Date of Award


Document Type

Campus Access Dissertation


Mechanical Engineering

First Advisor

Abdel-Moez E Bayoumi


In 2005, the University of South Carolina (USC) was given the unique opportunity of studying a 17 component Apache (AH-64) tail rotor drive train (TRDT) system's vibrations for the purpose of improving the Army's Condition Based Maintenance (CBM) program, a program established to monitor aircraft components via accelerometers in order to detect and diagnose failures. Since then, USC has constructed three test stands and operated a testing facility in order to acquire aircraft component data for the purposes of improving vibration monitoring algorithms. The original intent of this research was to monitor seeded fault bearings and differentiate those vibration signatures from the baseline and non-seed faulted bearings. The analysis was broken up into two sections ; baseline testing and seeded fault testing. It was quickly realized that studying vibration signatures for bearings surrounded by so many non-master components was going to be a daunting task due to such a largely cross-coupled system. By not utilizing "Golden Suite" parts for components not under test, the frequency data for the seeded fault tests were clouded by surrounding components which would later be seen as non-intentionally inserted fault components. In addition to noise from neighboring components, vibration signatures can be affected by loading.

The intent of this research is to identify a methodology for locating nontraditional frequencies of deteriorating rotating components, and their locations as related to AH-64 tail rotor components as mounted to the USC TRDT test stand. For the selected frequencies, this research also illustrates the importance of load monitoring, the influences of shaft misalignment, shaft's imbalances and their effects on vibration signatures, as well as the importance of flight regime recognition. The plan for this research is to devise a way of extracting useful power train information from the test data in order to reduce the amount of "false alarms" while increasing the ability to detect faults for the last three stages of bearing fault progression.

The desired outcomes are to reduce false triggers and misdiagnosis and reduce the amount of data monitored by eliminating frequencies that provide no insight into the state of a component's health. This would also reduce post processing time and storage space required to handle large amounts of data by focusing on frequencies that yield the most information in reference to a bearings state. Another desired outcome is to incorporate the affects of loading into the bearing life plot in order to normalize the acquired "in flight" data. Even with these challenges it would be desirable to provide information that improves the ability to create more robust CIs, possibly modify existing bearing life algorithms, and assist in creating a more successful condition based maintenance program.