Research On Maintenance Decision Modeling And Simulation Of Deteriorating System

SG-III is a large-scale Inertial Confinement Fusion (ICF) laser facility of our country in construction, which takes strong laser as the drive source in the laboratorial environment to implement heat nuclear fusion. As one of the major measurement indices of ICF laser facilities, maintainability is an important factor to ensue SG-III can run permanently, stably and reliably. Conducting research on maintainability of SG-III as soon as possible has important theoretical and practical significance on improvement of operating efficiency and reduction of running cost.At present, the research on the maintenance decision modeling and related analysis technologies focuses mainly on single-equipment system, which is difficult to solve issues on the maintenance decision of technology intensive complex systems such as SG-III. Aiming to the features of the energy component of SG-III, such as deterioration failure and unobserved states etc., and considering various maintenance policies and maintenance effect, this thesis adopts Semi-Markov Decision Process model, Stochastic Petri Net with Aging Token (SPNAT) and multi-model respectively to solve issues on maintenance decision modeling and analysis of energy storage capacitor, pre-ionization module and energy component. Main research of this thesis is as follows:First, aiming to the issue on maintenance decision of a single component whose deterioration obeying continuous Markov Process, considering actual maintenance effect and maximum times permitted of preventive maintenance, taking the minimum cost during long-term operation of components as objective, maintenance decision model and its algorithms based on Semi-Markov Decision Process, as well as the corresponding Monte Carlo simulation model are presented. The method is used to solve maintenance decision issues on energy storage capacitor of the energy component, which determine its optimal testing period and the threshold of preventive maintenance.Second, as the Semi-Markov Decision Process is difficult to describe issues on the maintenance decision of components under the policy of Non-Markov deterioration or complex maintenance policies, the thesis proposes a SPNAT maintenance decision modeling method, and presents a SPNAT maintenance decision model of components and systems including several maintenance policies, such as condition-based maintenance and corrective maintenance etc., which is used to solve the issue on maintenance decision of the pre-ionization model of the energy component under the policies of on-condition maintenance and corrective maintenance, to determine the optimal testing period and the opportunity maintenance threshold of the pre-ionization model.Third, aiming to solve issues on state space explosion maintenance decision modeling and related analysis technologies, and difficulties in description and solution of using single model in maintenance decision modeling of complex system, a method on maintenance modeling and simulation based on multi-model is proposed, which specify measures of performance based on reward model and the police of model combination based on Join/Replicate. Combining the fault-tree model of the energy component and the SPANT maintenance decision model of each component, a multi-model simulation tool, MISER is used to solve issues on maintenance decision of energy source component, to determine the optimal period of the energy component.At last, the paper summarizes its innovation and works, and conceives the further research in the future.