| dc.description.abstract | There is a long-standing interest in the study of condition-based maintenance (CBM) integrated with continuous monitoring schemes accompanied by enormous initial investments. The cost for continuous monitoring might be prohibitive. Many industries are still using inspection in equipment maintenance to find possible defects and to trigger repairs. The defects caused by deterioration can be revealed at inspection dates and recorded by a specific format of condition indicators that can be grouped. That is considered as discrete states through classifying the condition indicator. Real empirical data of hydrant pumps at fueling facilities in the airport is used to identify the maintenance properties. Data processing is requisitely needed to extract meaningful information for modelling. Two frameworks are proposed in parallel: One is that the condition indicator is established in conjunction with repair activity itself if the reliable measurement for typically indicating a pump status like vibration levels is not obtainable. The other is how the hidden defects (a near failure) are considered in the gradual deterioration process by introducing virtual nodes when sudden degradation is observed, such as double jump without visiting intermediate state between leaving and arriving state.
The data is used to estimate the parameters relevant to the CBM model to decipher the system’s behaviour well. The counting process is adopted to verify the trend of degradation, and parameters of the distribution are obtained by fitting the observed duration of each state.
A Markov process has been ubiquitous in the field of CBM modelling and reliability analysis for discrete state space. Markovian based model is the best option when the dataset of system condition is discretized by state. There is no doubt that the Markov process is a powerful tool to illustrate the degradation progress by a stochastic model, whereas the Markov process assumes the exponential distribution of sojourn time in the condition states, which results in a memoryless property that is equipment has a constant probability of failing irrespective of both the current usage and deterioration level. A semi Markov process (SMP) in this paper is more flexible, allowing a Weibull distributed sojourn times not possessing memoryless property, but exponentially distributed sojourn times is investigated to provide an unbiased comparison. Time homogeneous domain that relies on the time spent since the last transition is appropriate for a repairable system whose components are replaced whenever a failure occurs. Primarily, Monte Carlo simulation is generated for CBM modelling to determine the state probabilities and the amount of maintenance. Then, an analytical approach is also proposed using a two-stage method composed of developing an embedded Markov chain (EMC) and then calculating transition probabilities using EMC and sojourn time of SMP. Meanwhile, the multiphase Markov process is suggested to reflect the periodic inspection on CBM modelling. Therefore, the CBM model for pumps takes into account both SMP and multiphase Markov process. The developed CBM model is supported by the relevant calculation and simulation results. | |