Experimental research and artificial neural network-based prediction model on compressive strength of hardened cement pastes containing fly ash and silica fume at high temperatures

Abstract

The composition of the binder mixture affecting the workability of concrete at high temperatures was studied on a binder mixture of cement, fly ash, and silica fume. In this work, active mineral additives (fly ash and silica fume) were used to partially replace cement with a total substitution content of 20÷50% (fixed silica fume content with 5% and 15%). The binder samples were determined for compressive strength after being heat treated at different temperature levels from 25 to 1000 °C for 2h. The study showed that binder samples with a binder mixture composition containing 15% fly ash and 5% silica fume obtained the highest strength in the temperature range of 200÷800 °C compared to PC control samples and samples with other ratios. The samples significantly attenuated the compressive strength when exposed to more than 800 °C. In addition, this study constructed two predictive models using artificial neural networks to forecast the compressive strength based on input parameters including data on fly ash, silica fume and temperature (1); to predict the contents of fly ash and silica fume based on input parameters of compressive strength and working temperature (2). The models were built on the practical data showing the good fitting.