Fault recognition system of electrical components in scrubber using infrared images

Abstract

In this paper, an automatic recognition system is described for diagnosing faulted patterns of the electrical components in the scrubber system. The implementation of this faulted recognition scheme integrates several technological issues. Firstly, Preprocessing techniques are applied for diminishing the environmental effects and background temperature. And then, the shape of the rising temperature area can be obtained clearly for the electric components. Thermal shape and temperature distribution are selected for feature extraction. The thermal shape is chosen to distinguish components under a loading condition and the temperature distribution can be used to evaluate the deterioration severity of a component. Finally, a radial basis function neural network is built to identify various failure modes. The accuracy reach 89.4% under 80 hidden nodes in this designed faulted recognition system. It reveals that the feasibility of this model can be used for diagnosis and classify the failure mode of electric components in a scrubber system.