Morphological Filtering And Fractal Features Extraction Method For Acoustic Emission Signals Of Journal Bearings

Hydrodynamic bearing is one of important support parts for steam turbine inpower plant， bearing fault is a direct threat to the normal operation of theturbo-generator unit．Now acoustic emission technology is better used to monitor andanalyze the operations of the bearings，but hydrodynamic bearing is in poor workingconditions and tradition sliding bearing lubrication judgment index response to earlyfailures are not sensitive enoughIn order to efficient purification bearing acoustic emission signals， adoptmorphological filter for noise reduction processing．Put forward a method that selectstructural elements according to the content of the noise signal．A large number ofexperimental analysis proved that the most suitable structure of the hydrodynamic bearingacoustic emission signals is rectangular structure elements and determine its the optimumsize．Analysis of field data showed that signal time domain charactreristic parameter(RMSvalue、Peak、Kurtosis factor) mutations when the hydrodynamic bearing lubrication failureoccurs．The comparison of signal time-domain characteristic parameters incremental sizeproved that morphological filtering is better than wavelet filtering．In view of traditional acoustic emission feature index can’t characterize hydrodynamicbearing lubrication early failure sensitivly，take correlation dimension as the feature index ofsliding bearings acoustic emission signals．Take the field test acoustic emission signals of apower plant310MW turbine sliding bearings as example，analyze its fractal characteristicsand determine the sliding bearing lubrication status by correlation dimension．Since classicG-P correlation dimension with a large amount of calculation，the author put forward a kindof improved algorithm through quickly determining the scale-free interval and increasing theembedding dimension incremental．The improved algorithm reduced the computation timeto81.36%of the classical algorithm on the basis of the calculation result is guaranteed．In order to simplify the procedure of the signal analysis，based on Matlab softwareplatform developed morphological filtering and fractal characteristics of the acousticemission signal analysis system with the function of morphological filtering, time domain feature extraction and analysis of fractal characteristics．The accuracy and reliability of thedeveloped system has been verified through the comparison with PCI-type sound emissiondetector．...