A HIERARCHICAL MODELING TECHNIQUE OF INDUSTRIAL PLANTS USING MULTIMODEL APPROACH C. D. Stylios, N. G. Christova and P. P. Groumpos Laboratory for Automation and Robotics Dept. of Electrical and Computer Engineering University of Patras, GR-265 00 Rion, GREECE Phone: +30 610 997 295, Fax: +30 610 997 309 E-mail: {stylios,hristova,groumpos}@ee.upatras.gr Keywords: Hierarchical Structure, Intelligent Systems, Fuzzy Cognitive Maps, First Principles Model, Fuzzy Logic Model. Abstract This study investigates the problem of design adequate models for non-linear large and complex plants with high uncertainties. A new hierarchical structure is considered that utilize soft computing methodologies to model the supervisor. The proposed approach is based on the combination of different modelling techniques within a hierarchical supervised structure that has the ability to model system behaviour under different operational circumstances. A Fuzzy Cognitive Map (FCM) is used to aggregate multiple models and to create a hybrid model based on the current operational conditions of the industrial process. The proposed methodology is applied to model and simulate the operation of an industrial plant. 1 Introduction There is great need to develop reliable process models for the industrial plants within the different fields of computer-integrated manufacturing. The requirements for the higher quality of model design in the process industries have increased significantly in recent years. This leads to the investigation of design systems able to perform intelligent functions such as simultaneous utilization of memory capabilities, learning and high-level decision making procedures. Development of adequate models for industrial plants and complex systems is usually a complicated task because of the large uncertainties, caused by lack of direct measurements and necessity of inferential approach, high level of non linearity and different types of disturbances. One approach of building models that are accurate enough in a broad range of operational conditions may be successful if different modeling techniques are used that will create a hybrid methodology. Recent powerful development and use of intelligent technologies for the operation of complex industrial systems have been caused mainly by the intensive application of fuzzy logic and neural network methods [5]. Both fuzzy systems and neural networks have been shown to have the capability of modelling complex non-linear processes to arbitrary degrees of accuracy. Synergistic combinations of these methods can give even more effective performance results [5,6,8,9,10]. Fuzzy logic ideas have influenced relative areas, Kosko enhanced the power of cognitive maps [1] considering fuzzy values for the concepts of the cognitive map and fuzzy degrees of interrelationships between concepts. He introduced the Fuzzy Cognitive Map (FCM) theory as an integration of fuzzy logic and neural networks. Fuzzy Cognitive Maps (FCMs) have already been used to model behavioural systems in many Proceedings of the 10th Mediterranean Conference on Control and Automation - MED2002 Lisbon, Portugal, July 9-12, 2002.