A Failure Diagnosis System for the Burner of Absorption Chillers.

Sensors utilized for diagnosing the flame current of a flame-detector, for instance, are limited because the burner of the absorption chiller is small or medium in size. Subtle differences in the conditions can affect the readings of these sensors and make them unstable. The knowledge of burner daignosis based on such unstable sensors has several features. 1. The structure of the knowledge is relatively simple. 2. The knowledge is probabilistic because of the unstable readings and the non-deterministic nature of the knowledge possessed by engineers in the domain 3. The knowledge has a low quantitative accuracy. In this paper, we propose a system that employs simple binary logic to determine failure modes, and probalistic multi-valued logic to evaluate the probabilities of failure causes. This system achieves suitable expressions for knowledge of domain-engineers and features a compact inference engine that can be implemented into small-size computers such as a notebook-PCs.