Intelligent alarm handling in the steel manufacturing industry. The application of multivariate statistics to reheating furnaces and to an acid regeneration plant. Report no 1

O. Marjanovic et al.

In this project three specific programmes have been developed, each on a different process and with a different objective. These Programmes are

• Validation of thermocouple measurements in a reheating furnace (AvestaPolarit AB). It is shown clearly that an installed condition monitor would have detected an anomaly between measured and calculated temperatures and would have provided a valid estimate of temperature that could be used to temporarily replace the measurement during the time of anomaly. Such validation must provide the means for more effective energy management of the furnace by avoiding the positioning of control system set points at inappropriate temperatures.
• Development of a prediction model for NOx emissions in a reheating furnace (SSAB Tunnplåt).A prediction model has been developed which gives a satisfactory level of accuracy. Future developments should focus on incorporating the developed prediction model into the advisory system that would indicate which cause variables should be changed and by what amount in order to minimise NOx emissions while maintaining high productivity.
• Investigation of the use of Multivariate statistics for the modelling of an acid regeneration process (SSAB Tunnplåt). The investigations with the acid regeneration plant have proven to be less productive than those reported above for the reheating furnaces. However interesting aspects have been shown concerning the capability of the Multivariate Statistics to classify regions of process operation and to relate these regions to variations in key process variables