Fault diagnosis method of static inclined eccentricity of synchronous generator rotor based on FWA-RF

To address the challenge of identifying rotor inclined and eccentricity faults in synchronous generators under complex electromagnetic environments, a fault diagnosis method is proposed based on the features of air-gap magnetic flux density and stator vibration acceleration, combined with a Fireworks Algorithm (FWA)-optimized Random Forest (RF) model. The method utilizes finite element simulations to construct the generator model, extract fault features, optimize the RF structure using FWA, and identify fault types through classification voting. The results demonstrate that, compared to Support Vector Machines (SVM), Decision Trees (DCT), and conventional RF methods, the FWA-optimized RF model significantly improves fault classification accuracy before and after noise addition, exhibiting excellent robustness and enabling more accurate fault diagnosis for synchronous generators.