Characterisation of the shape of aggregates using image analysis

Aggregates are packed effectively packed in applications including pervious concrete, mechanical performance of which, depends primarily on packing characteristics. Size and shape distribution of aggregates affect packing. Several computational methods are used to represent shape of aggregates numerically, from analyses of digital images, yet, they have not been compared. This study aims to analyse representability of shape aspects of aggregates from different computational methods. Crushed rock-aggregates were grouped into five clusters and milled in LAAV machine for different number of revolutions (0–2000) to induce morphological changes, Samples were then sieved and aggregates from 5 to 30 mm in diameter were obtained (7191 in total). Aggregates were painted in black oil paint, laid on white sheets and digital images were obtained and analysed using an open source software ImageJTM while shapes were defined based on 14 computational methods. Statistical tests Pearson's Correlations, Principal Components Analysis and K-means Cluster analysis were used to assess shape factors. In conclusion, no shape factor singularly represented the morphological changes on aggregate particles. A combination of shape factors is required. The data matrix had three dimensions and three shape factors Circularity, Kumbrein solidity and Barksdale shape factor optimally represented aggregate shape index.