Development of mathematically motivated artificial intelligence models for the prediction of carbonate rock lime saturation factor for cement production

Abstract

Predicting the cement lime saturation factor (LSF) is vital in cement production as it determines the optimal ratio of lime to other components. This factor affects the quality and strength of concrete, enabling efficient resource utilization, cost control, and environmentally sustainable practices in construction and infrastructure projects. The laboratory tests to determine the elemental composition of cement material is time-consuming and costly. However, limited research has suggested that there is a correlation between parameters obtained from elemental and proximate analyses of these materials. In this study, some predictive models of the elemental composition of carbonate rock using soft computing and regression analyses have been developed from chemical database from 7 formations in Nigeria. These formation includes carbonate deposit in Ikpeshi (Edo State), Ubo River area (Edo State), Itobearea (Kogi State), Igarra (Edo State), Nsofang (Ikom) Igue (Edo State), and Ewekoro (Ogun State). One hundred and forty-four (144) samples including parameters of elemental and proximate analyses were used during the analyses to develop multiple prediction models. Dependent variables for multiple prediction models were selected as carbon, hydrogen, and oxygen. Using Calcium oxide, magnesium oxide, loss of ignition, and the ratio of calcium oxide to magnesium oxide as independent variables, three different prediction models were developed for LSF as the dependent parameter using multilayer perceptron (MLP), least square support vector machines (LS-SVM), Multivariant regression (MVR) and hunter point Backpropagation-artificial neural network (HPBR-ANN) techniques. In addition, a routine for selecting the best predictive model was suggested in the study. The reliability of the established models was tested by using various prediction performance indices and the models were found to be satisfactory. Therefore, the developed models can be used to determine the cement production lime saturation factor of carbonate rock for practical purposes.