Milling Chatter Online Monitoring Method Based on Nonlinear Adaptive Decomposition and Shannon Entropy

Abstract

The present invention provides an online milling chatter monitoring method based on nonlinear self-adaptive decomposition and Shannon entropy, comprising: step 1: collecting the actual measurement signal of the main shaft vibration during the milling process through the accelerometer installed at the end of the main shaft; The linear adaptive decomposition method decomposes the measured signal; step 3: according to the energy aggregation characteristics of the flutter signal, the energy-normalized Shannon entropy of the measured signal is comprehensively calculated by using the components of each order obtained from the decomposition, and used as a quantitative index for flutter identification; step 4: Calculate the decrease percentage of the Shannon entropy value of the measured signal in the current spindle rotation cycle compared with the previous cycle and compare it with the empirical threshold to determine the current milling status and perform stability monitoring and chatter warning. The invention has the characteristics of insensitivity to noise and high calculation speed, and is suitable for on-line monitoring in engineering practice.

Description

technical field [0001] The invention relates to the technical field of machining state monitoring, in particular to an online milling chatter monitoring method based on nonlinear adaptive decomposition and Shannon entropy. Background technique [0002] Milling chatter is a self-excited vibration phenomenon that occurs in the machine tool system during metal milling, and has become a common limiting factor in production efficiency and part quality in industrial production; especially in some large, high-end thin-walled parts (such as aircraft Mongolia In the process of high-speed and precision milling, the rigidity of the workpiece is low and the cutting force is large, so it is difficult to fundamentally eliminate the occurrence of chatter only relying on the selection of process parameters. Therefore, milling chatter online monitoring technology has been a hot topic in the manufacturing industry for many years. Flutter signals are nonlinear, non-stationary, and multi-compo...

AI Technical Summary

Problems solved by technology

Both Japan’s Agus Susantoa and Xi’an Jiaotong University’s Cao Hongrui used the Empirical Mode Decomposition (EMD) method to decompose the milling force signal and extract chatter features. Although the nonlinear signal can be effectively decomposed by EMD, this method does not overcome the existing It is difficult to popularize and apply it in engineering practice because of its noise sensitivity and mode aliasing; Liu Changfu from Northeastern University used the variational mode decomposition (VMD) method to decompose the vibration signal of the spindle, but this method is limited by the working principle of VMD (adaptive band filter bank), the obtained sub-components lose their original nonlinear characteristics, so they do not have a clear physical meaning, and the decomposition results make it difficult for people to intuitively understand

[0004] Patent document CN107229795A (application number: 201710408087.X) discloses a milling chatter recognition method based on variational mode decomposition and energy entropy. Use the energy entropy index to extract the flutter feature vector for each mode. Similarly, this method is also limited by the above-mentioned VMD algorithm and cannot obtain ideal decomposition results.

[0005] Based on the existing search literature, it is found that the commonly used milling chatter monitoring methods generally use the traditional EMD or VMD method to pre-process the chatter signal. This type of method mainly has the following problems: 1) It cannot retain the abnormality of the chatter signal. Linear and non-stationary characteristics, which have a negative impact on the subsequent feature extraction process, making each sub-component lose its actual physical meaning; 2) It is extremely sensitive to noise, and it is difficult to achieve flutter recognition in the case of low signal-to-noise ratio; 3) Some high-precision algorithms are too complex and time-consuming to calculate, and the process of determining algorithm parameters is very complicated, which is difficult to apply to online monitoring

Images