Research on Mechanical Fault Diagnosis of High-Voltage Circuit Breaker Based on Attention Parallel Acceleration

Aiming at the problem of high-voltage circuit breaker mechanical fault diagnosis, this paper proposes a highvoltage circuit breaker mechanical fault diagnosis model based on attention parallel acceleration by taking advantage of the correlation between the vibration signal and the mechanical state. First of all, the vibration signal feature extraction is realized through the convolutional neural network architecture of onedimensional aggregated attention mechanism; at the same time, the Laplace factor carried by the one-dimensional feature itself is used to construct the Laplace parallel acceleration channel architecture, which accelerates the convergence rate of the model; finally, the distribution of the same category of correlated subdomains is adjusted by combining the domain migration learning and the local adaptation algorithm to get the classification of vibration signals, and the three types of mechanical faults, such as closing spring and fatigue failure, are diagnosed efficiently. The vibration signal classification is obtained to effectively diagnose three types of mechanical failures, such as the fatigue failure of the closing spring. The experimental results show that the average accuracy of this model is more than $2 \%$ higher than that of existing detection methods, and its inference speed is at least $18 \%$ higher than that of existing one-dimensional convolutional neural networks under the condition of not exceeding $\pm 1 \%$ accuracy range interval.