A multi-fidelity transfer learning strategy for surface deformation measurement of large reflector antennas

Abstract

As the observation frequency of large-aperture antennas increases, the requirements for measuring main reflector deformation have become more stringent. Recently, the rapid development of deep learning has led to its application in antenna deformation prediction. However, achieving high accuracy requires a large number of high-fidelity deformation samples, which is often challenging to obtain. To address these problems, this paper establishes a high-accuracy antenna surface deformation measurement model based on a multi-fidelity transfer learning neural network (MF-TLNN). Firstly, a low-fidelity surrogate model is constructed using a large number of simulation deformation samples to ensure its robustness. Secondly, the MF-TLNN structure is designed and trained using a small number of high-fidelity samples obtained from actual measurements of the main reflector deformation via out-of-focus (OOF) holography method. Thirdly, a Zernike correction module is utilized to provide additional constraints and ensure the stability of the results. Experimental results show that the proposed method can closely approximate radio holography measurements in terms of accuracy and is almost real-time in terms of speed.