Analog circuit fault diagnosis based on common spatial patterns and extreme learning machine

作者：Gao, Kun^*; He, Yigang; Bo, Xianglei; Tan, Yanghong; Tong, Yaonan

来源：Chinese Journal of Scientific Instrument, 2015, 36(1): 126-133.

• Analog circuit fault diagnosis
• Common spatial patterns
• Dimension reduction
• Extreme learning machine
• Information representation
• Network training
• Principal Components
• Projection space

Principal component analysis is an analysis method with algebraic feature and linear mapping, the information representation is generated from the linear mapping information after dimension reduction. The more important information is kept in the projection space, and the remaining information is filtered out. All the information is integrated, and it is not that every eigenvector represents a principal component, to a certain extent, which affects the effectiveness of PCA method. This paper presents an analog circuit fault diagnosis method based on common spatial pattern, which improves the principal component method. This method adopts CSP algorithm to process the eigenvectors obtained with PCA, and then sends the obtained principal components to extreme learning machine to conduct network training and fault diagnosis. An example of Sallen-Key band-pass filter circuit was used to verify the proposed method, and the result shows the effectiveness of the proposed method.

• 出版日期2015
• 单位合肥工业大学