Technical Condition Indexes and Remaining Useful Life of Aggregated Systems

Abstract

 It is a challenge to determine the remaining useful life of items having several failure modes that are dependent on age and several condition monitoring sources. Remaining useful life has always been a wanted quantity in maintenance planning. If the exact time before a failure occurs were known, maintenance planning would be simpler.

The main objective of the thesis is to provide quantitative estimates of remaining useful life including error bounds. Based on information from process control systems, condition monitoring systems, and inspection reports, the technical condition of the items is quantified by technical condition indexes (TCIs). The TCIs are normalized quantities that can be aggregated upwards an established hierarchy to the desired level. Both an expert judgement principle and a proportional hazards model approach are used to establish the aggregated TCIs. In the latter approach significant covariates test are used to identify significant covariates.

Problems with partial or incomplete data where segments of recorded required maintenance actions and TCIs are missing are solved by left-truncation in the statistical model. The aging effect which is quantified by the Weibull proportional hazards model is used to estimate the remaining useful life. Realizing that the literature within the prognostic area tends to put more focus on age-independent models, the remaining useful life is also modelled as a degradationthreshold model. Here, a gamma process is used to model the deterioration trend of an aggregated TCI. The results from the two approaches are compared and discussed. A discussion of uncertainty and predictability is also performed. This information is important in decision making for further maintenance.

The models have been applied to a case study of a natural gas export compressor system. In the case study a detailed description of failure modes, condition monitoring, and inspection data is given. The case specific hierarchical aggregation of the technical condition index to compressor-level is also presented. Finally, the thesis make a summary of the work performed, and proposes further research related to the next generation of prognostic systems.