Performance assessment of permeability index prediction in an ironmaking process via soft computing techniques

Abstract

Permeability index is a crucial productivity indicator of the lower zone in blast furnaces to maintain the operation, energy consumption, and hot liquid metal production rates during the ironmaking process. Blast furnace operation parameters such as coke-to-ore ratio, wall pressures and temperatures, flame temperature, top gas pressure, temperature and composition, hot blast pressure and temperature, sounding levels, etc. and also the level of hot liquid metal and slag in the bottom of furnace, influence the permeability phenomenon directly. Hence, fluctuations and instantenous variations of permeability index parameter should be avoided by controlling inadequate drainage cycles and operational factors to achieve more efficient and stable operation in the furnaces. In this study, permeability index parameter of the Erdemir Blast Furnace #1, located in Turkey, is modeled and experimental computing work is carried out to assess the operation performance of the furnace, depending on selected input parameters. The demanding artificial intelligence and soft computing techniques, artificial neural networks and adaptive neural fuzzy inference system, and a well-known statistical tool, autoregressive integrated moving average model are executed throughout the study using previous furnace data, received during one day of operation. Selected performance measures, coefficient of determination ( R2) and root mean squared error, are used to compare the forecasting accuracy of proposed models. Consequently, the most satisfactory forecasting model of the study, adaptive neural fuzzy inference system, is proposed to be integrated into the plant control system as an expert modeler.