Aviation Arc Fault Diagnosis Based On EMD-SVM

With the rapidly development of aviation industry, aviation security attracts more and more attention. However, the working environment of aviation cable is special. The influence of temperature, radiation, vibration, etc will accelerate the aging of aviation cable, which may lead to arc fault. Aviation arc fault has a high energy and it is easy to cause fire, which is a serious threat to aviation safety. Therefore, the research of the arc fault diagnosis technology is an important subject to ensure aviation safety.At first, the generation of aviation arc fault is concluded in this paper, and the Cassie arc model is used to simulate the series arc in Simulink. Then the test plan is designed and the aviation arc test platform is proposed based on standard UL1699. By using the test platform, massive data of arc fault are collected including resistive loads, inductive resistive loads and nonlinear loads on 400Hz AC power supply in order to establish database of aviation arc for further study.Empirical mode decomposition (EMD) is applied to extract the feature of aviation arc current after fast Fourier transform. What's more, three-order autoregressive (AR) models of intrinsic mode functions(IMF) are established. AR model parameters are calculated by using Burg recursive algorithm and the model parameters can be taken as characteristic parameters of aviation arc current.Finally, least-square support vector machine (LS-SVM) is applied to detect the aviation arc fault. The processed datas on linear load and nonlinear load are divided into training set and test set to set up the LS-SVM classifier, which is trained for aviation arc fault diagnosis. The results show that LS-SVM has a good performance in aviation arc fault diagnosis on nonlinear loads and unknown loads. It proved that EMD-SVM presented in this paper is suitable for aviation arc fault diagnosis.