investigative reduction of iron pellets into directed reduced iron as multi-variates models based on their analytical and structural properties to predict the reducing indexes

Reducing iron pellets to iron sponge using the reducing gases of hydrogen and carbon monoxide can lead to different sponges due to the different properties of initial pellets. as an experimental theoretical study, the properties of produced iron sponge including metallization degree, sulfur and carbon contents, and disintegration of pellets, were modeled to analytical and structural properties of initial iron pellets. in this work, more than one hundred iron pellets, examined by instrumental techniques, were used to achieve the more practical models. contents of total iron, ferrous and metallic ones were determined by classical oxidation-reduction titration. size of pellets was measured by calibrated sieves. their analytical compositions were analyzed utilizing x-ray fluorescence spectroscopy. sulfur and carbon contents were also determined using combustion sulfur-carbon analyzer. physical properties of initial pellets were investigated by standard techniques of cold compressive strength and tumbler. the obtained data was modeled applying some multivariate linear and non-linear regressions like multivariate linear regression, partial least squares and neural networks. dataset was split into training and test sets to evaluate modeling performances using both cross-validation and external prediction. the optimized model resulted to the successful prediction of reducing indexes by good correlation coefficient, more than 0.99.