Mathematical modelling using advanced approach based on the neural networks has been applied to the control and the quality optimisation in the main processes of steelwork such as the ladle metallurgical treatment and continuous casting. Particular importance has been given to the improvement of breakout prediction system and the reduction in the rate of false alarm generated by the conventional breakout detection system. Prediction of the chemical composition and temperature of liquid steel in the ladle has been achieved by neural networks and linear model. This prediction can be considered as a soft sensor. Slab surface temperature stabilisation on the basis of the casting events has been controlled by a neural networks algorithm, that gives an improvement in the surface temperature fluctuation in comparison to the conventional control system which is based on the PID controller. Quality monitoring and classification is also achieved by a neural network which is related to the breakout detection system. This technique achieves a classification of different defects based on the different alarm signal given by the breakout prediction system. Fault detection and process monitoring is developed using neural networks modelling. All models are developed on basis of practical operating database obtained from the iron and steel industry.
| Identifer | oai:union.ndltd.org:DRESDEN/oai:qucosa.de:swb:105-6900128 |
| Date | 29 July 2009 |
| Creators | Bouhouche, Salah |
| Contributors | TU Bergakademie Freiberg, Maschinenbau, Verfahrens- und Energietechnik, Prof. Dr.-Ing. Jürgen Bast, Prof. Dr.-Ing. Jürgen Bast, Prof. Dr.-Ing. Dieter Janke, Dr.-Ing. H.-J. Hartmann |
| Publisher | Technische Universitaet Bergakademie Freiberg Universitaetsbibliothek "Georgius Agricola" |
| Source Sets | Hochschulschriftenserver (HSSS) der SLUB Dresden |
| Language | English |
| Detected Language | English |
| Type | doc-type:doctoralThesis |
| Format | application/pdf |
Page generated in 0.002 seconds