VIBRATION CONTROL OF SPACE STRUCTURES BASED ON AN IMPROVED PID NEURAL NETWORK USING OPTIMIZATION ALGORITHMS

To address the complex multivariable coupling issues inherent in the multi-actuator cooperative control of space structures, this study proposes an adaptive neural network proportional-integral-derivative (PID) controller optimized by a hybrid algorithm combining particle swarm optimization and genetic algorithm. A control framework tailored for multi-input multi-output systems is established, wherein the hybrid algorithm performs global optimization of the neural network's initial weights, while the gradient descent method facilitates real-time online parameter adjustment. This approach effectively resolves nonlinear coupling problems while minimizing dependence on precise mathematical models. System stability is rigorously proved using the Lyapunov function. Vibration suppression experiments were conducted on a multi-piezoelectric flexible beam and a multi-joint solar panel. Experimental results indicate that for the flexible beam, compared with the uncontrolled case, the proposed method shortens the vibration settling time by approximately 70.8%. In the solar panel system with unknown model dynamics, compared to fixed current control, the pointing settling time is reduced by approximately 48%. These findings validate the significant adaptability and robustness of the controller in the cooperative control of space structures.