An Integration Method for Rolling Bearing Fault Diagnosis
The paper presents an integration method of artificial neural network (ANN) and empirical mode decomposition (EMD) to identify fault severity in rolling bearing. A test apparatus is established, in which the rolling bearings with different faults and defect sizes are tested. Fault severity is divided into four grades of normal, light, middle and severe based on the defect size. Vibration signals are collected from the test rig. Due to the complexity of the signals, EMD has been exploited to decompose the signals into a series of intrinsic mode functions (IMFs). Then an ANN with one hidden layer is designed to diagnose the fault type and severity. The virtual values of first eight IMFs of a signal form the input vector of ANN. The output vector of ANN represented the fault severity with two binary digits. Appling the ANN to test the signals with unknown defect sizes, the diagnosis capability can arrive to about 95%. The results demonstrate that the integration method is successful in machine fault severity diagnosis.
L. Li et al., "An Integration Method for Rolling Bearing Fault Diagnosis", Advanced Materials Research, Vols. 228-229, pp. 293-298, 2011