Detection of AC Arc Faults of Aviation Cables Based on H-I-W Three-Dimensional Features and CNN-LSTM Neural Network

Abstract

Aiming at the difficulty in identifying subtle AC arcs in aviation cables, this paper proposes an arc fault detection method based on the combination of three-dimensional features and convolutional neural network-long short term memory (CNN-LSTM). Firstly, based on the SAE AS5692A standard, the vibration series test, cutting parallel test, and wet arc trajectory parallel test were respectively conducted and the arc current signals under four types of loads were collected to analyze the arc faults under different incentives. Then, the three-dimensional features of arc current including Hurst exponent, inter-harmonic variance, and wavelet energy entropy (H-I-W) were extracted with an improved algorithm so as to enhance the fault identification capability and overcome the limitation of single-dimensional feature detection. Finally, a grid search algorithm was used to find out the optimal parameters, and a three-dimensional reference input CNN-LSTM neural network was designed to detect arc faults. The experimental results showed that the average detection accuracy of the proposed method for the three AC arc faults respectively reached 98.52%, 99.23%, and 98.51%. The real-time performance of the proposed method was better than the comparison methods, proving the feasibility and effectiveness of the proposed method.

Series Arc Fault Detection under Vibration Condition Based on N-M-M-B

“…Gao [9] used the squeeze-excitation network-deep convolution generative adversarial network (SE-DCGAN) to enhance the Gramian angular summation fields (GASF) image of the arc fault data to solve the problem of limited SAF samples. Liu [10] extracted the Hurst exponent, inter-harmonic variance and wavelet energy entropy (H-I-W) as the three-dimensional feature of the arc, and constructed the arc fault diagnosis model by combining the convolutional neural network (CNN) and the long short-term memory network (LSTM), which enhanced the fault identification ability. Xin [11] proposed a lightweight arc fault detection method based on an EffNet module, which can reduce the complexity of the algorithm at the same detection accuracy level.…”

“…Xin [11] proposed a lightweight arc fault detection method based on an EffNet module, which can reduce the complexity of the algorithm at the same detection accuracy level. Although the method based on artificial intelligence has achieved good arc fault detection accuracy, it is not clear whether the results of References [8][9][10][11] are applicable to the arc fault detection of a motor frequency converter load.…”

“…In References [3][4][5][6][7][8][9][10][11][12][13][14], good results have been achieved in the aspects of arc fault signal preprocessing, feature analysis and detection model establishment, but there are still the following shortcomings. In References [3][4][5][6][7], when the fault data are preprocessed, the empirical mode decomposition (EMD) correlation algorithms have problems such as modal aliasing or difficulty in aligning IMF components, which will have a certain impact on the diagnosis results.…”

“…In References [3][4][5][6][7], when the fault data are preprocessed, the empirical mode decomposition (EMD) correlation algorithms have problems such as modal aliasing or difficulty in aligning IMF components, which will have a certain impact on the diagnosis results. In References [8][9][10][11], the deep learning algorithm performs well in arc fault detection, but there is still some room for optimization. In Reference [10], only the global features extracted by the hidden layer at the end of CNN module are transmitted between the long short-term memory (LSTM) module and CNN module, and the local features contained in the middle hidden layer of CNN are not included.…”

“…In References [8][9][10][11], the deep learning algorithm performs well in arc fault detection, but there is still some room for optimization. In Reference [10], only the global features extracted by the hidden layer at the end of CNN module are transmitted between the long short-term memory (LSTM) module and CNN module, and the local features contained in the middle hidden layer of CNN are not included. As a result, the global features extracted by the constructed neural network only contain high-level features such as sequence information, and it is difficult to describe low-level features such as single-peak signals.…”