FNO-Based Data-Driven Modal Control of Flexible Solar Arrays

Flexible solar arrays often induce complex rigid-flexible coupling, which affects spacecraft stability during maneuvers. Traditional model-based controllers are vulnerable to unmodeled dynamics, while conventional neural networks struggle with rapid oscillatory behaviors. This paper proposes a hybrid vibration suppression framework: an Analytical Mode Method (AMM)-based Linear Quadratic Regulator (LQR) controller combined with a Fourier Neural Operator (FNO) neural network compensation. The controller, derived from the Euler-Bernoulli beam formulation, manages nominal dynamics. The FNO-based network identifies and compensates for nonlinear residual dynamics in both the spectral and time domains. Simulations demonstrate that learning in the joint time-frequency domain achieves faster convergence and superior vibration suppression compared to conventional neural network compensators based on linear kernels. The framework offers a computationally efficient solution for real-time vibration control of large flexible space structures.