Robust identification of uncertain nonlinear systems with state constrains by Differential Neural Networks

Non parametric identifier based on differential neural networks is presented. Nonlinear systems with the a priori information about state constrains are considered. Differential neural network identifier includes a projectional operator as part of its structure in order to keep the state restrictions. Stability analysis based on Lyapunov-Krasovskii energetic-function lets proof the ultimated boundedness of error estimation. Free parameters stability of DNNI (adaptive weights) is also proved. Simulation examples regarding to biotechnological process and the Chua´s circuit depict the main advantages of the proposed structure against reported differential neural network with similar structure but without projectional operator.