SBS_FD: Fault Diagnosis of Harmonic Reducers Based on Symmetrized Dot Pattern, Bag of Visual Word, and Support Vector Machine Jointed Method

In order to realize fault diagnosis of harmonic reducers in real industrial scenarios, a fault diagnosis method based on image visual information, i.e., SBS_FD is proposed. Firstly, Symmetrized Dot Pattern images are obtained based on vibration signals, and Scale Invariant Feature Transform method is used to extract image information, wherein the mappings between fault signals and visual information are achieved. Then the bag of visual word and space pyramid matching method are used to process the visual information. Finally, the support vector machine is used for fault diagnosis. Comprehensive experimental results show that the diagnostic accuracy of four fault states can be up to 98% at all speeds, using proposed SBS_FD method, illustrating superior performance over other machine learning methods.