Research On Drill Wear Condition Monitoring Technology Based On The Drilling Force Signal

With the development of automatization manufacturing equipment such as MC and FMS etc, tool condition monitoring becomes a key technology of machining process automatization. So it is very important to study and develop condition monitoring technique of tool wear and breakage. The dissertation proposes a new view of using tool cutting capability as monitoring index through analyzing the characteristic of drilling process and drill wear type, establishes drill wear condition monitoring experimental system with the drilling force as the monitoring signal, and systematically researches some key technology such as remove noise and feature extraction of the drilling force signal and tool condition recognition etc about tool condition monitoring.The dissertation introduces multi-resolution performances of wavelet transform, discusses some signal processing methods based on continuous wavelet transform, orthogonal wavelet transform and wavelet packet transform, and this lay a solid foundation for applying wavelet transform to drill tool condition monitoring.Aim at unsteady feature of sensor signal and noise interference causing difficulty to feature extraction and state recognition, this dissertation proposes denoising method of drilling force signal based on module maximum value of wavelet transform. The results show that SNR increases greatly after signal denoising.The dissertation transforms drilling force signal into the coefficients matrix with abundant condition information and denotes it using gray picture, and introduces gray moment to describe information of gray picture in order to extract gray feature. The method provides a new approach to feature extraction of the drilling force signal.The dissertation discusses principle of wavelet fractal theory, puts forward using fractal dimension to describe essence characteristic of drilling force signal, gives the algorithm of wavelet decomposition box dimension, and applies it to feature extraction ofthe drilling force signal. The experiment results show that box dimension of wavelet decompose reconstruction signals can describe effecting rules of drill wear to drilling system, and can realize dimensionless feature extraction of the drilling force signal and drill wear condition monitoring with box dimension.The dissertation obtains wavelet packet energy spectrum based on wavelet packets energy monitoring theory and realizes unitary feature extraction of wavelet packet energy. The experiment results show that unitary characteristic vector of twist signal can realize drill wear monitoring effectively.The dissertation establishes RBF Neural Network model used in drill wear condition recognition, and realizes mapping from characteristic vector to the drill wear condition. Moreover using this method can successfully realize the rec ognition of three typical drill wear status.