Control of the concentration in a nonisothermal CSTR with artificial neural network İzotermal olmayan sürekli̇ kariştirmali bi̇r tank reaktörde yapay si̇ni̇r aǧi i̇le deri̇şi̇ m kontrolü

YÜCEER A. M., Atasoy I., Berber R.

Journal of the Faculty of Engineering and Architecture of Gazi University, vol.21, no.4, pp.721-727, 2006 (SCI-Expanded) identifier

  • Publication Type: Article / Article
  • Volume: 21 Issue: 4
  • Publication Date: 2006
  • Journal Name: Journal of the Faculty of Engineering and Architecture of Gazi University
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, TR DİZİN (ULAKBİM)
  • Page Numbers: pp.721-727
  • Keywords: Artificial neural networks, NARMA-L2, Predictive control
  • Ankara University Affiliated: Yes

Abstract

In this work, controlling of the outlet concentration of a continuous stirred tank reactor (CSTR) was studied using artificial neural network. Control of the outlet concentration in the reactor where a non-isothermal, exothermic and a first order irreversible reaction took place was accomplished by manipulating the coolant flow rate. Due to the highly nonlinear dynamic behavior of the system, Neural Network Predictive and NARMA-L2 (Nonlinear Auto Regressive Moving Average) controllers were used as neural network control architecture. The results obtained with neural network controllers were compared with those obtained by classical PID controllers. Both neural network controllers were able to succeed in responding toward the set point faster and earlier in time than PID controller. Thus, neural network controller architectures showed better performances when compared to PID control strategies. In comparison to the literature where the same example problem was solved by using Dynamic Neural Network and Nonlinear Internal Model Controllers, the results obtained in this work seemed better to represent the system behavior than the literature.