Research On Fault Monitoring And Diagnosis Of Numerical Control Machine

Because numerical control machines have high precision and good economic performance, they have been applied more and more in production areas. To solve the problem that numerical control machine’s fault is hard to find and difficult to deal with, a kind of numerical control machine’s fault monitoring and diagnosis system is developed, which has characteristics of standardization, bus construction and networked, etc. Various sensors’ real-time information is collected, and fault is monitored. Energy characteristics’ method based on wavelet packet is researched, which is used for vibration and noise signal’s feature extraction. Fault diagnosis expert system software is developed for fault reasoning and diagnosis. The specific work are as follows:(1)Numerical control machine’s fault mechanism is known about by reading literature with research situation and development tendency of signal monitoring, signal processing and feature extraction and fault diagnosis.(2) The hardware circuit board is developed, and temperature gathering unit and embedded fault monitoring unit based on ARM9processor are developed. Application software is programed and used for embedded fault monitoring unit and temperature gathering unit.(3)The collected vibration and noise data are decomposed and signal fault features are extracted by wavelet packet energy method. Energy eigenvalue is used as neural network’s input for fault recognition; A momentum and learning factor are added in basic BP neural network algorithm, weakening the neural network’s problem of slow convergence speed and easily stepping into the local convergence points. The recorded normal and abnormal noise is successfully identified in the experiment by the method.(4)The wavelet packet algorithm and neural network algorithm are combined with positive and negative reasoning and fault tree modeling based on knowledge base in the expert system software. Through simulation platform, the interactive fault diagnosis of M axis crawling and online diagnosis based on sensor data are experimented successfully, verifying that method of fault tree modeling and reasoning combined with wavelet neural network algorithm is feasible for numerical control machine’s fault diagnosis.At the last part of the thesis, the main contents are summarized and suggestions of the future research work in this field are given.