Fault-based Spectral Entropy Feature Extraction And Data Mining Technology Research

The gear transmission plays an important role in mechanical equipments, and the gear fault takes a big rate in the faults of machine. This paper chose the vibration signals of gear for studying, extracted the fault features to distinguish fault styles by the method of spectral entropy(SE), classed and clustered the signals of different gear sates. It also designed the visual instrument (VI) based on the processing before.In the diagnosis of gear faults, the key is how to extract the fault features from the random signals by proper methods. In fact, the signals will be polluted by noise. Some methods are sensitive to the noise and difficult to distinguish the faults, such as the power-spectrum. SE based on the statistics indicates the orderlessness and is robust to noise. In this paper, SE is introduced into the gear transmission system, the fault features of the gear with crack and abrasion are extracted, and compared to that of normal gear. SE is good at extracting and separating the features of gear signals, and can distinguish the unknown vibration signals reliably. It proved that SE is a feasible method to estimate the states of gear.Data mining which is an advanced method for extracting feature information can solve the insufficiencies of general methods of fault diagnosis, such as the "bottle-neck" in acquiring information and the rationality and reliability of consequence. This paper introduced the clustering into the fault diagnosis of gear. The K-mean can process the SE of one-dimension, and the method based on density and gird can process that of two-dimension. The result of DCT is good at separating signals while that of FFT is clever at predicting the trend of deterioration. It indicated that clustering is a good method for gear fault diagnosis and can forecast the fault styles..VI developed fast as a new technology in the fields of measurement and fault detection. This paper introduced the structures and characteristics of VI, developed the VI of fault diagnosis based on VC++ and MATLAB. The VI is based on SE and clustering, including some different modules, such as the signal inputting, SE compute, clustering and diagnosis.