Sustainable enhancement of mechanical properties in fly oil shale ash-based materials through AI-driven optimization

Abstract

The average compressive strength of raw and modified FOSA-based mortar was estimated utilizing the Response Surface Methodology (RSM) and Artificial Neural Network (ANN) models. Four critical factors were considered: the duration of the calcination, the temperature of the calcination, the percentage of replacement and the age of the sample. The results demonstrated that an increase in the percentage of FOSA replacement corresponded to a reduction in the strength of the raw material samples, whereas the substitution of treated FOSA resulted in an increase in strength. The ANN model showed superior performance, exhibiting a notable R correlation coefficient of 0.95698 and a mean square error (MSE) of 0.0038, thus underscoring its accuracy compared to the RSM model. A total of 477 cubes from 53 mortar mixes underwent testing at varying curing ages, specifically at 7, 28, and 56 days, with nine specimens for each mix. These comprehensive experiments and analyses substantiate the superiority of ANN model in predicting the average compressive strength for FOSA-based cement mortar.