ARTIFICIAL INTELLIGENCE-BASED PREDICTION OF COMPRESSIVE STRENGTH OF METAKAOLIN-SAW DUST GEOPOLYMER CONCRETE FOR SUSTAINABLE CONSTRUCTION APPLICATIONSPages 184-204
Abstract
This study examines the fresh and hardened characteristics of metakaolin-sawdust geopolymer concrete (MSGC) and develops models to predict its compressive strength. MSGC mixes were prepared with sawdust replacing fine aggregates at levels from 0% to 40%. Evaluations covered workability, setting time, bulk density, water absorption, and compressive strength, alongside artificial intelligence-based prediction. Increasing sawdust levels led to marked reductions in slump (172 mm at 0% to 0 mm at 30–40%) and substantial delays in initial setting time (53 minutes at 0% to 242 minutes at 40%). Bulk density fell from 2350 kg/m³ to 1400 kg/m³, while water absorption rose sharply from 3.5% to 25% as sawdust content increased. MSGC compressive strength decreased from 36.1 MPa (0%) to 3.8 MPa (40%) at 28 days. The control mix outperformed ordinary Portland cement concrete (OPC), and MSGC with up to 10% sawdust remained competitive (28.5 MPa). Predictive models developed using Artificial Neural Networks (ANN), Adaptive Neuro-Fuzzy Inference Systems (ANFIS), and Gene Expression Programming (GEP) showed the ANN model provided best accuracy, with R² = 0.9423. Overall, findings confirm MSGC’s potential as a sustainable alternative for construction.
Keywords:
Workability,
ANN,
OPC,
Sawdust,
Slump.
|
Hyginus Obinna Ozioko, 
PDF 