Abstract
In the field of machinery condition diagnosis, it is most important and difficult to extract excellent symptom parameters for detecting failures. Spectrum of vibration or sound signal is usually used to monitor conditions and distinguish failures. However, in many cases, since the relationship between failures and spectrum cannot be made clear by theoretical analysis, and spectrum is always contaminated by noise, the symptom parameters are difficult to be extracted from the spectrum, by which the feature of the spectrum can be reflected for machinery condition monitoring. In order to overcome this difficulty, this paper proposes a sequential diagnosis method by statistical tests and possibility theory for condition monitoring of plant machinery. The feature frequency of vibration signal measured in each state can be extracted by the statistical tests, and symptom parameters in frequency domain can be calculated to monitor conditions of any machines of which failures can be detected by vibration or sound signal. In order to distinguish failures, the membership functions of the symptom parameters are established by probability density function, and failures can be sequentially distinguished by possibility theory. A practical example for gear equipment diagnosis is shown to verify the efficiency of the method.
