The electrode regulator system is a complex system with many variables, strong coupling and strong nonlinearity, while conventional control methods such as proportional integral derivative (PID) can not meet the req...The electrode regulator system is a complex system with many variables, strong coupling and strong nonlinearity, while conventional control methods such as proportional integral derivative (PID) can not meet the requirements. A robust adaptive neural network controller (RANNC) for electrode regulator system was proposed. Artificial neural networks were established to learn the system dynamics. The nonlinear control law was derived directly based on an input-output approximating method via the Taylor expansion, which avoids complex control development and intensive computation. The stability of the closed-loop system was established by the Lyapunov method. The current fluctuation relative percentage is less than ±8% and heating rate is up to 6.32 ℃/min when the proposed controller is used. The experiment results show that the proposed control scheme is better than inverse neural network controller (INNC) and PID controller (PIDC).展开更多
基金Project(N100604002) supported by the Fundamental Research Funds for Central Universities of ChinaProject(61074074) supported by the National Natural Science Foundation of China
文摘The electrode regulator system is a complex system with many variables, strong coupling and strong nonlinearity, while conventional control methods such as proportional integral derivative (PID) can not meet the requirements. A robust adaptive neural network controller (RANNC) for electrode regulator system was proposed. Artificial neural networks were established to learn the system dynamics. The nonlinear control law was derived directly based on an input-output approximating method via the Taylor expansion, which avoids complex control development and intensive computation. The stability of the closed-loop system was established by the Lyapunov method. The current fluctuation relative percentage is less than ±8% and heating rate is up to 6.32 ℃/min when the proposed controller is used. The experiment results show that the proposed control scheme is better than inverse neural network controller (INNC) and PID controller (PIDC).