| Boring operations is quite subject to chatter, especially when boring deep holes.Chatter leads to poor surface quality, shorter cutting tool life, and limited machiningcapacity. Most of the researches on chatter detection have adopted sensor signals directlyrelative to machining operations, such as acceleration, cutting force and audible sound,which are not suitable for promoted applications. So an approach for chatter detectionbased on spindle motor current is developed. The main research work is as follows:According to the demand of engineering application, the feasibility of chatter detectionusing motor current is analyzed and the problems that may exist are considered. Theoverall program is developed according to the functional requirements for chatterdetection, and the corresponding experimental system and program are developed;To acquire the monitoring signal sensitive to chatter, a method to acquire sectionsunder the cutting condition is studied, and spindle motor current and feed motor currentare compared in both time domain and frequency domain. As a result, spindle motorcurrent is selected for chatter detection in boring operations, and the denoising methodbased on wavelet packet decomposition is adopted to preprocess the signal;To get the optimal feature set for pattern recognition, both time domain and frequencydomain features are extracted according to the change law with the development of chatter,and the method based on distance evaluation technique and correlation analysis is adoptedto select features. Finally, the optimal feature set includes mean, maximum, standarddeviation, one-step autocorrelation function and frequency center;To detect the symptoms of chatter in time, probabilistic neural network(PNN) isadopted for pattern recognition and is trained by an active learning method. Theexperimental verification results show that the trained PNN gets a classification accuracyrate above95%. So the approach can meet the demand of the chatter detection application. |
PDF Full Text Request