摘要：Aiming at the characteristic that the vibration signals of hydraulic pump usually have strong nonlinearity and low signal-to-noise ratio, this article presents a novel hydraulic pump degradation feature extraction method based on improved local characteristic-scale decomposition and multi-fractal spectrum. First of all, the original vibration signal is decomposed into the independent intrinsic scale components by local characteristic-scale decomposition, and the main intrinsic scale components which contain the sensitive degradation information are selected by mutual information. And then, the multi-fractal parameters of the main intrinsic scale components are calculated. The presenting capability of four fractal spectrum parameters on hydraulic pump degradation state is analyzed, and as a result, the multi-fractal spectrum width Δ α is finally selected as the degradation feature parameter. Finally, the degradation features are inputted into a binary tree support vector machine to recognize the degradation states. The application results indicated that the proposed method can recognize the degradation states of hydraulic pump effectively.
关键词：Degradation feature extraction; local characteristic-scale decomposition; multi-fractal; hydraulic pump; binary tree support vector machine
