Research On Service System Of Milling Cutter Life Prediction Based On Deep Learning

With the widespread application of information technology such as industrial sensors,automatic control systems,and industrial Internet in the industrial field,a large number of tool sensor signals are collected and transmitted in real time.Because the collected data itself has outlier data,and often in the data transmission process There will be some data missing,which will make it impossible to accurately predict the remaining service life of the tool during the machining process.Therefore,filling in vacant data and detecting outlier data from the collected tool sensor signals can improve the accuracy of the remaining service life of the tool,and it is of great significance to improve the processing accuracy of the workpiece and the processing efficiency of the production.In this paper,the tool condition monitoring signals(cutting force signals,vibration acceleration signals and acoustic emission signals)collected during milling are used as the research object,and the deep learning framework is used to study the vacancy data filling algorithm and outlier data detection algorithm under the multi-dimensional sensor signal.And the tool life dynamic prediction algorithm.Finally,a tool life prediction service system is established by using computer technology and research content.The main research contents of this paper are as follows:(1)Improvement of the algorithm for filling vacant data of the tool sensor signal.Aiming at the situation that the traditional data filling algorithm is not suitable for multi-dimensional sensor signal data filling,the deep convolutional neural network is used to extract the relationship between the multi-dimensional sensor signals,and the long and short-term memory neural network is used to capture the timing of the sensor signal.Relationship,so as to establish a predictive model of tool sensor signal vacancy data filling.And for the multi-output prediction algorithm,the loss function weighting method is improved to improve the accuracy of multi-dimensional sensor signal prediction.And compared with the traditional data filling algorithm to verify the effectiveness of the improved algorithm.(2)Improved algorithm for outlier detection of tool sensor signal.In view of the different tool sensor signal data distributions reflected in the different wear stages of the tool,the concept of concept drift is introduced,and the detection standards used for outlier data detection in different wear stages are also different.Therefore,the tool status is monitored from the perspective of data distribution.Whether the signal has concept drift is detected,so as to update the threshold of outlier data detection,improve the accuracy of outlier data detection,and compare experiments with traditional outlier detection algorithms to verify the effectiveness of the method.(3)Dynamic prediction of tool life.Aiming at the inconsistent tool monitoring signal data distribution under the same working conditions and the same model,this paper uses the deep convolutional neural network to mine the tool life characteristics and adds an attention mechanism Strengthen the learning of tool life characteristics and establish a tool life prediction model.Based on the idea of transfer learning,a dynamic update method of tool life prediction model is proposed to obtain a more reliable tool life prediction model,improve the accuracy of tool life prediction,and compare experiments with traditional tool life prediction models to verify the effectiveness of the method.(4)The design and realization of the service system for predicting the life of milling cutters.Based on the research of tool sensor signal vacancy data filling,outlier data detection and tool life dynamic prediction,through analysis of the functional requirements of milling cutter life prediction services,through hypertext markup language(HTML),cascading style sheets(CSS),Java Script and other front-ends Technology,Python application development technology,My SQL database,etc.realize a milling cutter life prediction service system with functions such as vacancy data filling,outlier detection and tool life dynamic prediction.