Athena Zitrou

Robustness of maintenance decisions: Uncertainty modelling and value of information

In this paper we show how sensitivity analysis for a maintenance optimisation problem can be undertaken by using the concept of expected value of perfect information (EVPI). This concept is important in a decision-theoretic context such as the maintenance problem, as it allows us to explore the effect of parameter uncertainty on the cost and the resulting recommendations. To reduce the computational effort required for the calculation of EVPIs, we have used Gaussian process (GP) emulators to approximate the cost rate model. Results from the analysis allow us to identify the most important parameters in terms of the benefit of ’learning’ by focussing on the partial expected value of perfect information for a parameter. The analysis determines the optimal solution and the expected related cost when the parameters are unknown and partially known. This type of analysis can be used to ensure that both maintenance calculations and resulting recommendations are sufficiently robust.

An Influence Diagram Extension of the Unified Partial Method for Common Cause Failures

Abstract Modelling Common Cause Failures (CCFs) is an essential part of risk analyses, especially for systems such as nuclear power plants, which are required to have high reliability. The Unified Partial Method (UPM) is the main approach of the UK for modelling CCFs. This paper presents an Influence Diagram model for CCFs which extends UPM and represents uncertainty on system performance. This allows more detailed modelling of CCFs in terms of root causes and coupling factors, creates a context for using information in the industry database, and captures the non-linearity in the way system defences influence reliability. A structured expert elicitation process is used to construct the Influence Diagram model and to identify the non-linear structure of the domain, using an example of Emergency Diesel Generators (EDGs) from nuclear power plants. Insights and experiences from the elicitation process are described.

Bayes geometric scaling model for common cause failure rates

This paper proposes a mathematical model to associate key operational, managerial and design characteristics of a system with the systems susceptibility towards common cause failure (CCF) events. The model, referred to as the geometric scaling (GS) model, is a mathematical form that allows us to investigate the effect of possible system modifications on risk. As such, the presented methodology results in a CCF model with a strong decision-making character. Based on a Bayesian framework, the GS model allows for the representation of epistemic uncertainty, the update of prior uncertainty in the light of operational data and the coherent use of observations coming from different systems. From a CCF perspective these are particularly useful model features, because CCF events are rare; hence, the operational data available is sparse and is characterised by considerable uncertainty, with databases typically containing events from nominally identical systems from different plants. The GS model also possesses an attractive modelling feature because it significantly decreases the amount of information elicited from experts required for quantification.

A model for availability growth with application to new generation offshore wind farms

A model for availability growth is developed to capture the effect of systemic risk prior to construction of a complex system. The model has been motivated by new generation offshore wind farms where investment decisions need to be taken before test and operational data are available. We develop a generic model to capture the systemic risks arising from innovation in evolutionary system designs. By modelling the impact of major and minor interventions to mitigate weaknesses and to improve the failure and restoration processes of subassemblies, we are able to measure the growth in availability performance of the system. We describe the choices made in modelling our particular industrial setting using an example for a typical UK Round III offshore wind farm. We obtain point estimates of the expected availability having populated the simulated model using appropriate judgemental and empirical data. We show the relative impact of modelling systemic risk on system availability performance in comparison with estimates obtained from typical system availability modelling assumptions used in offshore wind applications. While modelling growth in availability is necessary for meaningful decision support in developing complex systems such as offshore wind farms, we also discuss the relative value of explicitly articulating epistemic uncertainties.