Rotating equipment defect detection using the algorithm of mode isolation

Findings from a project involving rotating equipment defect detection using the Algorithm of Mode Isolation (AMI) are presented. The prototypical system evaluated is a rotating shaft, supported by hydrodynamic bearings at both ends, with one disk mounted to the shaft. Shaft cracks and bearing wear are the two equipment defects considered. An existing model of the prototypical system from the literature is modified to simulate the presence of a transverse shaft crack at mid-span. Ritz series analysis, in conjunction with a previously published description of the compliance related to the presence of a transverse shaft crack, is used to describe the decrease in shaft stiffness associated with the crack. The directional frequency response function (dFRF) is shown in the literature to provide benefits over the standard frequency response function (FRF) in both system identification and shaft crack detection for rotating equipment. The existing version of AMI is modified to process dFRFs and termed Two-Sided AMI. The performance of Two-Sided AMI is verified through system identification work. The results confirm the benefits of using the dFRF for system identification of isotropic systems. AMI and Two-Sided AMI are experimental modal analysis (EMA) routines, which estimate modal properties based on a frequency domain expression of system response. Three defect detection studies are fully described. In the first, bearing wear detection methods are evaluated. The results show that damage metrics based on modal residues are sensitive to bearing wear. Next, an in-depth investigation of shaft crack detection is performed. The shaft crack results in a time-varying system. Therefore, this analysis is also used to evaluate performing EMA on non-modal data. The effects of noise and coordinate system choice on shaft crack detection are also investigated. Crack detection through EMA processing of noisy, non-modal data is found to be feasible. The eigenvalue-based damage metrics show promise. Finally, a dual-defect (shaft crack and worn bearing) study is described. The results suggest that AMI is usable for defect detection of rotating machinery in the presence of multiple system defects, even though the response data is not that of a time-invariant system.