Research On Adaptive Neural Network Controller For Negative Pressure Servo System

The negative pressure servo system can generate high precision and dynamic pressure signals,which can provide a realistic air pressure environment for the semi-physical simulation of the aircraft.There are strong nonlinearity and uncertainty problems in the negative pressure servo system caused by the flow characteristics of the servo valve,the heat transfer effect of the capacitor cavity,and the internal leakage of the servo valve,which make the linear controller ineffective in the large range pressure control.To solve these problems,a neural network model reference adaptive control strategy was proposed,and an adaptive neural network controller was designed.The system was simulated and experimented,and the experimental results show that the adaptive neural network controller designed in this paper can achieve consistency in widerange pressure control.The following work was carried out in this paper.First,the mathematical model of the system was established,and the strong nonlinearity and uncertainty problems of the system caused by the flow characteristics of the servo valve,the heat transfer effect of the capacitor cavity,and the internal leakage of the servo valve were analyzed theoretically.The system was simulated and the results showed that the control effect of the linear controller gradually becomes worse with the change of the pressure point.In addition,there are uncertainty problems in the actual system,and it is difficult to establish an accurate system model,and it is difficult to control the system effectively using the modelbased control strategy.Considering that neural networks have nonlinear mapping capability and self-learning capability,which are suitable for solving nonlinear and uncertainty problems,a model reference adaptive control strategy based on neural networks was proposed.Then,the neural network model reference adaptive control strategy was specifically designed in a negative pressure servo system.In order to transfer accurate control error gradient information to the neural network controller,a neural network discriminator for the negative pressure servo system was designed using the NARX model.In order to determine the ideal control performance of the system,a reference model was designed according to the system characteristics and the system control requirements,and a dynamic neural network was used to design the negative pressure servo system neural network controller,and the stability of the negative pressure servo system based on the neural network controller was analyzed.The proposed control strategy was simulated,and the results show that the designed discriminator can accurately identify the system model,and the designed controller can overcome the nonlinear influence of the system and achieve the consistency of the large range pressure control.Finally,an experimental platform for the negative pressure servo system was established,and the neural network controller for the negative pressure servo system was studied experimentally.The experimental results show that the neural network controller can overcome the influence of nonlinearity and uncertainty in the system,improved the dynamic and static performance of the system compared with the linear controller,achieved consistent control effect at each pressure point,solved the problems of nonlinearity and uncertainty in the negative pressure servo control,and realized the consistency of the large range pressure control.