Factor space differentiation of brick clays according to mineral content: Prediction of final brick product quality

• 139 heavy clay mineral content are calculated using LPNORM.
• The samples are divided using PCA into 4 groups concerning mineral contents.
• The mathematical models are developed for each of the obtained groups.
• Statistical models are used for brick quality prediction.

Chemical composition and XRD qualitative analysis were used to calculate mineral contents of 139 brick clay raw materials using LPNORM. The second order polynomial models (SOP) for all the samples, which express the relation between mineral contents and the characteristics of fired laboratory products, did not fit to experimental data satisfactorily, due to low coefficients of determination (r2). In order to improve the models, the samples are divided into four groups in factor space (four quadrants), according to their mineral content similarity, using principal component analysis (PCA). Predictive models of compressive strength (CS), water absorption (WA), firing shrinkage (FS), weight loss during firing (WLF) and volume mass of cubes (VMC) are obtained for each of the groups. Second order polynomial (SOP) models are developed, and the influence of certain minerals to brick clay bricks quality within the groups is discussed. Developed models were able to predict the final quality of products in a wide range of mineral content and temperature treatment data, showing coefficient of determination (r2) in range between 0.704–0.995. In order to estimate the adequacy of these models, the results were applied to the experimental data and compared according to additional statistical tests, so the next values are determined: coefficients of determination, reduced chi-square (χ2), mean bias error (MBE), mean percent error (MPE) and root mean square error (RMSE).