Enhanced Sparse Polynomial Chaos Expansion for Electromagnetic Compatibility Uncertainty Quantification Problems

In electromagnetic compatibility analysis, the quantification of uncertainty frequently encounters challenges caused by the curse of dimensionality, particularly in high-dimensional models with complex data characteristics. This article proposes an enhanced sparse polynomial chaos expansion method that combines hyperbolic truncation, E-optimality criterion, and the subspace pursuit algorithm to improve both computational efficiency and model accuracy. The proposed framework enables accurate estimation of polynomial coefficients with fewer samples by jointly optimizing the selection of basis functions and sample points. A cross-validation procedure is further incorporated into the algorithm to enable automatic and adaptive selection of the sparsity level, enhancing generalizability across different uncertainty models. A comprehensive convergence analysis is conducted, revealing that the method achieves reliable accuracy when the sample size is approximately equal to the number of basis functions and the leave-one-out error of sparsity level converges. Numerical simulations on a shielded cable crosstalk model and a lightning return-stroke electromagnetic field model demonstrate that the proposed method maintains high quantification accuracy while significantly improving computational efficiency. These results confirm the robustness, adaptability, and effectiveness of the proposed method for high-dimensional stochastic problems in electromagnetic compatibility analysis.