Application Of Mathematical Morphological Filtering And Local Mean Decomposition In Gear Fault Diagnosis

As the transmission of power and changing the speed of commonly used components, gears and gear box widely used in the field of modern industry of machine tools, aerospace, power systems, agricultural machinery, transportation machinery, metallurgy and mining, etc.. And once the gears and gear box failure may cause the entire device cannot run normally. Therefore, the fault diagnosis of gears is of great significance.Aiming at gear vibration signals present in the background noise and vibration shock and other issues, this paper used a combined mathematical morphological filter and local mean decomposition. Gear fault vibration signal’s background noise can be filtered and vibration shock signal can be extracted through multi-structural and multi-scale mathematical morphological filter. On the basis of the combination of local mean decomposition process gear vibration signal, and then extract energy characteristic parameters of the local mean decomposition and normalized. Finally BP neural network is used to identify the classification of gear the various running status. The main work is as follows:(1) The characteristics of gear vibration signals, the main form of fault gear and gear fault influence on motion parameters.(2) Study the basic operators of mathematical morphology and morphological operators, the impact of structural elements on the mathematical morphological filtering, and gives the mathematical morphological filtering technology advantage in the vibration signal processing, we propose a multi-multi-scale mathematical morphological structure of the combination filter devices. Discussed in detail the basis of multi-multi-scale mathematical morphological structure of the combined filter structure elements of the construction method to the characteristic frequency of the intensity factor criterion, the use of a combination of structural elements sensitive adaptive mean filter multi-multi-scale structure, made more low-frequency signal extraction.(3) The basic theories, algorithm, the main characteristics, the problem of end effect and the energy feature extraction method of LMD are discussed in detail. And the problem of end effect is detailed analysis. Then a self-adaptive waveform matching extending method is proposed, which is used as an improved method. In the method the inherent law and changes of gear signal endpoints need be taken into account. Then the endpoints of gear signal proceed a self-adaptive waveform matching extending, thus the unconstrained situation at the endpoint was changed. The simulation and experimental results of the tests show that the method can effectively suppressed the end effect of local mean decomposition.(4) By analyzing gear vibration signals measured under different conditions with the method of mathematical morphological filter and local mean decomposition. The energy characteristic parameters of LMD are extracted and normalized. Finally BP neural network is used to classification. The method proved to be simple and easy, being an effective method of gear fault diagnosis.