Community Research and Development Information Service - CORDIS

Abstract

Different methods using model-based, knowledge-based, and AA-based techniques have been investigated concerning their potential to improve blast furnace control. Assessments of their operational applicability and efficiency have been included.

The theory of model based predictive control has been developed and its performance has been compared to different types of controllers. The best performance was obtained by changing existing operational manual control rules so that they performed an integrating action and employed dynamic coal/coke linkage.

Several diagnostic techniques have been investigated for early identification of furnace instability. Principal component analysis together with other statistical techniques have been used to identify channelling away from the stock rods and provide a furnace stability index.

The use of filtering and trending of measured and expected top gas hydrogen has been shown to provide early identification of cooling water leakage into the lower part of the furnace.

Models for predicting the hot metal temperature at a blast furnace have been developed. The mean tap value of HM temperature can be predicted by a NN-model with an accuracy of nearly 8C standard deviation. The prediction horizon is 3.5 to 4 hours.

Different methods for extracting control knowledge from observed process data have been investigated, assessed, and procedures developed on how to apply them.

Inductive decision tree methods can elicit the control knowledge from process data automatically. But the decision trees show structures that are not always easily understood. Model-based methods are not appropriate for eliciting control knowledge because of their inflexible structure and high requirements of computing time. The accuracy obtained with fuzzy logic extracting methods does not yield the desired values. red values.

Additional information

Authors: PLM H D, BFI, Bietriebsforschungsinstitut des VDEH, D³sseldorf (DE);HAGUE M J, British Steel, Teeside Laboratories, Middlesborough (GB)
Bibliographic Reference: EUR 19342 EN (2000) 152pp., EUR 23.50
Availability: Available from EUR-OP sales agents
ISBN: ISBN 92-828-9238-7
Record Number: 200012056 / Last updated on: 2000-06-16
Category: PUBLICATION
Original language: en
Available languages: en
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top