Multifractal Study And Application To Gear’s Vibration Signal

As the important transmission parts, gears are widely used in various mechanicalequipments because of its high transmission efficiency, transmission ratio, transmissionand other characteristics of the acclaimed accurate. The signals’ nonstationary andnonlineary caused by complex gear transmission system increase the difficulty of thegear fault pattern recognition. Based on summary of the common types andcharacteristics for gear failures, the multifractal detrended fluctuatuin analysis(MF-DFA) and multifractal spectrum theory, which usually are used to extract thenonlinear non-stationary and multifractal characteristic of the gear fault vibrationsignals, are employed to gear fault feature extraction in our work. Combined with theGaussian Mixture Model, the gear fault detection are realized.The main researchescontents include:1、In this thesis the main failure modes of gear are summarized. By means of thevibration mechanism of gear analyzing, the common characteristic of gearbearing fault signals was also summarized.2、The MF-DFA is used to extract feature of gear fault signals, and the twodimension MF-DFA is proposed. It is proved that these methods are effectivethrough simulation signals, the experiment project and system are set up,choose the appropriate signal acquisition test equipments, acquisit the gearfault experiment data to verify these methods are practical.3、The single scale-free area multifractal spectrum theory is used to extract featureof gear fault feature signals, and the feature extraction method of multiplescale-free area multifractal spectrum theory is proposed. It is be proved thatthese methods are effective through simulation signals, using the same testmethod, these methods are proved to be practical by the simulation of gear faultexperiment datas.4、Combined with the gaussian mixture model, the proposed methods arecompared. The gear fault classification experiments’ results show that themultiple scale-free area multifractal spectrum theory possesses higher computational sfficiency and the recognition rate, and more appropriate for thefeature extraction of gear fault signals.