Parameterisation of domain knowledge for rapid and iterative prototyping of knowledge-based systems

In critical infrastructure applications, timely and consistent fault detection diagnosis is an increasingly important operational process, especially in the energy sector where safety of utmost importance. To realise this, engineers have to manually analyse data acquired from several assets using predefined diagnostic processes, but this a time-consuming process requiring significant amounts specialist expert knowledge. Data-driven approaches support diagnosis, other similar problems, can produce accurate results comparable what achieve fraction time. However, majority these data-driven techniques are black box lack explainability which often necessary for explaining decisions about power generation industry. Knowledge-based systems, such as rule-based systems been shown provide not only also explanation reasoning behind some related applications. there time cost associated with development knowledge-based particular knowledge elicitation domain expert’s formalised encoded into system. This challenge commonly referred bottleneck. paper, we present novel approach performing set symbolic primitives (rise, fall, fluctuate, stable) parameterise typical time-series condition monitoring data. The represented by common language that easily be communicated (and from) experts. allows quick experts knowledge, formalisation implementation rapidly produced Further due parametrisation it possible iteratively improve base updating parameters based on new unseen was applied Tennessee Eastman dataset, simulated real-world industrial process. It found accurately quickly capture required detect faults within case study provided fully explained reasons why each detected relating explanations previously defined.