The quality prediction of iron ore pellets in grate-kiln-cooler system using artificial neural network

A model of three single layer back propagation(BP) artificial neural networks has been established to predict the compression strength of final pellets and dried pellets, the same as the shatter strength of the green pellets, according to the production data from SHOUGANG Mining Company. The Levenberg-Marquardt optimization arithmetic was used to train the model with the production data. After training, the error of the prediction result is less than 3%. The developed model can meet the requirement from production with a high accuracy and a wide flexibility.