Research On The Wear Prediction Model Of Milling Tool Considering The Working Condition Information

Intelligent manufacturing has become an important trend in the global manufacturing industry,in the field of automated production,CNC milling machine has become an important part of intelligent manufacturing by virtue of its strong adaptability,stable processing quality and high production efficiency.As the core element of machine tool production and processing,the tool is one of the key factors for the success of milling processing,and it is also a component that is prone to consumption and damage,and accurate,timely and effective monitoring of the wear state of the tool is one of the important ways to improve the milling processing capacity.The working conditions of the tool in the actual production are complex and changeable,resulting in the poor performance of traditional wear prediction methods.Therefore,how to reduce the influence of working conditions and ensure the accuracy of tool condition monitoring has become one of the urgent problems to be solved in the development of milling processing to intelligence.This thesis takes the end mill as the research object,combining machine learning methods and deep learning methods to establish tool wear prediction model,focusing on the accuracy of tool wear prediction under different working conditions,the tool wear prediction model and model migration based on deep learning method are explored.The aim is to provide a more efficient method for predicting tool wear under different operating conditions,the specific research content is as follows:(1)Multi-working condition tool wear experimental design and data set construction,analyze the experimental conditions of the public data set,and point out the problems existing in the existing data set in terms of data quality and data integrity.Based on this,a cutting experiment signal acquisition platform is built,and tool wear experiments under different working conditions are designed to construct a multi-working condition tool wear data set,which provides data support for the subsequent research on tool wear prediction model.(2)Wear change analysis and feature extraction and optimization,the tool wear change trend and wear form are studied,the multi-source signal generated during the experimental cutting process is collected,the data preprocessing is carried out,the signal is denoised by variational mode decomposition and wavelet threshold noise reduction method,the time domain characteristics,frequency domain characteristics and time frequency domain characteristics of the signal data are extracted,and finally 50 signal features with high correlation are screened out by Pearson correlation coefficient for subsequent tool wear prediction.(3)For the tool wear prediction of single working condition,multiple regression method and machine learning method are used to establish tool wear prediction model,and a convolutional long short-term memory neural network tool wear prediction model based on attention mechanism is proposed,and the model structure and parameter tuning are elaborated.Through the comparative analysis of multiple model prediction results,the effectiveness and advantages of the model applied to tool wear prediction are proved,it lays the foundation for tool wear prediction considering working condition information.(4)For the tool wear prediction under different working conditions,the above convolutional long short-term memory neural network tool wear prediction model is studied by model transfer method.The prediction results show that tool wear prediction under different working conditions can be realized through model migration,which proves the feasibility and effectiveness of model migration.Finally,on the basis of the above research,the condition monitoring system of milling tools is developed and designed,and the effectiveness is verified by example operation.