Multi-feature Fusion Based on RBF Neural Network for Aviation AC Arc Fault Identification

摘要：

　　Aiming at the characteristics of an aviation arc fault, such as short duration, random current distortion, and difficult detection, etc.This paper proposed a multi-feature aviation AC arc fault identification method based on radial basis (RBF) neural network.Under the condition of 115V/400Hz, we perform the series and parallel arc test separately to collect the main circuit current signal during the typical linear load and non-linear load.The kurtosis and cosine distance are used to extract the time-domain characteristics of the current signal.The fast Fourier transform (FFT) is utilized to analyze the frequency-domain features, and construct the multi-dimensional eigenvector matrix based on time-domain and frequency-domain features, input RBF neural network for arc fault recognition.The experimental results show that compared with the nearest neighbor algorithm (KNN), multi-layer perceptron (MLP) neural network and decision tree, the RBF neural network has faster recognition speed and the highest recognition accuracy (98.9%), which is suitable for aviation AC series and parallel arc fault detection.

关键词： Series and parallel arc fault Multi-feature matrix Radial basis (RBF) neural network

作者: Ruihua Cui Chuanyu Wang

作者单位: State Key Laboratory of Reliability and Intelligence of Electrical Equipment, Hebei University of Technology, Tianjin,300130

会议类型: 国际会议

会议名称: The 7th Intrenational Conference on Reliability of Electrical Products and Electrical Contacts(第七届电工产品可靠性与电接触国际会议)

会议地点: 江苏苏州

会议语种:英文

页码: 380-387

在线出版日期: 2019-11-04（万方平台首次上网日期，不代表论文的发表时间）