RELIABILITY AND AVAILABILITY MODELING OF MECHANICAL SYSTEMS USING STOCHASTIC PETRI NET

Reliability and availability modeling of complex mechanical systems such as power plants in a single model, considering binary/tertiary states of individual components, using Markov approach is difficult due to state space explosion problem. Inclusion of various types of dependencies in the model further aggravates it. Moreover, the Markov approach is applicable when failure and repair time distributions are exponential. As the time to failure of mechanical components in most of the cases follows Weibull distribution, the Markov approach cannot be realistically applied for such systems. To overcome this, Stochastic Petri net (SPN) hierarchical modeling is proposed in this paper. The model is based on three aspects of the system: hierarchical level, basic structure and dependency. The three hierarchical levels are considered in the model. Individual component (level '3') SPN model is developed assuming Weibull failure distribution, while the individual subsystem (level '2') SPN model is developed considering the arrangement of components within the subsystem. The individual model at level '2' is reduced to an equivalent single net model and its equivalent transition rate is derived from its basic structure assuming the independence of components. The assumption of independence is relaxed for the system model ( level '1') and the dependencies (e.g. repair, standby redundancy, load sharing) are incorporated. The repair distribution in the system model is assumed exponential and repairs are considered at the subsystem level. Reachable markings are generated for the system model to obtain the reduced state space model. The proposed methodology provides realistic assessment of reliability and availability values. The suggested methodology is demonstrated for the 'Cooling Tower System'.