Internal Vibration Source Inversion of Gas Insulated Switchgear by Ensemble Kalman Filter - Finite Element Analysis and Particle Swarm Optimization

Vibration characteristics are significant mechanical state information in gas-insulated switchgear (GIS). However, obtaining internal vibration data from enclosure data for equipment diagnosis is still challenging due to the unclear nonlinear transfer relationship between the internal structure and the enclosure. This study presents a novel inversion method combining ensemble Kalman filter enhanced finite element analysis (EnKF-FEA) with particle swarm optimization (PSO) to reconstruct internal vibration sources using GIS enclosure vibration data. Firstly, the EnKF-based data assimilation method corrects the simulation accuracy of the model. The mass distribution is one of the state variables of the simulation model, with a sensitivity of 4.98, and is therefore taken as the correction object for data assimilation. Second, through transient structural FEA and structural intensity (SI) analysis, power flow paths were identified to optimize measurement vibration point selection. Finally, the target function was constructed by comparing the simulation and experimental data of multi-point vibrations on the enclosure. The forward FEM was integrated with the PSO algorithm to achieve the inversion of vibration source information within the GIS. This inversion algorithm overcomes the limitations of traditional methods based on transfer functions, which heavily rely on regularization effects. The above non-invasive inversion framework reconstructs internal excitation sources through sparse enclosure measurements, aiming to overcome the limitations of traditional transfer function-based methods. It provides a practical pathway for reliable internal anomaly detection without the need for direct internal sensors, enabling dependable internal anomaly detection while reducing reliance on direct internal measurements. This offers important technical support for achieving GIS health management through non-destructive monitoring.