Research On Application Of Multi-sensor Information Fusion In Tool Wear Condition Monitoring

The development level of mechanical automation reflects the economic level and scientific and technological strength of a country.The level of mechanical processing is greatly affected by the degree of tool wear.Because tool wear is unavoidable during machining,and single sensor monitoring is insufficient,multi-sensor real-time online monitoring of tool wear status becomes the key to solving this problem.The multi-sensor information fusion process needs to study the sensor technology,information collection technology,information fusion technology,through signal collection,feature selection,feature extraction and other links to achieve effective complementation of multi-sensor information,complete online monitoring of tool wear.The main research contents of this article are as follows:First,based on the relevant theories of tool wear,this article first considers the sensor type,signal characteristics and tool model,selects test equipment,and builds a signal acquisition system platform for vibration and cutting force signals;In the process,the vibration signal and cutting force signal of each pass and the tool wear VB value after each pass are collected;finally,according to the relevant theory of the tool wear process,the factors affecting the durability of the tool are analyzed,and the orthogonal test method is used The vibration signal and the cutting force signal were respectively tested for the effect of factors on the data,and the effects of the degree of wear,spindle speed,feed speed and cutting depth on the two signals were analyzed.Secondly,the factors that affect the durability of the tool are analyzed.The orthogonal experiment method is used to perform the "effect experiment on the factors" on the vibration signal and the cutting force signal and The degree of influence of the two signals analysised by the wear degree,feed speed,spindle speed and cutting depthThen,the time domain,frequency domain and time-frequency domain analysis methods were used to analyze and extract the collected signals.First,the two signals are analyzed using time domain and frequency domain analysis methods.Then,in the characteristic frequency band range obtained by analysis,the wavelet packet analysis method is used to decompose the signal frequency band and calculate the energy value of the different frequency bands of the signal.The test method analyzes the frequency bands where the energy values of the vibration signal and the cutting force signal change significantly,and extracts the optimal frequency band as the fusion feature value.Finally,BP neural network-optimized Bayesian algorithm is used to establish an information fusion intelligent recognition system based on the combination of feature layer and decision layer.The characteristics of BP neural network with variable learning rate are selected for feature layer fusion,and then the optimized Bayesian fusion algorithm with Kalman filtering is used for decision layer fusion,and finally the fusion tool wear prediction value is obtained.A series of experiments were carried out to monitor tool wear status,which confirmed the reliability and feasibility of tool wear status monitoring combined with BP neural network and optimized Bayesian fusion algorithm,which opened up new ideas for the development of tool wear monitoring direction.