Study On Stress Wave Signal Of Initial Fault Acquisition And Separation For Low-speed Machinery

Condition monitoring using vibration analysis is an established and effective technique for detecting the loss of mechanical integrity of a wide range and classification of rotating machinery. Equipment rotating at low rotational speeds presents an increased difficulty to the diagnostician, since conventional vibration measuring equipment is not capable of measuring the fundamental frequency of operation. Also, component distress at low operational speeds does not necessarily show an obvious change in vibration signature.Against the difficulty of the fault diagnosis for low-speed rotating machinery, this paper presents a study of high-frequency stress wave analysis as a means of detecting the early stages of the loss of mechanical integrity in low-speed machinery. The feature frequency of fault stress waves is extracted by using wavelet analysis and blind source separation and then a fault diagnosis can be made.Firstly, a detailed study for fault mechanism of low-speed rolling bearing is made and then the disabled modalities and characters of fault are emphasized in the paper. The feature frequency of the fault signals can be educed by accounting. Therefore it lays the foundation for further investigation of fault diagnosis based on stress waves of rolling bearing with slow speed.Secondly, the properties and characters of blind source separation and wavelet analysis are introduced in brief, and then these two methods are felt together in fault diagnosis of slow-speed rolling bearings. Blind source separation is a method of signal processing, which can separate individual signal from mixtures of multiple faults. The background noise is eliminated by using wavelet decomposition and the feature frequency of fault stress waves is extracted.Finally, a simulation experiment on fault diagnosis of low-speed rolling bearing in laboratory. Then the fault stress wave signals which are obtained from the 600kW Wind Turbine are analyzed in the paper and the feature frequency of individual fault signal is extracted from mixtures of multiple faults successfully.