Use this URL to cite or link to this record in EThOS:
Title: Bayesian approaches for complex system prognostics
Author: Bin Zaidan, Martha Arbayani
ISNI:       0000 0004 5348 7617
Awarding Body: University of Sheffield
Current Institution: University of Sheffield
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Access from Institution:
Condition-based maintenance is an emerging paradigm of modern health monitoring, where maintenance operations are based upon diagnostics and prognostics. Prognostics promises to optimise maintenance scheduling, resources and supply chain management, leading to reductions in operational disruption, spares inventory, maintenance labour cost and hazardous conditions. The main objective of this research is to develop generic datadriven prognostic approaches to address several challenges associated with complex system prognostics, where in this particular work, the developed techniques are applied to the degradation data obtained from civil aerospace gas turbine engines. This thesis contains four key contributions. Firstly, deterministic Bayesian prognostics is used to deal with large uncertainty in degradation data. The novelty and value in the presented formulation lies in a fuller Bayesian treatment of observation error than prior art while retaining the closed-form solution desirable for real-time, deterministic computation. Secondly, the Bayesian hierarchical model (BHM) is introduced to optimise the use of fleet data from multiple assets. This formulation allows Bayesian updates of an individual predictive model to be made, based upon data received from a fleet of assets with different in-service lives. The results obtained demonstrate BHM capability in dealing with some extreme scenarios, occurring in complex system prognostics. The next contribution lies in developing variational inference for the existing BHM to overcome the computational and convergence concerns that are raised by sampling methods needed for the inference of the original formulation. This technique delivers an approximate but deterministic solution, where the quality of approximation is found to be satisfactory with respect to prediction performance, computational speed and ease of use. In the final contribution, an integration concept is proposed, combining the Bayesian data modelling technique with an information theoretic change-point detection algorithm to solve a wide class of prognostic problems, such as information arising from irregular events occurring during the life-cycle of an asset. This integration concept has a great potential to be implemented in complex system prognostics as it demonstrates several advantages of the deterministic BHM in combination with change-point detection to utilise, optimally, all available multiple unit data as well as data available at various levels of the system hierarchy.
Supervisor: Harrison, Robert F. ; Fleming, Peter J. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available