Uncertainty Quantification of Radiation Pattern in Radome-Enclosed Antennas via Integrated Gaussian Process Regression and Grey Wolf Optimization

This letter proposes an efficient framework for uncertainty quantification (UQ) of radiation patterns (RPs) in radome-enclosed antennas (REAs), addressing variations in wall thickness and permittivity. A novel integrated Gaussian process regression (I-GPR) surrogate model is developed to overcome the computational complexity of conventional GPR in RP prediction. Furthermore, the global grey wolf optimization (GWO) algorithm is integrated to systematically determine the upper and lower bounds of RPs at each observation angle. Compared to conventional Monte–Carlo methods, the proposed approach reduces simulation costs by up to 91.9%. Moreover, the predicted uncertainty bounds achieve a root mean square error below 0.04, outperforming benchmark techniques. Numerical validations demonstrate that the proposed method achieves robust UQ performance under radome parameter variations of up to 60% in wall thickness and 47.6% in relative permittivity, ensuring reliability across broad operational conditions.