Abstract:

     The paper presents a history fault database based fault diagnosis model, taking ETCs as an example. A fault symptom vector is initially generated through acquiring abnormal data stream and preprocessing them. It is then fed into a history fault database to match a fault pattern. The unmatched vector serves as a training sample to obtain the corresponding rule between fault symptom and fault cause through decision tree algorithm C4.5. The rule is saved into the history fault database and provide the reference for future diagnosis. Test shows that the diagnosis model has certain reliability and validity.

Key words: fault diagnosis modeling, data preprocessing, decision tree