Orbit is a significant symptom in the fault diagnosis of rotating machine. The orbit is a 2-D image and can be described by moment invariants, the shape property of 2-D image, which is a description with translating-, rotating-, and scaling-invariants for 2-D image. The descriptive method of orbit image is investigated and an automatic orbit shape recognition based on artificial neural network (ANN) with moment invariants is proposed in this paper. The ANN of orbit shape recognition is trained by the training patterns generated by computer simulation for plenty of orbit shapes. It is shown that the trained ANN is of good recognition performance and generalization capability when applied to recognition of the measured orbits. This method can be used to the intelligent expert system of fault diagnosis to obtain automatically online orbit symptom in shafts vibration monitoring of turbine generator, which will improve the automatization of obtaining fault symptom and the automatic diagnosis in the expert system.
Orbit Shape Automatic Recognition Based on Artificial Neural Network
Du, D, & He, Q. "Orbit Shape Automatic Recognition Based on Artificial Neural Network." Proceedings of the ASME 2005 Power Conference. ASME 2005 Power Conference. Chicago, Illinois, USA. April 5–7, 2005. pp. 489-492. ASME. https://doi.org/10.1115/PWR2005-50208
Download citation file: