Tool Wear Monitoring And Remaining Useful Life Prognosis

The wear of tools can lead to low dimensional precision and poor surface integrity during machining. When tool wear is severe, it may also result in tool breakage and damage of the workpiece. To avoid these situations, it is necessary to monitor the wear of tools and predict the remaining useful life of tools. This thesis studied the tool wear diagnostic and prognostic model.Firstly, monitor the cutting process using dynamometer, accelerometers and acoustic emission sensor, and extract the time domain feature, frequency domain feature and timefrequency domain feature from the collected signals. By comparison, the features extracted from the cutting force are better than the features extracted from the vibration and acoustic signal. Among the extracted features, the best force feature is the power at blade passing frequency from the signal of feed direction, the best vibration feature is the amplitude at blade passing frequency from the signal of feed direction, and the best acoustic emission feature is the average value.Then select the sensitive and robust features based on a distance metric. Finally model the tool weal process using Hidden Markov Model(HMM), and predict the continuous tool wear through a probabilistic approach. Different from the tool wear classification based on HMM, this model can obtain continuous tool wear value, which can provide more accurate information for the final control process. Besides, the model can also predict the remaining useful life of the tool, which unifies the problem of diagnostics and prognostics in one model, and simplifies the modelling process.Compared with the model based on Support Vector Regression, the model based HMM is more accurate and stable. The reason for this is the proposed method models the whole wear process while the SVR is a model based on sample matching, and do not take the wear process in consideration.Compared with the model using force, vibration and acoustic emission signals, the model only using vibration and acoustic emission signals has a similar accuracy. As the vibration and acoustic emission are easy to collect and needn't change the structure of the machine tool, the proposed method has a better application prospect.