|dc.description.abstract||As the consequence of the more widely used complex systems, there has been a shift in maintenance strategy. The traditional corrective maintenance is gradually replaced by preventive maintenance or even more advanced philosophy such as condition based maintenance and prognostics and health management. This thesis introduces prognostics and health management that can implement the advanced condition-based maintenance for the more complex and dynamic systems. First, the content of prognostics and health management is discussed, and then the procedure is described as data acquisition, data processing, diagnostics and prognostics, and maintenance decision making. In addition, historical literatures about diagnostic and prognostic models are systematically and throughly reviewed. It consists of model-based models and data-driven models and the paper focuses more on data-driven models, due to its simplicity and generality. Since degradation is one of main causes for system failure either for either machinery or electronics, degradation status assessment and prognostics are discussed in this paper. Gamma process is suitable for monotonic degradation, but there is a prerequisite that the degradation indicator is observable. In order to overcome this limitation, ANNs are used to calculate the value of indicator by monitoring relevant measurable covariates. One hybrid model is proposed consists of ANNs model together with Gamma process for degradation prognostics. Bayesian estimation for updating the scale parameter of gamma process is suggested to improve the accuracy.
Chock valve is studied as a case to demonstrate how the hybrid model can be applied to estimate valve residual useful life. The case study approves the results from hybrid prognostic model, the distribution of RUL can support the maintenance decision making. This proposed hybrid model can not only be applied to subsea valves erosion prognostics, but also can be applied to other equipment degradation prognostics problem.||