LAUSR

Abstract Fault detection and identification methods have received extensive study in the literature, but very rarely demonstrate the economic impact of these interventions. In this paper, a simulation of a two-tank system with economic performance functions is presented as a test-bed for process monitoring methods. The simulation includes sensor noise and process variation, control loops, and safety interlocks, as well as sensor failures. Simple process monitoring, by principal component analysis and contribution methods, is applied to single- and successive-sensor failure simulations, and the economic benefit of the proposed condition-based maintenance approach demonstrated, even for cases in which the faulty sensor was not correctly identified. Challenges with fault identification, root cause analysis and fault prognosis are discussed. Proposals are made for further development of the simulation, which has been made available for future research.