The paper presents the rules induction algorithm designed to identify and locate faults in the analog systems. Compared to the most popular numerical methods, such as Artificial Neural Networks, this approach enables presenting knowledge behind the reasoning mechanism. This way it is possible to better understand relations between measured symptoms and particular faults. The method exploits the preliminaries of the AQ algorithm, which was proposed for the discrete data. The modification proposed in this paper covers the analysis of continuous features and adjusting the method to work in the uncertainty conditions. Evaluation of the approach, using DC motor driven servomechanism is performed.
展开▼
