Prediction of silicon content in blast furnace hot metal using Partial Least Squares (PLS)

An application of partial least square (PLS) for the prediction of silicon content in blast furnace hot metal, is presented. Dataset containing 120 X type variables, which included process variables such as coke rate, coal rate, amount and chemistry of other materials, hot metal and slag and furnace data, and one Y variable, that is, silicon content in hot metal, were recorded hourly for one month. Hourly prediction was done for the 31st day from the the hourly training data of previous 30 days,and it was found that the trend of predicted silicon content followed the actual values. The results show that the PLS model can be helpful in monitoring the hourly prediction of silicon, and in estimating the next day's average silicon content.