This paper presents a knowledge-based system, called ASTRA, for state assessment, preventive diagnosis, and intervention planning of power transformers. The main goal of ASTRA is to support novices or non-expert operators to make the first decisions about the state of a transformer and to single out cases in which expert consultation is needed. ASTRA can also provide useful support to experts, through effective organization and presentation of available information and focused reasoning about it. ASTRA features a novel distributed architecture, where independent specialists cooperate to reach the final solution by interchanging partial results. Each specialist has autonomous reasoning capabilities and is able to perform specific tasks. Uncertainty affecting both data and knowledge is explicitly represented and propagated during the reasoning process, from data collection to the presentation of the final results to the user. Approximate reasoning is based on a new powerful method based on fuzzy numbers and Driankov's uncertainty calculus. The conclusions reached by ASTRA are presented to the user in an intuitive graphical form and are supported, on request, by explanations. Suggestions about appropriate interventions along with indications about their urgency are provided as well.
Return to publication list