Research On Fault Diagnosis Of Rotating Machinery Rotor With Blade Tip-timing Data

The vibration of rotor system which serves as the core component of rotating machinery is extremely complicated because of the result of various fault types. The traditional methods for shaft vibration monitoring have been researched extensively on the areas of fault diagnosis such as unbalance, misalignment and rubbing, but are limited in the aspect of disk crack, blade failure and turbulence excitation vibration. At the same time, blade tip-timing technology has superiority in terms of blade and blisk fault features extraction due to the application of monitoring the operating parameters of rotating blades. In this dissertation, the research focuses on the fault diagnosis technology to integrate the information from blade tip-timing system and rotor shaft vibration monitoring system for the purpose of error reduction and accuracy improvement. History experience is combined to identify the malfunction parameters on the condition that the fault occurrence and its type have been known. The specific details are as follows:A method for the frequency identification from blade tip-timing data matrix is presented. The matrix is based on ESPRIT and obtained from two sensors which work as the double signal channels, and mounted with a specific angle. This paper is analyzed the influence on accuracy caused by fluctuating of revolving speed and SNR.The reconstruction method for blade tip-timing signal from equispaced sensors is studied. The reconstructive kernels consist of B-spline function which is analyzed in the performance and error reasons by simulation. Experiment results indicate that the reconstructive signal is helpful for the identification of the blades vibration amplitude which is one of the features for rotating assemblies fault diagnosis.The vibration responses of fault force are solved on the base of rotor unbalance and rubbing fault model which are built by transfer matrix and finite element method respectively. The influences on response features caused by relevant parameters are also analyzed. Based on these analyses as well as wavelet spatial filtering theory, it proposes a wavelet packet coefficients correlation filtering algorithm. According to the known fault type, this algorithm can be applied to specify fault feature extraction from strong background signal or concurrent faults. Combining with wavelet redistribution spectrum and Hilbert-Huang transform, this filtering algorithm achieves favorable result in rotor rubbing feature extraction from acceleration signal. The precision would be influenced by rotor shaft fault response during the blade working condition judgment. The wavelet packet coefficients correlation filtering algorithm is applied to solve this problem. At first the ESPRIT method is used to identify the frequency of blade tip-timing data, and then the energy of blisk unbalance response amplitude which is identified by a quantitative diagnosis method is performed as the threshold value. The experiment results indicate that this method could restrain the interference caused by the blisk unbalance in the blade vibration signals to some extent.The rubbing fault locating-analysis method is explored by experiment. The axial rubbing location of rotor is estimated according to the distributing variation of node vibration response, and so does the blade-case rubbing location based on the blade tip-timing sensor equispaced around the case. The feasibility of diagnosis method of the blisk rotor rubbing location quantitatively is verified according to the two information sources blade tip-timing and shaft vibration monitoring.