Research on intelligent recognition of axis orbit based on Hu moment invariants and fractal box dimension
Jie Shao, Xinyu Pang, Yang Zhao-jian, Juanli Li · 2017
The shape feature of axis orbit is very important to judge the fault of rotor system. After the sampling and quantization of the image of axis orbit, the edge of the image will be inaccurate. Therefore, the discretization of image has a great influence on the calculation of higher order moments of the moment invariants. High order moments are mainly used to describe image details. The fractal dimension is sensitive to the details and complexity of the image. Therefore, this paper proposes a method of combining fractal dimension with moment invariants. And Canny operator is used to detect the edge of axis orbit image. The Hu invariant moments of the image and the fractal box dimension are used as the feature vectors of BP neural networks. The 40 groups of samples of the typical fault axis orbit of the experiment are trained and sampled to test. The recognition rate of the 8 groups of axis orbit is up to 100%, and the recognition effect is satisfactory. The results show that the method has high recognition speed and high recognition accuracy, and has good practical value for the intelligent fault diagnosis of the rotor system by using the axis orbit.