An adaptive algorithm of Universal Learning Network for time delay system

This paper presents a new adaptive algorithm of Universal Learning Network (ULN) and its application to identify time delay of nonlinear black-box plant model. The ULN, a superset of many kinds of neural networks, consists of two kinds of elements: nodes and branches corresponding to equations and their relations in traditional description of dynamic system. Following the idea of ULN, the time delay parameters on the branches of ULN can be re-parameterized by the adaptive algorithm, based on the character in simulations of time delay system identifications and state stability analysis. One of distinctive features of the adaptive algorithm is that it can identify the pure time delay of the object model during identifications. The applicability and effectiveness of the adaptive algorithm are proved by simulation results. The general architecture and adaptive algorithm give ULN more representing abilities to model and control the nonlinear black box systems with time delay.