NEURAL NETWORK ASSISTED NONLINEAR CONTROLLER

In this paper, design of a nonlinear controller for a Bioreactor Benchmark Problem is presented. The system under control simulates the dynamical behaviour of a biochemical process. The process has a few state variables but controller design is highly involved due to the nonlinear behaviour of the process, existence of limit cycles in the uncontrolled dynamics and the inherent delays. The proposed controller is also nonlinear, proven to be stable in the sense of bounded input/bounded output and locally stabilizing in the sense of Lyapunov. Additionally, neural networks are utilized in the controller design. A neural network is trained to partially identify the process dynamics and the trained network is used in the closed-loop control system. An error analysis is given for the neuro-identifier assisted controller performance. 1