Synthesis of Neural Network Stabilizing Control for Objects with Mechanical Limiter

The article presents a new approach to creating a neural network simulator of a control object with a mechanical limiter, which allows you to identify the parameters of the limiter in real time. It is shown that the most effective way to identify such dynamic systems is to use neural networks based on ReLU activation functions. The effectiveness of the developed method is demonstrated by the example of the synthesis of stabilizing control for an aerodynamic pendulum using neural network predictive control.