Millimeter-level surface flatness is critical for spaceborne synthetic aperture radar deployable planar antennas to ensure high-resolution imaging performance. This study addresses the core challenge of maintaining flatness accuracy under multisource disturbances–including manufacturing/assembly errors and joint clearance–in a Formula: see text carbon-fiber-reinforced antenna with honeycomb panels. To overcome the progressive distortion of traditional finite element models (FEMs) in long error-transmission chains, we propose a hybrid FEM–least-squares support-vector machine (LSSVM) framework. Herein, FEM captures deterministic physical laws, while LSSVM compensates for nonlinear random errors induced by joint clearance, establishing an accurate actuator-displacement mapping. Further, a dynamic sensitivity threshold method adaptively selects actuator subsets by tracking extremum migration, enabling rapid convergence with cumulative adjustments. Full-scale experiments demonstrate that the method achieves 0.5 mm flatness under random disturbances–surpassing genetic algorithms by 93% in computational efficiency. This work provides a validated solution for active precision control of large deployable space structures.
Shi et al. (2026) studied this question.