Machining process/tool wear monitoring system based on real-time sound recognition

A machining process/tool wear monitoring system based on real-time sound recognition and neural network algorithm was developed. The system is composed of a high-speed digital filter bank unit, preliminary test/learning module, database module, and real-time monitoring module. The filter bank is constructed using DSP's and calculates sound patterns through digital filtering program in real-time. Abnormality in overall process is monitored by comparing a current sound pattern with the standard pattern, which is calculated and registered in the database for each NC block through test machining prior to production machining. Since the overall change of the sound pattern due to tool wear is too small to be detected by simple comparison of pattern, the way to monitor tool wear more reliably by using the rate of change of pattern components and neural network algorithm is proposed. Evaluation tests were performed for the NC lathe operation in the real shop floor. Environmental noise was reduced remarkably by closing the protection door of machine. Monitoring results showed usefulness of the developed system. Especially tool wear monitoring could judge the time of tool change as exactly as an expert did.