Abstract: It is very important to acquire the easily understood diagnosis knowledge rules of rubbing fault, in order to further understand the rubbing fault mechanism and effectively diagnose the rubbing fault. In this article, a rule extraction method based on the functional point of view is studied, and the key algorithms are introduced, such as the discretization of continuous attributes, the generation of train samples of neural network (NN), the training of NN, the generation of instance samples from the trained NN, and the rule extraction. The Iris dataset is used to verify the rule extraction method. Finally, rotor-stator rubbing fault samples are obtained by an aero-engine rotor experimental rig, the rule extraction method is used to extract the rubbing fault diagnosis knowledge rules from fault samples, the obtained rules are verified and analyzed, and the results fully show the correction and rationality of the new method.

Key words: aero-engine, rotor-stator rubbing, knowledge acquisition, neural network, rule extraction