A fundamental problem in the development and validation of PHM technologies is the general shortage of realistic fault signature data. While healthy signatures can be obtained from operational systems, faults are relatively rare and difficult to observe. The PHM community must rely primarily on bench level seeded fault test data collected under a limited set of conditions.
To augment physical data, a modeling and simulation toolset is being developed for the vibration signatures of faulted components in propulsion subsystems such as gearboxes. The toolset includes sophisticated dynamic models of vibration forcing for common rotating components such as bearings and gears based on detailed analysis of their physical interactions, including the effects of faults such as bearing spalling or gear tooth cracks. The response of the overall system, and thus the vibration signal seen at a particular sensor location can then be predicted using either FEA or a transfer function analysis of actual hardware.
The purpose of the toolset is to leverage seeded fault test data (for example to study operating conditions or fault types that were not tested), improve fault diagnosability through optimal sensor placement, and enhance development, testing, and validation of diagnostic systems. Several examples are presented comparing simulated vibration signals to actual test data.
Author: Nancy Lybeck, Brogan Morton, Andrew Hess, John Kelly
Publish Date: Mar 15, 2013