This paper deals with the diagnosis of industrial equipments through the use of Expert Systems. Intending to develop procedures that result in diagnosis knowledge bases for Industrial Maintenance, we have considered 2nd Generation Expert Systems. We have proposed a modified model and a diagnosis procedure. We used for the diagnosis strategy a "top-down best-first search", that combines two types of uncertainty treatment: (i) entropy, to find the best way in the search throughout knowledge structures, (ii) belief in the symptoms, to validate the resultant diagnostics. This proposal has the following advantages: a more complete knowledge base, a better explanation and presentation of the resultant diagnostics. We have developed a prototype considering real informations about centrifugal pumps.
expert systems; industrial maintenance; technical diagnosis
