Topics: Control theory, Control engineering, PID controller Pages: 30 (4815 words) Published: March 29, 2013
Expert Systems
with Applications
Expert Systems with Applications 32 (2007) 911–918

Brain emotional learning based intelligent controller applied to neurofuzzy model of micro-heat exchanger
Hossein Rouhani a,*,1, Mahdi Jalili b,2, Babak N. Araabi b,
Wolfgang Eppler c, Caro Lucas b

Mechanical Engineering Department, University of Tehran, Tehran, Iran Control and Intelligent Processing Center of Excellence, Electrical and Computer Engineering Department, University of Tehran, Tehran, Iran

Institute of Data Processing and Electronics, Forschungszentrum Karlsruhe, Germany

In this paper, an intelligent controller is applied to govern the dynamics of electrically heated micro-heat exchanger plant. First, the dynamics of the micro-heat exchanger, which acts as a nonlinear plant, is identified using a neurofuzzy network. To build the neurofuzzy model, a locally linear learning algorithm, namely, locally linear mode tree (LoLiMoT) is used. Then, an intelligent controller based on brain emotional learning algorithm is applied to the identified model. The intelligent controller is based on a computational model of limbic system in the mammalian brain. The brain emotional learning based intelligent controller (BELBIC) based on PID control is adopted for the micro-heat exchanger plant. The contribution of BELBIC in improving the control system performance is shown by comparison with results obtained from classic PID controller without BELBIC. The results demonstrate excellent improvements of control action, without any considerable increase in control effort for PID + BELBIC. Ó 2006 Elsevier Ltd. All rights reserved.

Keywords: Intelligent control; Emotion based learning; Neurofuzzy models; Locally linear models; Nonlinear system identification; Heat exchanger

1. Introduction
Although industrial processes usually contain complex
nonlinearities, most of the conventional control algorithms
are based on a linearized model of the process. Linear


Corresponding author.
E-mail addresses:, (H. Rouhani), (M. Jalili), (B.N. Araabi), (W. Eppler), (C. Lucas).

Present address: Laboratory for Computer-Aided Design and Production, Institute of Production and Robotics, Swiss Federal Institute of Technology Lausanne (EPFL), CH 1015 Lausanne, Switzerland.
Present address: Laboratory for Nonlinear Systems, School of Computer and Communication Sciences, Swiss Federal Institute of Technology Lausanne (EPFL), CH 1015 Lausanne, Switzerland.
0957-4174/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2006.01.047

models can be identified in a straightforward manner from
process test data; e.g. via step or impulse response. However, if the process is highly nonlinear and subject to large frequent disturbances, a nonlinear model will be necessary
to describe the behavior of the process. For such systems
nonlinear identification methods should be used to describe the dynamic behavior of the system, which can be achieved
by means of neural networks. An alternative approach is to
design a nonlinear model consisting of several linear functions. The major output function is derived from a combination of linear models. Many training algorithms and structures are suggested for the mentioned networks such

as locally linear model tree (LoLiMoT), adaptive network
based fuzzy inference system (ANFIS), Takagi–Sugeno
(TS) and piecewise linear networks (PLN) (Eppler & Beck,
1999; Jang, 1993; Nelles, 1997; Sugeno & Kang, 1988). In
this work we will make use of LoLiMoT for training


H. Rouhani et al. / Expert Systems with Applications 32 (2007) 911–918

algorithm of the neurofuzzy network because of its rapid
and accurate operation in control applications.
We will use brain emotional based learning intelligent...
Continue Reading

Please join StudyMode to read the full document

Become a StudyMode Member

Sign Up - It's Free