Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.613530
Title: Improving planned and condition-based maintenance decision support
Author: Lim Chi Keong, Reuben
ISNI:       0000 0004 5365 6548
Awarding Body: Cranfield University
Current Institution: Cranfield University
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
In both civil and military aviation, maintenance plays a large role in ensuring continued safe operation and accounts for a significant portion of operating costs. Typically, a conservative planned maintenance (PM) program is initially developed to ensure the aircraft reliability but this often leads to over-maintenance. With more in-service experience, operators seek to customize the maintenance interval accordingly in order to reduce workload and cost without compromising safety. With prevailing use of health usage monitoring systems (HUMS), the maintenance can even transit from PM to condition-based maintenance (CBM) where further safety and costs benefit may be reaped. Whilst some guidance for such changes exists, it remains challenging for maintainers in practice as suggested methods often require significant component failure or test data; which are unavailable or too expensive to obtain. As such, this research reviews the challenges faced by maintainers when extending PM intervals or implementing CBM and seeks ways to support decision making for the changes. For PM, the challenge to extend the maintenance interval with little or no past failure is addressed. Existing reliability methods were reviewed and two improved methods to estimate the reliability lower confidence bounds were developed. The first approach adopts the use of Monte Carlo simulation applied to the Weibull plot equation while the second uses a probabilistic damage accumulation model together with bootstrap techniques. Both methods are used to assess the reliability of extending the replacement interval of a gearbox bearing and are shown to perform better than existing methods as they provide tighter reliability confidence bounds. For CBM, a survey on sensor technologies and diagnostic algorithms showed that vibration-based sensor is most widely used to detect fault. The study then demonstrates a CBM implementation using vibration-based HUMS data from in-service helicopters. Analysis of the FFT spectra shows that the fault patterns corresponding to progressing stages of bearing wear can be clearly observed. The fault patterns are extracted as features for unsupervised classification using Gaussian Mixture Models and used to infer the different bearing health states. Signal detection theory was then applied onto the classified feature to determine the detection thresholds for fault diagnosis. A simplistic prognostic model using trend extrapolation to determine the replacement lead-time is then performed and use for maintenance planning. In an effort to ease the implementation of CBM, ways to improve prognostics application is explored. The Switching Kalman Filter (SKF) was adapted for both diagnostic and prognostic under an autonomous framework that requires little user input. The SKF uses multiple dynamical models with each one describing a different stage of bearing wear. The most probable wear process is then inferred from the extracted feature data using Bayesian estimation. As different stages of bearing wear can be tracked using the dynamical behavior of the measurements, pre-established threshold for fault detection is no longer required for diagnostics. The SKF approach provides maintainers with more information for decision-making as a probabilistic measure of the wear processes are available. It also offers the opportunity to predict RUL more accurately by distinguishing between the wear stages and performing prediction only when rapid and unstable wear is detected. The SKF approach is demonstrated using in-service feature data from the AH64D TRGB and the results have shown the proposed methods to be a promising tool for maintenance decision-making. As an extension of research on methodologies to improve PM and CBM decision support, a thioether mist lubrication is explored for its feasibility as a backup lubrication system for helicopters. The aim is to reduce the mishap severity category which in turn eases the extension of PM interval or its replacement with a CBM task. An experimental setup was developed to test the thermal properties of a spur gearbox with thioether mist lubrication under various load and speed conditions and it was shown that only a very small volumetric flow of lubricant is required to preserve the gears from damage in oil starved environment. As such, a thioether based mist backup system can potentially reduce the risk of oil starvation failures significantly.
Supervisor: Mba, David Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.613530  DOI: Not available
Share: