Incipient fault diagnosis of analog circuits based on wavelet transform and improved deep convolutional neural network

Abstract

To enhance the reliability of analog circuits in electrical systems, this letter proposes a novel incipient fault diagnosis method by integrating wavelet transform(WT) and improved convolutional neural network. Different from traditional methods, where feature extraction and classification are separately designed and performed, this letter aims to automatically learn fault features and classify the type of faults simultaneously. An improved convolutional neural network named multi-channel compactness convolutional neural network (MC-CNN) is proposed,which can obtain complementary and rich diagnosis information from multi-scale components extracted by wavelet transform. Moreover, we adopt center loss as an auxiliary loss function to maximize the interclass separability and intraclass compactness of samples. The proposed method is fully evaluated with the Sallen-Key bandpass filter circuit and the fouropamp biquad high-pass filter circuit. The experimental results demonstrate that the proposed method is very effective in feature extraction for fault diagnosis, and has higher diagnosis accuracy than other typical fault diagnosis methods.

A Novel Analog Circuit Fault Diagnosis Method Based on Multi-channel 1D-Resnet and Wavelet Packet Transform

“…12 shows that the diagnostic accuracy of the remaining fault states has reached 100%. In order to prove that the proposed diagnosis method has a strong diagnostic ability, we compare the diagnostic results obtained by the MC-1D-ResNet classifier with other existing methods, including the methods in [5], [27] and [34]. As shown in Table 4 the method proposed in this paper has the best diagnostic result, that is, the classification accuracy of all fault states is 100%.Through the above analysis, it is concluded that our method has a wider diagnostic range and higher accuracy than other methods, which shows that the presented intelligent diagnosis method has better diagnostic performance than other methods.…”

“…13 is the result of the confusion matrix for CUT 2, including the diagnosis results of each fault state. As shown in the [5], [27] and [34]. Table 5 shows the comparison results of each method, and the methods in [5,27] get poor performances on fault states.…”

“…As shown in the [5], [27] and [34]. Table 5 shows the comparison results of each method, and the methods in [5,27] get poor performances on fault states. However, our method shows the best performance even when diagnosing the fault status of all components.…”

“…Yang et al adopted 1D-CNN network to extract the fault characteristics and classify faults [13]. Du [27]. However, the traditional 1D-CNN improves the ability to extract features for input data by stacking convolution layers, which would lead to many model parameters, large amount of calculation and difficult model training.…”

Sound pattern recognition method of transformer core based on convolutional neural network

“…Under the background of the new era, the economic market construction and urbanization construction process continue to deepen, and the power supply load standards of urban commercial and residential areas continue to improve, which leads to the increasingly serious environmental noise problem. Based on this, many scholars at home and abroad have carried out research on related problems, and through the optimization of transformer equipment, the effective control of vibration noise is realized, the noise level of transformer is reduced, and the noise environmental pollution problem is treated [1][2][3] . Based on this, this paper studies the vibration and noise test platform of the iron core model, analyzes the working state of the iron core displayed by the vibration signal and the sound signal, and on this basis, introduces the convolution neural network which extracts the voiceprint features of the iron core, and verifies the effectiveness of the model noise reduction by experiments.…”