Study Of Signal Enhancement And Intelligent Diagnosis Method For Gearbox Early Fault

Gearbox has a very wide range of applications in mechanical system, which can change speed and translate power. In general, gearbox usually working in a harsh environment, so easily result in failure. So it is very significant to research on gearbox condition monitoring and fault diagnosis techniques, in particular for differentiate the early fault of gearbox. The impact signal caused by early fault is often weak so that would submerged in ambient noise, making it difficult to recognize, and the paper was studied the problems that how to diagnose the early single failure of gearbox.The gear and bearing vibration models are introduced respectively, both the gear vibration signal and bearing vibration signal are belong to second order cyclostaionary signals. Based on Wiener adaptive filter theory, cyclic adaptive filter is designed to de-noise the gearbox vibration signal. It is proved that cyclic adaptive filter technology is effective to extraction the early weak fault signal, and to enhance the signal pulse component. And meanwhile in this paper thirty four characteristic is extracted in the time domain, frequency domain, time-frequency domain, and non-Gaussian fields. In view of the lack of other traditional method for characteristic select, the article has proposed attribute reduction method based on rough sets, confirmed the Alpha stable distribution parameters as features of early fault, and selected a subset of features.Finally, based on the subset which after reduction of features, training and testing the Na?ve Bayes classifier by the data test on gearbox dynamic simulator, verified that the method is effective, and achieved goo d results.