Durability of clay roofing tiles: assessing the reliability of prediction models

The indirect prediction of frost resistance of construction materials has been widely investigated in recent years giving rise to several models based on pore size, open porosity, water absorption and/or capillary rise. This study is aimed at appraising these models on 13 industrially-manufactured roofing tiles with a different frost resistance determined by severe freeze/thaw testing (EN 539-2). Samples were characterised by measuring pore size distribution (MIP), pore specific surface (BET), open porosity and water absorption (ASTM C67 and C373), capillary rise (UNI 10859) and phase composition (Rietveld-XRPD). No model is able to foresee reliably the product frost resistance since all models exhibit a strong dependence on the data population and probably succeed only with a homogeneous sample in terms of both composition and manufacturing technology. Among them, the Arnott's model seems to be the most reliable to discriminate among scarcely (<100 freeze/thaw cycles) and highly frost resistant products (>250 cycles); however, a complete understanding of the excellent performance (>400 cycles) provided by some roofing tiles is still lacking. Looking at their phase composition, tiles with the best performance contain also abundant calcium-magnesium silicate phases formed during firing while products with a frost resistance lower than expected are characterised by large amounts of amorphous phase or residual mica-illite and quartz. This circumstance indicates new ways to achieve highly frost resistant products that is alternative to the conventional design imposing microstructural rearrangements (low porosity and coarse pore size) through drastic batch changes and/or firing at higher temperature.