Eco-friendly mortars with granite powder and fly ash and their prediction with artificial neural networks

Abstract

This paper focuses on mortar in which cement is partially replaced with siliceous fly ash and granite powder and subjected to abrasion. To optimise the process of testing abrasion resistance, artificial neural network (ANN) was proposed to predict results. Predicting abrasion resistance of mortar, especially with addition of wastes is not widely explored and requires greater academic attention. The possibility of applying machine learning is reducing amount of destructive tests methods in favour of prediction, therefore sustainable approach. This study confirms that it is possible to create eco-friendly mortar with siliceous fly ash and / or granite powder and that these materials do not deteriorate the surface of mortar samples. The most similar results of abrasion resistance to reference samples with only Portland cement were obtained for combination of fly ash and granite powder. The application of ANN to predict abrasion resistance of tested mortars showed very accurate results with the network's worst quality at 0.967 (with 1.0 as ideal result). The quality of the network relies on properly selected and detailed inputs. This study proved satisfactory results of predicting abrasion resistance based on mortar's components only.