Identification of nonlinear dynamic systems with large time delays based on universal learning network

Recently, neural networks are widely used to identify or model large scale and dynamic nonlinear systems. Current neural networks commonly have to have many nodes and links when it is used to solve large complicated problems. As a result, the structure of networks is quite miscellaneous and incompact, so that the overfiting phenomena of network weights often appear and generalization ability become worse. In additionally, the time delay of current neural network is not optional, differing with practical situation. In order to overcome the shortcomings of current neural network, the paper introduce a Universal Learning Network (ULN). The paper introduces a new type of neuril networkUniversal Learning Network to identify nonlinear dynamic systems in chemical engineering field. It also discusses the practical algorithms for realize simulation. Sample simulation results demonstrate that the Universal Learning Network can identify the pH neutralization process well with satisfactory precision. It can also describe the nonlinear and dynamic characteristics of the systems. Owing to the fact that only one neural network can also realize the identification of the sample system, the Universal Learning Network is a simple and effective method for simulating pH neutralization process.