Community Research and Development Information Service - CORDIS


The present project deals with the investigation of fuzzy logic and artificial neural network techniques for controlling the iron making process.

The project is divided into five main parts: data handling, ANN applications. Forecasting, simulation and control.

Several types of data pretreatment are tested and compared. Exclusion of abnormal data set, frequency filtration of data, cross-correlation analysis and principal component analysis techniques are used for the preparation of virgin data. This part of the project has confirmed the importance of data handling on the final quality of the results.

Many approaches are used to analyze the forecasting problem in a blast furnace. In particular, two different methods are compared: auto-aggressive and multi-aggressive.

ANN, statistical and fuzzy logic methods are used to develop the blast furnace process simulators. A large number of simulators are obtained using artificial neural network methods even if this technology shows, for these applications, both precision and instability problems. Statistical methods reach about the same quality as those of the ANN; however the fuzzy logic approach obtains very encouraging results.

A fuzzy control structure for the hot metal temperature dynamic stabilization is developed and off-line tested by means of a mathematical BF simulator. Even if it can be considered as just a first analysis, a lot of really promising results were obtained. The main positive aspect was the possibility to obtain good results starting from very simple control rules and simple membership function structures.

In order to improve the control over the final sulphur content resulting from the desulphurization process, ANN is considered as an alternative method to conventional approaches. No significant improvement is obtained; however, better control over the desulphurization process can also be reached taking into account the less important parameters. This solution requires changes res changes

Additional information

Authors: CSM, ;ROME (IT), ;CENIM, ;MADRID (ES), ;VASL, ;LINZ (AU), ;TEA, ;PISA (IT),
Bibliographic Reference: EUR 19348 EN (2000), 249pp.
Availability: Available from EUR-OP Sales agents
ISBN: ISBN 92-828-9221-2
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top