A self-adaptive alarm method for tool condition monitoring based on Parzen window estimation

Tool condition monitoring (TCM) takes an important position in CNC manufacturing processes, especially in damages avoiding of working parts and CNC itself. This paper presents a self-adaptive alarm method using probability density functions estimated with the Parzen window based on current signals, which gives an adaptively and rapidly corresponding alarm when the cutting tool fracture occurs. A CNC with cutting tools was obtained by Guangzhou CNC Company for test purpose, and the relative experiments were done in the state key laboratory. Current signals of the spindle motor and the main feed motor were acquired during the tool life. A probability model estimated with the Parzen window is established for current data fusion to alarm adaptively. At the meantime, the acoustic emission (AE) signals were acquired for comparison purpose. Experimental results show that this technique is flexible and fast enough to be implemented in real time for online tool condition monitoring.