Abstract: From the 1960s, the machinery fault diagnosis has rapidly developed as an emerging subject, whose prominent feature lies in close combination of theoretical research and engineering practical application. Over the past half a century, the machinery fault diagnosis has been developed into its maturity, which results in relatively comprehensive theoretical framework on signal acquisition and sensing technology, fault mechanism and symptom relationship, signal processing and diagnosis method as well as intelligent decision and diagnosis system. This subject has also made several original theoretical achievements, such as holospectrum diagnosis and crack dynamic quantitative diagnosis with wavelet finite element. Meanwhile, it has been effectively implemented in a number of engineering applications, such as machinery, metallurgy, petrochemical industry, energy and aviation, etc. However, there still exist some shortcomings in this subject, such as research blanks in fault mechanism, limitation in diagnosis methods, weakness in intelligent diagnosis equipment. Therefore, how to develop the basic research and explore the source of machinery fault diagnosis? What are development directions in the future? These are urgent problems needed to be addressed. Based on the review of the recent researches on basic theory of machinery fault diagnosis and its engineering practical application at home and abroad, the deficiency of the basic research and the breakthrough potentials of machinery fault diagnosis are presented.

Key words: Fault diagnosis, Fault mechanism, Feature extraction, Life prediction