| The research presented here concentrates on experimental and analytical methods to identify the twelve rotordynamic coefficients of high-Reynolds-number hydrostatic bearings. Fluid-flow-induced noise (turbulence, cavitation) presents a significant problem in testing these bearings, since this noise exists in the same frequency range as the test signals. This common frequency range eliminates the possibility of rejecting the noise through filtering.; A frequency-domain analysis method has been developed to extract the rotordynamic coefficients from data obtained by multi-frequency sequential excitation testing. Power-spectral-density estimates are used to reject noise that is independent of the applied excitation. The method is verified experimentally by single-frequency excitation, and repeatability tests.; A second power-spectral-density based identification method has been developed and applied to simultaneous excitation test data. Here, two independent random excitations are applied simultaneously to the bearing, and their effects are separated by computing appropriate power-spectral-density estimates, while, at the same time, maintaining excellent noise rejection properties. This method yields results identical to the sequential excitation procedure, and reduces the actual test and analysis times by one half.; Last, an analysis method in the time domain has been developed and implemented. In essence it is a direct application of least squares to the bearing housing equations of motion. For this method, it is shown that for the low-frequency noise present in the measurements, differentiation of the signals is acceptable, and preferable over integration. The twelve rotordynamic coefficients are obtained from a recursive formulation of least squares, and the results are shown to have acceptable convergence. Also, the coefficients extracted from this method are found to reconstruct the bearing response more accurately than the coefficients extracted from the power-spectral-density methods. |
