“…Under the frame work, recurrent neural network is developed to accommodate the on-line identification, in which the weights of the neural network are iteratively and adaptively updated through the model errors. The commonly used parameter identification procedures of the dynamic system include sliding-neural network (Ali et al, 2010), PID neural networks (Li and Liu, 2006a), functional series TARMA model (Poulimeno and Fassois, 2009), neural network modeling (Lee and Oh, 1997), hybrid genetic algorithm (Li and Liu, 2006b), frequency domain method (Peng et al, 2004), interval parameter estimation (Nazin and Polyak, 2005) and recursive incremental estimation (Zhou and Willianm, 1996). The simulated experiments were generally adopted to study the control performance of the earth-pressurebalance system of a shield machine economically and safely.…”