A chemometric approach to evaluating the mechanical properties of castable refractory ceramics for the aluminium foundry industry

Chemometrics applies statistical and data-driven methods to design experiments and extract chemically-relevant information from multivariate datasets. In this study, chemometric modelling using the Chemometric Agile Tool (CAT) software was employed to quantify the influence of individual raw materials on the mechanical properties of an SiO2-based castable refractory designed for aluminium-industry applications. Refractory ceramics are essential in aluminium production due to their thermal stability and mechanical integrity under high-temperature compressive loads, which ensures dimensional control and longer service life during molten aluminium handling. Accordingly, the mechanical properties of twenty-two formulations were evaluated by systematically varying raw material proportions around a commercial reference mixture to generate a comprehensive experimental domain. Specimens were prepared and tested in accordance with ASTM standards C133-97 and C674-13 to evaluate compressive and three-point flexural strengths. Compressive strength testing emerged as the most reliable indicator for this material class, and a multiple linear regression (MLR) model quantified the contribution of each raw material to cold crushing strength. The MLR results explained over 90% of the variance in the completed datasets, revealing the positive influence of the cementitious portion and specific fused silica fractions, as well as the complex role of colloidal silica and boron nitride. Moreover, increased total water content showed a negative correlation with compressive strength, likely due to higher porosity and defect formation. Finally, recommendations for formulation strategies and further property evaluation and optimisation are provided, demonstrating the value of Design of Experiments (DoE) in industrial formulation development.