A typical fault intelligent diagnosis algorithm based on VGG-SVM for N+1 rotor systems of steam turbine generators

With the continuous advancement of the Industrial Internet of Things, high-end turbomachinery, especially multi-support rotor equipment, is increasingly developed toward artificial intelligence. The healthy operation and maintenance of equipment have garnered widespread attention. Aimed at turbine generator sets with N+1 supports and supercritical rotor characteristics, a novel integrated fault diagnosis model algorithm based on VGG-SVM is proposed. This algorithm is capable of identifying multi-dimensional fused fault features and demonstrates advantages of high accuracy, good stability, and strong robustness. Typical characteristic mechanisms of structural faults, such as rotor imbalance, misalignment, and loosening, are introduced. A two-span three-support rotor test bench is constructed to conduct fault simulation experiments, during which data are collected and diagnostic tests of the algorithm are performed. Results indicate that the diagnostic accuracy of the integrated algorithm is on average 10%–20% higher than that of other algorithms, with superior model generalization capability. Finally, the algorithm is tested with 120 sets of test data, and the diagnostic accuracy is found to exceed 99%.