Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.636309
Title: Advanced computer modelling in steel making
Author: Cox, I. J.
Awarding Body: University of Wales Swansea
Current Institution: Swansea University
Date of Award: 2004
Availability of Full Text:
Access from EThOS:
Abstract:
The most critical factors to control during steel making are carbon composition, steel temperature and time. At Corus Port Talbot Steel Works the priority is to achieve the aim temperature and carbon target ‘window’ at the end of the steel making process. An analysis of the historical end-point carbon and temperature performance revealed that in many cases the aim temperature and aim carbon composition was exceeded, and worst still, fall outside acceptable limits with expensive side effects. The use of computer control in the steel making process is essential to obtain accurate end-point temperature and carbon control in liquid steel. The current computer model employed to execute this task at Corus Port Talbot Steel Works is a procedural model that must be maintained by a person with considerable steel making knowledge. An initial investigation into the use of Artificial Neural Networks (ANNs) for prediction of oxygen and coolant requirements during the last few minutes of steel making proved that this technique can accurately model oxygen but has difficulty when modelling coolant requirements. An analysis of historical steel data on the carbon-temperature diagram was performed and a relationship between the position of the steel data (on the carbon-temperature diagram) at the in-blow sample point and values of the end-blow oxygen and coolant addition was discovered. Several further ANN models were created using steel process data where the target window was achieved. These models displayed improved accuracy when compared to linear and polynomial models. An assessment of each ANN model output concluded that the ANN models would provide improved end-point carbon and temperature control with more steel heat arriving in the end-point target window. The next stage is to investigate a strategy for providing feedback and updating the ANN model. Finally, when the ANN model is deployed, the control should be closed-loop so as to minimise the need for human intervention.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Eng.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.636309  DOI: Not available
Share: