Application of data reconciliation and gross error detection techniques to enhance reliability and consistency of the blast furnace process data

Abstract

Effective control of a blast furnace (BF) process requires accurate estimates of key process indicators (KPIs), namely, productivity, coke rate, direct reduction per cent, adiabatic flame temperature, bosh gas volume and top gas utilization. Some of these KPIs are obtained directly from the measurements, and some are derived by carrying out material and energy balances on measurements of different feeds, their compositions and temperatures. Due to errors in the measurements, the estimates of the KPIs can be inconsistent or misleading, which may result in misinterpretation of the current state of the BF process. Hence, it is necessary to reconcile the measurement data before these are used either for interpreting the current furnace state directly or as an input to other models. In the proposed methodology, data reconciliation and gross error detection techniques are used to improve the accuracy of the estimates of process variables and parameters, by ensuring that they satisfy process constraints such as elemental balances of iron, nitrogen, carbon, oxygen and hydrogen. Since the BF is a fed‐batch process, a customized version of these techniques has been developed and applied real time to an operating BF. The method is shown to be useful in deriving consistent estimates of the hot metal production rate, identifying gross errors in the online gas analyser and for estimating unmeasured parameters, such as top gas flow rate, its moisture concentration and calorific value which are useful for the purpose of stove heating in the downstream process.