Scholarlink

<p>With the increasing complexity of electronic circuits, identifying the location and type of faulty components has become more challenging. This paper proposes a dual-channel analog circuit fault diagnosis method based on voltage and current information. Focusing on circuit output location information, the method constructs a circuit model using the concept of circuit tolerance to generate a fault dataset, addressing the scarcity of real-world fault data. Discrete wavelet transform is employed to convert time-domain data into frequency-domain data, and a Redundant Feature Suppression (RSF) method is designed for feature extraction. A DCNN-BiLSTM fault diagnosis model is built by stacking a deep convolutional neural network with a bidirectional long short-term memory network. The experimental results demonstrate that the proposed method achieves an accuracy rate of 94.33% in identifying 14 types of faults in the Sallen-Key filter circuit and an accuracy rate exceeding 90% in identifying 24 types of faults in the four-op-amp high-pass filter circuit. Ultimately, a physical experiment was conducted to evaluate the practical effectiveness of the proposed method.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Method of Redundant Feature Suppression in Circuit Output Positions for Analog Circuit Soft and Hard Fault Diagnosis

  • 期刊论文

  • 发表日期：

    2025年8月1日

查看全文
  • Zhijun Yue,
  • Ling He,
  • Jin He,
  • Guanci Yang,
  • Yi Yang,
  • Yinhong An

摘要

With the increasing complexity of electronic circuits, identifying the location and type of faulty components has become more challenging. This paper proposes a dual-channel analog circuit fault diagnosis method based on voltage and current information. Focusing on circuit output location information, the method constructs a circuit model using the concept of circuit tolerance to generate a fault dataset, addressing the scarcity of real-world fault data. Discrete wavelet transform is employed to convert time-domain data into frequency-domain data, and a Redundant Feature Suppression (RSF) method is designed for feature extraction. A DCNN-BiLSTM fault diagnosis model is built by stacking a deep convolutional neural network with a bidirectional long short-term memory network. The experimental results demonstrate that the proposed method achieves an accuracy rate of 94.33% in identifying 14 types of faults in the Sallen-Key filter circuit and an accuracy rate exceeding 90% in identifying 24 types of faults in the four-op-amp high-pass filter circuit. Ultimately, a physical experiment was conducted to evaluate the practical effectiveness of the proposed method.