期刊名称：International Journal of Innovative Research in Science, Engineering and Technology
印刷版ISSN：2347-6710
电子版ISSN：2319-8753
出版年度：2014
期号：ICIET
页码：156
出版社：S&S Publications
摘要：Neural networks (NNs) have beensuccessfully applied to solve a variety of applicationproblems including nonlinear modelling andidentification. The main contribution of this paper ismodeling and identification of pH process based onrecurrent neural networks. The most powerful types ofneural network-based nonlinear autoregressive models,namely, Neural Network Auto-Regressive MovingAverage with eXogenous input models (NNARMAX),Neural Network Output Error Models (NNOE) andNeural Network Auto-Regressive model with eXogenousinputs models (NNARX) will be applied comparatively ofthe pH process identification. Moreover, the evaluation ofdifferent nonlinear Neural Network Auto-Regressivemodels of pH process with various hidden layer nodes iscompletely discussed. On this basis the features of eachidentified model of the highly nonlinear pH process havebeen analyzed and compared. The performance analysisshows that the nonlinear NNARX model yields moreperformance and higher accuracy than the other nonlinearNNARMAX and NNOE model schemes. The proposedmethod to identification is not only of the pH process butalso of other nonlinear and time-varied parametriicindustrial systems.
