In this paper, 871 data were collected from literature and trained by the 4 representative machine learning methods, in order to build a robust compressive strength predictive model for slag and fly ash based alkali activated concretes. The optimum models of each machine learning method were verified by 4 validation metrics and further compared with an empirical formula and experimental results. Besides, a literature study was carried out to investigate the connection between compressive strength and other mechanical characteristics. As a result, the gradient boosting regression trees model and several predictive formulas were eventually proposed for the prediction of the mechanical behavior including compressive strength, elastic modulus, splitting tensile strength, flexural strength, and Poisson's ratio of BFS/FA-AACs. The importance index of each parameter on the strength of BFS/FA-AACs was elaborated as well.

A two-dimensional (2D) image-based methodology was proposed to measure the specific surface area (SSA, specified as the surface area per unit volume) of irregular aggregate by random sectioning. Conventional methods including spherical assumption, Brunauer-Emmett-Teller (BET) and computed tomography (CT) tests were used and compared in this study. Results show that spherical assumption provides the lowest SSA among these methods since the feature of anisotropy in dimension is not considered. SSA by BET test has one order of magnitude higher value than others, which is attributed to the fact that BET method measures each position of particles that nitrogen molecule can be adsorbed on during the applied relative pressure, based on the ‘pixel’ of nitrogen molecule. The proposed random sectioning method presents very similar SSA result compared to CT method, indicating that it can be considered as a reliable method. To improve the estimation of SSA by random sectioning method, factors that may influence SSA result were analyzed. Results indicate that the number of samples should be high enough to reach a constant result and the thresholding algorithm should be adequate. Besides, a higher resolution of pixel provides a higher SSA value. The comparison among these methods demonstrate that it is necessary to determine the scale at which the features of the surface are supposed to be captured before selecting the optimal testing method.