Research On Tool Wear Recognition Based On Cutting Sound Signals

The precise monitoring of the tool wear condition is the key to guarantee the metal cutting procedure. There is always the researchers'target that how to develop an exact, reliable and low cost tool wear monitor system. Compared with the other tool wear monitor technologies, the cutting sound signal analysis technology is more suitable to the tool wear monitor because it has many merits such as high sensitiveness, respond quickly and so on. Moreover compared with the acoustic emission signal, the cutting sound signal has lower frequency and the low cost acquisition devices as well as the set up location can be adjustable. But there are also many problems should be resolved, for example the environmental and machine tool disturbances. According to those reasons, this paper hopes to research how to avoid the defects and utilize cutting sound signal's merits and eventually establish a simple, reliable and low cost tool flank wear monitor and recognize system.Firstly, we established an experimental monitor system to collect the cutting sound signals in different tool flank wear conditions. Through time statistics and frequency power spectrum, the characteristic frequency between 2kHz～3kHz could be found and the influences of different cutting speeds and feed rates to tool flank wear could be also analyzed. Using wavelet methods to decompose cutting sound signal into eight wavelet frequency zones, then extracted every wavelet frequency zone's energy distribution proportion value to as characteristic vectors which can reflect tool wear. Using these characteristic vectors to train the neural network and further optimize it. Finally, programmed a tool flank wear identify software with LABVIEW. The experimental results indicated that the cutting sound signal and the tool flank wear had a well pertinence and tool flank wear identify software could effectively distinguish tool wear.