Integrated Software Platform for Fleet Data Analysis, Enhanced Diagnostics, and Safe Transition to Prognostics for Helicopter Component CBM

Abstract


 
 
 Although typical Health and Usage Monitoring Systems (HUMS) intend to support a transition from scheduled part replacements to performing maintenance upon evidence of need, they generally exhibit a limited ability to diagnose component faults early and accurately in complex systems such as a helicopter drive train. Consequently, the traditional approach to implementing Condition Based Maintenance (CBM) programs is slow, requires substantial amounts of human supervision (including case-by-case data analysis and results verification), and ultimately shuns prognostic activities. Causes of these limitations, which ultimately lead to an underrepresentation of prognostics in fielded CBM systems, include: (i) the sensitivity of sensors and condition indicators to signal noise and operating modes; (ii) use of empirical condition indicators not fully understood at the fleet-wide level; (iii) uncertainty in damage progression tracking; (iv) the inherent risk of condition prognosis; and (v) the lack diagnostic and prognostic validation with known fault cases.
 To improve the performance of CBM systems and facilitate transition from scheduled maintenance to reliable implementation of diagnostics and prognostics, a team of developers from Impact Technologies, the U.S. Army Research Laboratory and the Georgia Institute of Technology, with support from the U.S. Army have been working over the past 21⁄2 years to develop a methodology that is capable of addressing the challenges listed. This work has been a part of the Air Vehicle Diagnostics and Prognostics Improvement
 
 
 
 Program (A VDPIP), a collaborative agreement to develop, test and evaluate modular software components that provide enhancements to diagnostic systems already in service, as well as add failure prognosis capabilities for critical Army aircraft components. This paper presents the integrated diagnostic enhancement and prognostic architecture, as well as the software suite developed under the collaborative program, and discusses how a hybrid and systematic approach to sensing, data processing, fault feature extraction, fault diagnosis, and parallel health- based and usage-based failure prognosis can be used to improve the performance of a wide variety of HUMS and CBM activities in support of implementing prognostics. The software architecture contains generic components and algorithms building on model based and data driven methodologies that are applicable to a variety of critical components in complex systems such as those found in a helicopter drive train