Research On Tool Condition Monitoring Based On Milling Sound Signal Analysis

Real-time monitoring of tool wear is the key technology of automation and intelligent processing.It has been a hot research topic both at home and abroad.In the cutting process,the tool wear condition affects the accuracy and surface quality of the work-piece directly.Therefore,it is of great importance for improving product quality,production efficiency and automation level to realize real-time monitoring of tool wear condition.This paper carried out the following research based on milling process.First of all,this paper introduced the characteristics of milling process,summarized the research status of tool condition monitoring at home and abroad,and analyzed the advantages and disadvantages of various monitoring methods.The paper put forward the method that took the milling sound signal as the main and the force signal as the auxiliary object in order to monitor the wear condition of milling cutter.This paper also designed the acquisition and analysis software system of the sound signal based on the Lab VIEW and set up the tool wear condition monitoring system.The milling experiments were carried out to complete the signal acquisition under different tool wear conditions.Secondly,this paper analyzed the machine tool idling noise and the normal milling noise in time and frequency domain and ultimately dealt with the interference of the machine noise through the high pass filter.This paper obtained the correlation between the signal characteristics and tool wear condition by the time and frequency domain analysis.Besides,the influence of the milling parameters o n the sound and force signals was discussed by the single factor experiments.Then,the signal was divided into eight frequency bands by using the theory of wavelet analysis in order to study the change of energy in different frequencies under different tool wear conditions,and obtained the most related frequency range according to the tool wear condition.Next,this paper extracted the percentage of energy in different frequency bands and the mean value of X and Y direction of milling force signal as the characteristic values of tool wear condition.Finally,the pattern recognition and its commonly used methods were studied.And the three-layer BP neural network was selected as the recognition classifier of the tool wear condition.The paper took the normalized characteristic values as the input vector of the neural network and the tool wear condition as the output vector.The neural network was trained by the training samples and the BP neural network structure(8-9-3)was selected by analysis and comparison.The trained BP neural network was tested by the test samples.And the actual outputs of the neural network were normalized at last.The results showed that the BP neural network can accurately identify the tool wear condition.This paper successfully established the relationship between tool wear condition and milling sound signal and force signal through the theoretical analysis and experimental research and realized the non-contact measurement and accurate recognition of the tool wear condition,providing new thoughts and methods in tool wear condition monitoring for the actual production.