Semi-Automated Knowledge Capture and Representation for the Development of Knowledge Based Systems

Abstract

Plant fault detection and diagnosis is an increasing operational necessity, especially in the nuclear sector where safety is of the utmost importance. Currently, operators have to manually inspect data acquired across multiple assets using predefined diagnostic processes, placing a high time burden on the analyst. Data-driven approaches to solving this problem can produce accurate results approaching what the analysts can achieve but in a fraction of the time. However, the majority of these techniques are black box in nature and therefore lack the explicability, often required for critical assets in the nuclear industry. Knowledge-based systems can be used for a variety of applications to provide not only accurate decisions but also the explanation and reasoning behind these decisions. However, the knowledge elicitation process places a significant time cost associated with the development of knowledge-based systems.In this paper, an approach is proposed for the development of knowledge-based systems that allow for accurate knowledge capture and formalisation that forgo formal knowledge elicitation sessions. By firstly producing a symbolic representation of the time-series data, abstracting similar trends to produce a list of potential rules, it was found that there was a significant time saving using this approach without equivalent loss of accuracy. A knowledge-based system developed in this way would allow for accurate and transparent fault diagnosis in any discipline, without placing a huge time burden on domain experts.