(210h) Multifunctional Magnesium Oxide Cement Composites for Building Products

Magnesium oxychloride (MOC) cement is recognized as a transformative building material due to excellent compatibility with a variety of filler materials and its superior mechanical, thermal, and abrasive properties relative to ordinary Portland cement. The sensitivity of MOC’s properties to the stoichiometry of the initial ternary mixture of raw materials presents a significant challenge in predicting its final performance. This complexity is further heightened when considering an MOC composite, where the incorporation of various fillers, commonly utilized in commercial applications, brings an additional layer of variability to the production properties. Consequently, there is a demand for advanced predictive models to ensure optimized utilization and reliable performance in commercial applications. In this work, Gaussian process (GP) regression is used to model the effects of process control parameters, formulation, and composition on multiple mechanical performance metrics for fiber-reinforced MOC boards. Various summation and product combinations of squared-exponential kernels were compared using cross-validation and the final performance of each model was evaluated using held-out validation test datasets. The output variables consisted of flexural strength, modulus of elasticity, brittleness index, and lateral nail resistance. Based on the validation results, the proposed kernels performed exceptionally well at predicting lateral nails resistance properties and modulus of elasticity, reasonably well flexible strength properties, and the worst at brittleness index.