Diagnosis of Bearing Fault Using Morphological Features Extraction and Entropy Deconvolution Method

Abstract

It is observed that the bearing failure of rotating machinery is a pulse in the vibration signal, but it is mostly immersed in noise. In order to effectively eliminate this noise and detect pulses, a novel an image fusion technology based on morphological operators inference is proposed. The correctness of morphological operators lies in the correct selection of structural elements (SE). This report presents an effective algorithm for SE selection based on kurtosis, which makes the analysis free empirical method. When analyzing three different groups of faults, the results show that this method effectively and robustly generates impulse. It enables the algorithm to detect early faults too. Recently, minimum entropy deconvolution (MED) was introduced to the machine in the field of condition monitoring, to enhance the detection of rolling bearing and gear failures. MED analysis helps to extract these pulses and diagnose their source, namely defects bearing components. In this research, MED will be reviewed and reintroduced, Application in fault detection and diagnosis of rolling bearings. MED parameters are selected and its combination with pre-whitening. Test cases are presented to illustrate benefits of MED technology. The simulation has been done on MATLAB and a graphical user interface has been created for analysis of bearing and detection of bearing faults using morphological features.