Research On Vibration Signal Process In Online Dynamic Balancing Based On Second Generation Wavelet Transform

The test system of CNC machine tool spindle requires dynamic unbalanced measure which is extracted from collected vibration signal to compensate the unbalanced spindle in the quality aspect. Dynamic unbalance vibration signal collected by vibration sensors was often subjected to severe noise pollution. In particular, when the imbalance is too small, the signal to noise ratio of vibration signal is poor which can lead to a reduction of balanced test accuracy and stability. Therefore, noise elimination process for the weak periodic vibration signal is an effective means of extracting characteristic signal from the vibration noise in the strong noise interference.This thesis analyzes the characteristics of the vibration signal caused by the unbalanced rotor and constructs a mathematical model of the vibration signal.It studies the problem of frequency aliasing of the second generation wavelet transform and the second generation wavelet packet, and also introduces the constructor of the adaptive redundant second generation wavelet without aliasing in detail. According to the discontinuous or constant bias of wavelet coefficient problems existing in the constructor resulted by traditional methods of threshold, the thesis introduces a function of improved bi-variable threshold and propose an improved function of eliminating the vibration signal noise based on the second generation wavelets of adaptive redundant. Combined with the inter-correlation, it eliminates the noise of the simulated signal to extract the signal of dynamic unbalanced which verifies the improvement of the de-noising effect brought by improving the threshold function improved by bi-variable.According to the aliasing problem existing in the second generation wavelet packet, the thesis proposes an improved algorithm of the second generation wavelet packet:resets the frequency component whose approximate coefficients does not satisfy the sampling theorem in decomposing the wavelet packet and the frequency component of adjacent sub-bands mixed in detail coefficients. At the same time, it shifts frequency component remaining in the detail coefficients, and restores the detail coefficients in the restructuring process of the frequency shifted of wavelet packet. The thesis verifies that the algorithm of the improved second generation wavelet packet can retain the characteristics of the original signal well and remove the noise effectively. It helps algorithms of prony to extract frequency and phase of dynamic unbalance signal more accurately.