Prognostics and health management of safety-instrumented systems: approaches of degradation modeling and decision-making

Modern industries are developing towards a high-integrated direction with overwhelming complexities bringing benefits and potential risks with catastrophic consequences simultaneously. To reduce the occurrences of undesired events or mitigate their consequences, safety-instrumented systems (SISs), as a type of technical safety barrier, have been widely installed in different applications with the aim to protect people, the environment, and other valued assets. Examples of SISs can be emergency shutdown systems in oil & gas production, airbags in cars, fire sprinkler systems in buildings, etc. Many SISs operate in a demanded mode, meaning that they are only activated to perform safety functions while the unexpected occurs. For such systems mainly dormant in normal operation, it is important to conduct proof tests for checking system states and following-up maintenance in case of failures, to keep SISs highly available so as to ensure safety. In current studies, these activities are assumed following a predefined scheme with fixed intervals, independent from the actual system state. However, when more SIS state information can be collected by sensors and in manual tests, the prognostics and health management (PHM) strategy is expected to be more reasonable and cost-efficient. This PhD project thus aims to explore a new approach to evolve the SISs management from time-based to performance-based taking the technological advancement in data collection. This primary objective is then divided into five sub-objectives from the modeling approach and decision-making aspects that are addressed in the form of four journal articles and two conference papers. This PhD thesis bridges SISs performance assessment and degradation process through addressing different influence factors in the operational phase, including aging, and impact of demands, etc, for the decision-making in PHM by proposing: 1.A stochastic process-based degradation model with a specific threshold to describe the time-dependent system performance deviations with the target performance requirement. This model releases the as-good-as-new assumption even though the system is verified as being functional in tests. The proposed stochastic process-based degradation model provides an advantage of calculating the conditional system performance based on the collected information in tests. 2.An approach to quantify the side-effect of operational history on system degradation by introducing abrupt Gamma-distributed increments following a homogeneous Poisson process with arrival rate λde. Impacts of random demands are thus considered in performance evaluation. 3.A maintenance strategy with multiple follow-up actions to adapt the manifested system state in tests. The role of preventive maintenance on SIS management is emphasized in the operational phase. Relying on effective collected information contributes, such a strategy helps to keep an SIS at the required safety level while reducing the frequency of corrective maintenance. 4.A new decision-making support tool on updating testing and maintenance activities with coordinating the system unavailability and life cycle cost. The conditional system unavailability in the required safety integrity level will be the priority principle for updating test intervals, accompanying lower estimation intervention cost in the life cycle. The practical utility of the thesis resides in the provision of a comprehensive consideration of the time- and event-dependencies of SIS performance, as well as safety and economic meanings of testing and maintenance activities. In particular, the first is to provide hints of system deterioration and relevant health management to reliability analysts when they evaluate SIS design. The second is for operational managers of SISs as the decision-makers, to help them to update testing and maintenance plans and identify the optimal intervention opportunities. To conclude, this thesis will contribute to the implementation of PHM on SISs and other systems with similar operational characteristics. The research results on degradation assessment and predictive maintenance optimization can be generalized to more applications where production and maintenance need to be in synergy in consideration of safety and economics. Further research is, however, necessary for testing and validating the proposed methods with practical cases