to construct a fuzzy neural network with adaptive learning ability.The related concepts in fuzzy theory are introduced,and the derivation of main(4)parameter learning methods in the model is given in detail.By analyzing the significant characteristics of tool wear state,the tool wear state and the interval of each wear state were pre-divided by the criterion,and then the optimal tool wear stage interval was determined according to the average recognition accuracy criterion.Combined with the experimental data,the feasibility and effectiveness of the proposed method are verified.(5)LSTM has the feature of effectively learning the development law of time series data.Based on LSTM network and using Bi-LSTM to simultaneously use the past and future information in the data,a prediction model of tool remaining service life based on Bi-LSTM+Attention mechanism is established.The Adam algorithm is used to optimize the main parameters in the model,and the Dropout mechanism is introduced to prevent overfitting.By comparing with the basic LSTM and Bi-LSTM network models,the superiority of the remaining service life prediction model proposed in this paper is proved.