Finite‐element analysis combined with an ensemble Gaussian process regression to predict the damper eddy current losses in a large turbo‐generator.

A method combining the finite‐element method (FEM) numerical calculation with machine learning is developed and used to calculate the eddy current losses of the rotor damping slot wedge of generator with different structures and electromagnetic properties. FEM simulation data is used as the input, and the stacking method is used to build the ensemble Gaussian process regression model for eddy current loss prediction to predict and analyse the calculation results. The error is tiny. The FEM results are highly consistent with the prediction data. As demonstrated by comparison experiments, the prediction accuracy of the stacking Gaussian process regression (SGPR) model is greater than that of other models. Therefore, the SGPR model provides a new auxiliary design method for large generators that can help improve generator design efficiency. [ABSTRACT FROM AUTHOR]