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

Romano Patrick, Matthew J. Smith, Carl S. Byington, George J. Vachtsevanos, Kwok Tom, and Canh Ly
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_10_005.pdf1015.87 KBAugust 30, 2010 - 9:29am

Although typical Health and Usage Monitoring Systems (HUMS) have the potential to support the goal of transitioning from time based part replacement decisions to performing maintenance upon evidence of need, their ability to diagnose component faults in their early stages in vehicle drive systems is limited, and, consequently, the task of prognosis in Condition Based Maintenance (CBM) programs is often underrepresented in the field. One of the primary causes of these limitations is the sensitivity of diagnostic processes to signal noise, specific component fault modes, and variations in environmental and operating conditions (this latter including loads, speeds, etc.). Another constraint on detecting faults earlier is imposed by the performance of the condition indicators used by diagnostic algorithms, because the indicators are in many instances chosen empirically, and their behavior in the presence of many variables and parameters is not fully understood or explained. A representative example is given by a fan support bearing of the oil cooling subsystem of the H-60 series of helicopters currently in service for the U.S. Army. The transition from time-based to condition-based maintenance for this bearing is relevant due to the criticality of the component and the relatively high incidence of replacements and faults. There is interest in eliminating the use of “time before overhaul” (TBO) definitions, which currently drive maintenance and retirement schedules for this component. A prerequisite to achieve this elimination is thus that diagnostic operations perform robustly even in the presence of the multiple kinds of disturbances affecting data acquired by HUMS sensors. This paper presents an integrated software architecture capable of improving fault detection methods for health monitoring of a damaged helicopter transmission component, including (1) sensing and data processing tools to extract useful fault information from raw sensor data, (2) selection and extraction of condition indicators/features, (3) fusion of data at the feature level, (4) tools for performing statistical analysis of fleet data and consideration of ground truth data, (5) consideration of usage or loading profiles to assess wear and/or damage progression rates in components, (6) a methodology to detect as early as possible with specified degree of confidence damage in a helicopter drive train component, and (7) techniques to predict damage or wear progression in the component regardless of whether a fault has been detected, using a Bayesian estimation framework.

Publication Control Number: 
005
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed