Research On Ultrasonic Tool Wear State Identification And Life Prediction Technology

Ultrasonic composite processing technology is widely used in the field of Nomex honeycomb composite processing because of its small cutting force and high processing precision.Ultrasonic tool is an important part of ultrasonic composite processing equipment.Studying its wear state and remaining life has theoretical significance for the identification of acoustic system faults.It has important engineering value for reducing enterprise economic expenditure and improving part processing quality.The main research work and results of this paper are as follows:1.Straight blade cutter wear stage division and monitoring signal acquisition experiment.By collecting and observing the straight edge cutter,the typical wear form of the straight edge cutter is obtained,and the wear bandwidth of the cutting surface is selected as the wear measurement value.The ultrasonic cutting experimental platform was built to complete the cutting force of different cutting lengths and the collection experiment of cutting acoustic signals.The relationship between the length of the cutter,the root mean square value of the main cutting force and the tool wear value was studied.The wear state of the straight edge cutter was divided into three stages,and the wear state curve was drawn.2.Extraction and dimension reduction of straight blade cutter wear characteristics.The base noise of the collected acoustic signal is analyzed,and the acoustic noise of the acoustic signal is removed by an elliptical high-pass filter.The time domain,frequency domain and wavelet packet analysis method are used to extract the feature signals of the force signal and the filtered sound signal,and the principal component analysis technique is used to complete the dimensionality reduction of the feature vector.3.The establishment of the straight edge knife wear state recognition model.A number of BP neural network classifiers are constructed,and K-nearest neighbor clustering analysis method is adopted to make each BP neural network classifier realize adaptive acquisition of weights.The wear-recognition decision of multi-BP neural network classifier is obtained by linear weighted fusion technique.Through experimental tests,it is found that the correct rate of straight blade knives wear based on multi-classifier fusion is 92.3%,which is better than the traditional single BP neural network classifier.4.Establishment of Prediction Model for Residual Life of Straight Blade Knife.The BP neural network and HMM model were used to establish the residual life prediction model of straight blade.Using the test samples to test the two models trained,it is found that the accuracy of the remaining life prediction of the straight edge cutter based on the HMM model is higher,especially in the severe wear stage of the straight edge cutter,the prediction accuracy is better than that of the BP neural network model.