AE-based Fault Diagnosis Of Milling Cutter

On-line monitoring of the metal cutting process for machining parts to ensurequality, reduce tool failure and machine failure, increase the level of automation isimportant. Milling process will produce rich acoustic emission signals, which areintimately related to the tool condition. In this paper, acoustic emission signal isanalyzed, and the fault diagnosis of cutter is studied.First, build a milling acoustic emission signal acquisition and analysis system onthe basis of virtual instrument technology. Based on acoustic emission signal, dataacquisition system, denoising system, parameter analysis system and the frequencydomain analysis system are developed by using LabVIEW combined with MATLAB.On the basis of design test method of various cutting tools, acoustic emission signalsin different operational states are collected and processed by wavelet thresholddenoising.Then, the time domain parameters and frequency domain analysis of acousticemission signals are used to analysis.According to the method of acoustic emissionparameter analysis, the total ring counts and total event counts are going on adownward trend as tools wear mounts up, the total energy and RMS however are quiteopposite. After spectrum analysis, the conclusion is that with the intensification ofcutter wear, the energy moves to lower frequency, the power spectral amplitude tendsto go upward, but three frequency parameters such as gravity, root mean squared andstandard deviation are going on a declining curve.After that, the technique of neural networks is applied to mill state recognition.The recognition theory of BP neural network which based on milling acousticemission signal is researched; the work of BP neural network is analyzed; thedetermination of BP neural network architecture and sample and some other importantissues are discussed; the principles and methods to solve the problems such as thequantity of network layers, the number of each layer nodes, the initialization ofnetwork parameter, the selection of sample, the construction of specimen, the processof sample are provided; and BP neural network is established in order to achievemilling cutter recognition.At last, the constructed system and the proposed method is validated by millingtest in this paper. The test shows that the AE signal acquire analysis system based on LabVIEW has high response speed and friendly interface. It is easily operational andpracticable. Meanwhile, the method of cutter fault recognition based on BP neuralnetwork is valid.Compared with the other identification methods，the proposedmethod overcomes the shortcomings of over-reliance on the experience andbackground knowledge.