Modeling Industrial Processes with Hybrid-Neural Intelligent Systems (Intelligent advisory system for a Linz-Donawitz steel converter)

Intelligent Systems Research Group

Project aim:  The objective of the project to develop hybrid-neural intelligent technology to model and control complex industrial systems. The application field of the hybrid intelligent technology is steelmaking: modeling of a Linz-Donawitz converter (Basic Oxygen Steelmaking converter) and constructing an intelligent advisory system for determining the necessary amount of oxygen for obtaining steel of prescribed quality.  Period: April 1999. - April 2001. Sponsors: National Committee for Technological Development (OMFB) Financial support: 30 million HUF Members of the development consortium:   Dunaferr Co. (leader of the project)  BUTE Department of Measurement and Information Systems, University of Miskolc, Department of Physical Chemistry. Participants at the BUTE Department of Measurement and Information Systems: Gábor Horváth Béla Pataki György Strausz Péter Berényi

Some background and motivations: In many industrial complex modelling or control problems exact or even approximate theoretical/mathematical relationship between input and output data cannot be formulated. The reasons behind this can be the insufficient knowledge we have about the basic underlying physical behaviour, chemical reactions, etc., or the high complexity of the input-output relationship. In the latter case we could formulate "exact" mathematical equations describing accurately the operation of the industrial system/process, however these equations are too complex to solve them in real time or they need such pieces of information (e.g. measured data) about the process which cannot be obtained at all. In such cases experimental model can be constructed only. To construct experimental models a general modelling structure and appropriate experimental knowledge e.g. a large amount of measured data are needed.  To build experimental models for utilizing the different representations of heterogeneous information appropriate model architectures have to be selected. Hybrid intelligent systems that are built from neural networks and expert systems and that can utilize the traditional algorithmic solutions combine the advantageous properties of the different approaches.  Details: The modelled process The inverse model The hybrid intelligent approach Publications