A hybrid fault diagnosis method for mechanical components based on ontology and signal analysis

Fault diagnosis of mechanical components has been attracting increasing attention. Researches have been carried out to reduce unnecessary breakdowns of machinery. Signal processing approaches are the most commonly used techniques for fault diagnosis tasks. Ontology and semantic web technology have great potential in knowledge representing, organizing and utilizing. In this paper, a hybrid fault diagnosis method for mechanical components is proposed based on ontology and signal analysis (HOS-MCFD). The method is a systematic approach covering the whole process of fault diagnosis: feature extraction from raw data, fault phenomenon identification using continuous mixture Gaussian hidden Markov model and fault knowledge modeling and reasoning using ontology and semantic web technology. A semantic mapping approach is presented to relate signal analysis results to ontology elements. The hybrid method integrates the advantages of signal analysis and ontology. It can be applied to deal with fault diagnosis more accurately, systematically and intelligently. This method is assessed with vibration data of rolling bearings. The experimental results prove the proposed method effective.