Fault diagnosis using redundant data in analog circuits via slime module algorithm for support vector machine

• Published in: Journal of ambient intelligence and humanized computing Vol. 14; no. 10; pp. 14261 - 14276
• Main Authors: Yu, Danlu, Zhang, Aihua, Gao, Zhiwei
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.10.2023; Springer Nature B.V
• Subjects: Analog circuits; Ant colony optimization; Approximation; Artificial Intelligence; Circuits; Classifiers; Computational Intelligence; Design optimization; Electric potential; Engineering; Fault diagnosis; Foraging behavior; Genetic algorithms; High pass filters; Machine learning; Methods; Morphology; Neural networks; Optimization algorithms; Original Research; Particle swarm optimization; Performance enhancement; Principal components analysis; Redundancy; Robotics and Automation; Simulated annealing; Slime; Support vector machines; User Interfaces and Human Computer Interaction; Voltage; Fault diagnosis; Support vector machine; Slime mould algorithm; Redundant data; Principal component analysis
• ISSN: 1868-5137; 1868-5145
• DOI: 10.1007/s12652-023-04664-z

A large amount of redundant data brings challenges for fault diagnosis to achieve satisfactory performance. Therefore, it is particularly important to identify the fault quickly and accurately. In order to solve this problem, this paper applies a novel optimization algorithm, which is slime mould algorithm (SMA) combined with support vector machine (SVM), for fault diagnosis. Firstly, the experimental circuit is analyzed by Monte Carlo to obtain the voltage signals of different fault states. Then the collected voltage signal is subjected to wavelet packet transformation to extract the feature set of the data, and Principal Component Analysis (PCA) is used to reduce the dimension to eliminate redundant data. Finally, the SMA-SVM classifier is used for fault diagnosis and the results are analyzed. Two circuits are chosen as fault circuits, in order of complexity, the four-opamp second-order high-pass filter circuit and the Leapfrog filter circuit. Since SMA optimizes SVM to improve the performance of the classifier, the paper compares SMA with grid search method, particle swarm Optimization (PSO), genetic algorithm (GA), simulated annealing algorithm (SA), and ant colony algorithm (ACA) in terms of the results of the optimized parameters, the time of fault classification and the accuracy of diagnosis after optimizing the classifier. The results show that the SMA-SVM classifier not only demonstrates the advantages of SMA's excellent merit seeking ability and fast convergence, but also has better robustness.