Contents I. Introduction
Recently, neural network has been applied for many researches [1]. Due to having Fault-tolerance, ability of parallel processing and learning ability, neural network is very suitable to control nonlinear dynamic system. Because a recurrent neural network which has many feedback loops can deal with compound dynamic response, a recurrent neural network is better than forward neural network [2], [3]. A recurrent neural network can use natural temporal operation [2] to operate time-varying input and output data, and can control unmodeled dynamic system and external disturbance effectively. A fuzzy neural network has fuzzy consequence mechanism [4] to deal with uncertain information and has ability of on-line learning. [5], [6]. Lately, many papers adopted fuzzy neural networks to deal with nonlinear and uncertain problems [7], and developed a functional link neural network (FLNN). Input variables of FLNN can expand linear unattached functions [8]– [10], omitting hiding layers and simp the construction of multilayer apperceive mechanism, it makes a FLNN construction has much faster convergent speed and less account load than the construction of multilayer apperceive mechanism. In [9], a monolayer functional link neural network was applied for nonlinear dynamic identification system. A functional link-based fuzzy neural network (FL-based FNN) was developed to realize fingerprint identification and sound identification in [10].
