A Real Time Model For Prediction Of Blast Furnace Hot Temperature Through Neural Network

Kumar, Avinash and Mrunmaya, K P and Maharana, Sudeep and Chowdhury, Subrata Kumar and Sah, Rameshwar and Kaza, Marutiram (2013) A Real Time Model For Prediction Of Blast Furnace Hot Temperature Through Neural Network. In: Proceeding of the International conference on science and technology of iron making and steel making , December 16-18, 2013, CSIR-NML Jamshedpur.

[img]PDF
Restricted to NML users only. Others may use ->

231Kb

Abstract

The thermodynamic processes inside the Blast Furnace are very complex and it’s ex difficult to analyze the ongoing process through physical measurements. The same can be estimated by monitoring measurable parameters which represent and indicate the state of the furnace. The hot metal and slag temperature, hot metal and slag chemical a blast temperature and volume are some of the parameters which have a bearing on the state of the furnace. Hot metal temperature is an important factor that not only depicts of the furnace but also determines the quality of pig iron. Hence, aiming at maintenconsistent the hot metaltemperature to achieve a stable operation of the furnace iso importance. Neural networks are parallel machines that have the ability to build and define relationships between various parameters and are self-learning to dynamically respon variations in operating conditions. The neural networks are capable of generating linear as nonlinear relationship between the parameters. This paper describes the approach adestimate the blast furnace parameters denoting the internal conditions which can ser guide line to the operator to take corrective actions in order to maintain smooth operatiofurnace and consistent production. The model developed has been tested and validated o periods with online data from furnace instrumentation to build the reliability of the preThe model is then integrated with the blast furnace automation to predict the Hotemperature in real time. The accuracy of prediction achieved is more than 90 %.

Item Type:Conference or Workshop Item (Paper)
Official URL/DOI:http://eprints.nmlindia.org/7076
Uncontrolled Keywords:Blast Furnace; Neural Networks; Forecasting; Simulation; Time Series; Hot Metal Temperature (HMT).
Divisions:Material Science and Technology
ID Code:7076
Deposited By:Sahu A K
Deposited On:14 Nov 2014 13:04
Last Modified:14 Nov 2014 13:04
Related URLs:

Repository Staff Only: item control page