Lifetime profit modelling of ageing systems utilising information about technical condition

This dissertation focuses on a framework to support decision making in the management of ageing (oil and gas) facilities. Within the oil and gas sector on the Norwegian Continental Shelf the topic has become important as the installations are approaching the end of their intended lifetime. The objective of the thesis is to provide decision support for managing ageing systems and equipment with respect to the inspection, overhaul and replacement strategy in a life cycle perspective. The replacement strategy also include evaluation of obsolescence i.e. changes in external requirements and thus functional demands that call for a replacement. The thesis presents a model designed to investigate and seek optimal solutions when it is possible to classify the items present condition and predict future development based on previous condition monitoring results and future functional demands. The deterioration process is described by a Markov process, and the sequential decision problem is modelled as a discrete time Semi-Markov Decision Process. The transition probabilities of the controlled time-variant Markov process are described in a condition transition probability matrix. To account for the end-of-horizon effect and time dependent external parameters e.g. varying production profile, the optimal solution is found by use of the value iteration procedure (stochastic dynamic programming). The model has been applied to a case study including inspection and repair of an offshore gas turbine. Two main degradation processes have been studied and added to the model. In the case study, an investigation has been made into the effect on the optimal plan due to forthcoming known turnarounds (planned shutdowns for inspection, maintenance and modifications), and knowledge about new technology. The provision of methodologies that can support decision making for future maintenance and operation activities is challenging, but hopefully the thesis presents ideas that will improve understanding of the applicability of Markov Decision Processes in this matter