Fault Diagnosis for Mechatronic Systems Based on Fusion Model

Now the fault diagnosis of mechatronic systems generally has these characteristics, such as changeful operating conditions, serious signal coupling of noise and fault, and typical fault small samples. Based on multi-scale wavelet transform, extended rough neighborhood and the classification of support vector machine (SVM), a more effective method of mechatronic systems fault diagnosis was proposed. Firstly, the vibration signal was decomposed to each energy band according to frequency doubling and crossover of speed by multi-scale wavelet transform. Secondly, generalized distance was introduced, then the noise was eliminated and weight of energy-band importance was distributed through extending neighborhood rough sets. Finally, the model of dynamic fault diagnosis was established according to classification method of SVM in a small sample and non-linear space. This method is applied to the fault diagnosis of gearbox and proved that accurate and efficiency of the modeling method have greatly improved by the experiment results.