Application of morphological wavelet and fuzzy entropy in gear fault classification

As one of the key assemblies of armored vehicle transmission system, the gear failure of transmission often leads to the failure of the whole transmission box, which leads to the failure of the whole transmission system. However, due to the complexity of transmission structure and multiple excitation sources, transmission gear fault information is often submerged in strong noise signals. Therefore, the research on fault identification and classification of transmission gear under strong noise environment has become the research focus of transmission fault diagnosis. The research shows that the characteristics of the fault transmission gear vibration signal are typical non-stationary and non-linear, so it is very important to study the effective extraction and classification method of transmission gear fault signal. Huang et al. Extracted the normalized spectrum amplitude in gearbox fault diagnosis. Crabtree et al. Detected the early fault of gearbox from the perspective of multi characteristic parameters. Zimroz et al. Extracted the instantaneous speed of the gearbox. Tang Xin’an et al. Realized the preliminary diagnosis of gearbox fault with the help of time domain statistical index. Hu pengqing et al

Based on Hilbert Huang spectrum, the energy feature in the neighborhood of meshing frequency is proposed to detect the broken tooth fault of the sun gear of planetary gearbox. As a new nonlinear method, fuzzy entropy is used to measure the probability of generating new pattern when the dimension of sampling sequence changes. Its physical meaning is similar to sample entropy. Moreover, because the exponential function is used to replace the unit step function in fuzzy entropy, the continuity of the exponential function makes the fuzzy entropy value calculated by the sampling sequence present the trend of continuous and smooth change with the change of parameters. However, before feature parameter extraction, the original sampled signal will be interfered by various noises, so choosing the appropriate denoising method to preprocess the original signal will play a positive role in the subsequent extraction of fault feature parameters.

1) Due to the noise interference of the original fault signal, it is difficult to effectively identify the fault category because of the cross of fuzzy entropy directly used to calculate. After using morphological wavelet de-noising, the noise information in the original signal is well suppressed, and the calculated fuzzy entropy can be effectively used to distinguish different fault types.

2) The fuzzy entropy calculation and fault classification are carried out based on the measured gear signals of different working conditions in gear fault test-bed, which provides a new way for gear fault state identification and classification of armored vehicle transmission.

As a class of nonlinear wavelets, morphological wavelets not only keep the morphological characteristics of mathematical morphology, but also inherit the multi-resolution characteristics of traditional wavelets, and have good anti noise performance and good detail retention characteristics. Goutsias et al. Constructed morphological wavelets by using mathematical morphological operators instead of linear operators of traditional wavelets. Morphological wavelet is used to denoise the rotor signal, and good results are obtained. Therefore, based on the in-depth study of morphological wavelet and fuzzy entropy, a gear fault classification method based on morphological wavelet and fuzzy entropy is proposed in this paper. The vibration signals under different working conditions are preprocessed by morphological wavelet de-noising, and then the fuzzy entropy is extracted. At the same time, the fuzzy entropy is extracted from the original signal for comparison. The results show that the fuzzy entropy has better fault classification ability .

Scroll to Top