Identification of distributed parameter system with recurrent trajectory via deterministic learning

Abstract

In the paper, a novel method for identification of distributed parameter system (DPS) with recurrent trajectory via deterministic learning is proposed. The system dynamics rather than the system parameters or system structure is identified, which is completely different from the existing literature. The DPS, which is often described by partial differential equation (PDE), is first transformed into a set of ordinary differential equations (ODEs) by using a special case of finite difference method, method of lines. Then, the system dynamics at each discrete node is identified by using deterministic learning. Without loss of generality, the DPS dynamics at any spatial point in the spatial domain Ω can also be accurately identified by using the proposed method. It will be more appropriate for DPS identification in many physical systems whose both structure and parameters are unknown. Simulation results have shown the feasibility and efficient of the proposed method.