Rubbing Fault Diagnosis of Rotary Machinery Based on Wavelet and Support Vector Machine

The diagnosis method of rubbing fault in rotary machinery was investigated by support vector machine combined with wavelet transform. The rubbing fault of a rotary machine was simulated with a rubbing-block. The decomposed signals at every level were continuous for the case without rubbing fault, while the decomposed signals were bursting signals at level 1, level 2 and level 3, and continuous signals at level 4, level 5 for the case with rubbing fault. The correct rate of test samples was more than 92%, which indicated that the method can be used to identify rubbing faults effectively. The complexity of the SVM model was decreased and the calculation was simplified by using wavelet transform.