Research on PCB Drilling Tool Wear Condition Monition Based on Vibration and Neural Network
|School||Shanghai Jiaotong University|
|Course||Mechanical Manufacturing and Automation|
|Keywords||PCB Micro-hole Drilling Tool wear Vibration WaveletTransform Artificial Neural Network|
Real-time monitoring of tool wear condition is the key technology ofadvanced manufacture system and is a very important link in machineprocessing, especially for high-density drilling machining, such as PCB(Printed Circuit Board). How to realize automatic monitoring of tool wearcondition is the significant factor for improvement of machine quality,productivity and automation of production, which has been considered asa very key technology and important significant problem that is not yetbeen solved. Based on the technology status of tool wear conditionmonitoring, a tool wear condition monitoring method in PCB micro-holedrilling is presented in this paper, the research as follows:This paper, firstly, establish the tool condition monitoring signalsampling system based on a self-built micro-cutting platform andaccelerometer sensor. After experiments, analyze the vibration signals ofdifferent tool wear conditions in time domain, frequency domain as well as wavelet transform. Based on wavelet transform, three kinds of toolwear features are abstracted, including energy, root-mean-square valueand kurtosis and the method is demonstrated through comparing threekinds of signal-normal, initial wear and severe wear. Then this paperestablishes tool wear condition monitoring system based on BP neuralnetwork. Through the test of BP neural network system, we found thedefects. To solve the problems of BP neural network, the method of fuzzyneural network is introduced to tool condition monitoring. The conditionsystem based on that successfully built the non-linear relationshipbetween tool wear conditions and signal features, and improves therecognition speed as well as system stability.The research on the experiment design, signal collection, signalanalysis, feature selection and extraction, pattern recognition has beenexplored in the thesis. The precision and stability of monitoring systemare improved.