1996

Pakanen, Jouko

On-line diagnostic testing is one choice, when practical and robust fault detection and isolation methods are considered for automated processes. Performing a test means exciting a process by means of prescribed input signals, supervising responses and comparing results with a process model. An on-line diagnostic test is repeated similarly every time, in similar process conditions, making modeling an uncomplicated task. Fault detection is a direct consequence of the comparison, but fault isolation is based on elementary constraints, decomposed from the process model. A rough description of a fault can be achieved by heuristic reasoning, which enables application of the method in practice. A more specified fault description is accomplished by learning from old solutions. The reasoner accumulates information making decisions of the classifier gradually more precise through acquired experience. The method is best for successive installations, in which knowledge can be cumulated. On-line diagnostic tests are generic in character, but in this paper they are configured for an air handling unit of an office building and applied in its preheating subprocess. The paper presents the development, simulation and field tests of the fault detection and isolation method and its configuration as a part of a diagnostic system.