Identification Of Three Tanks Water Control System Based On Neural Network

In this thesis, three tanks water control is composed by A3000 process control equipment There are traits of time-delay and non-linear in this system, it is difficult to realize by using traditional mechanism method to identify this system. For the non-linear mapping capability , artificial neural networks have been applied widely in the field of system identification and control engineering. Firstly, white noise input inspiring signal is designed by using LabVIEW software, and used in dynamic property testing experiment of three tanks water control system, and note the input\output dates. Secondly, based on MATLAB language, BP (Back Propagation) neural network and Elman neural network is adopted to identify the system. and the series-parallel structure is adopted. Levenberg-Marquardt (LM) algorithms, Quasi-Newton algorithms (BFGS) and self- adaption modified learning rate algorithms are used to train the BP neural network, and , in the Elman neural network, Quasi-Newton algorithms (BFGS) and self- adaption modified learning rate algorithms are used. The simulation results indicate that the generalization capability is not so well by using the error performance function of average-square, and this state can not be change by using LM algorithms with the error performance function of unitary average-square, but the generalization capability is improved by using the other two algorithms with this function. And in the Elman neural network, the effect of identification is well by using Quasi-Newton algorithms and self- adaption modified learning rate algorithms, there no any problem in generalization capability, and the structure of the neural network is more simple, but the identification precision is not reach but approach to the given error, less than the BP neural network. These two neural networks need to do more in-depth study to work as the system identification models exactly.