Condition-based component replacement of the pneumatic valve with the unscented particle filter

This paper investigates the condition-based maintenance (CBM) that concerns the component replacement strategy based on the estimation of the failure probability distribution. To obtain the accurate estimation of the distribution, specifically in non-linear case, an improved model-based Monte Carlo method, the unscented particle filter (UPF), is introduced. With the estimation of the failure probability, the replacement is determined by minimizing a decision variable called the expected cost per unit time, which considers both the replacement upon failure and preventive replacement. Simulated experiments are performed with regards to a pneumatic valve, a normally-closed and gas-actuated valve, whose dynamic physical model is studied a lot in recent years. The experiment results illustrate that with the accurate prediction of the probability distribution of the component's remaining life, we can effectively realize the condition-based component replacement and risk-informed life-extension in many application domains, such as nuclear, aerospace and chemical ones.