Rotor-bearing system identification using time domain methods

Several existing and new time domain methods have been compared and proposed for rotor bearing system off-line and online modal parameter and physical parameter identification. This dissertation concentrates on modal and physical parameter identification with incomplete input measurement, single input and single output measurement, or output measurement only at limited locations. This kind of measurement environment is typical for current rotating machinery. The transient analysis is thus important for the identification and it has been shown that increased input frequency change will induce large transient signals which can be used for system identification with output measurement only.; Existing modal identification methods are reviewed. Among them, the Backward ARMA method is found to be most suitable for modal parameter identification with single input and single output measurement. A new method, Extended Backward ARMA, is derived to identify system zeros and automatically detect the number of system zeros. The existing modal identification methods for output measurement only case, i.e. Random Decrement and Yule-Walker Equation, need to be modified to handle the rotating machinery online measurement condition. Three new identification methods, i.e. Direct AR Modeling, Improved Random Impulse Injection, and Improved Yule Walker Equation, have been proposed for identification with the rotating machinery online measurement condition and with output measurement only. The Direct AR Modeling has been shown to be very effective for the modal parameters identification under certain circumstances. Improved Random Impulse Injection and Improved Yule Walker Equation are shown to be able to reduce the required averaging number, to have higher frequency resolution, and to be able to identify well-damped modes.; The physical parameter identification methods with complete input and output information are reviewed. Two new methods, i.e. Modified Newton-Raphson method and Transformed Eigenvalue method, have been proposed for Inverse Eigenvalue analysis. Modified Newton Raphson method takes into account the unique nature of rotor bearing system. Transformed Eigenvalue method is proven to be more computationally efficient, have larger convergence range. These new methods not only apply to physical parameter identification with complete input and output information but also apply to physical parameter identification with output measurement only.; The combination of above mentioned identification methods for modal parameters and physical parameters with or without the complete knowledge of input signals is proposed for online and off-line rotor bearing system identification.