MIX PROPORTION OF FLY ASH AND SILICA FUME MORTAR USING THE ANN-GRADIENT DESCENT MODEL FOR PREDICTING COMPRESSIVE STRENGTH

Abstract

This study develops an Artificial Neural Network with Gradient Descent (ANN-G) model to predict the compressive strength of geopolymer mortar based on fly ash and silica fume mix proportions. Accurate prediction of compressive strength is essential for optimizing geopolymer mixes and promoting sustainable construction practices. The research employs the Taguchi experimental design to optimize the geopolymer mix for target strengths of 30 MPa, 35 MPa, and 40 MPa. The ANN-G model predicts compressive strengths of 26.92 MPa, 35.15 MPa, and 40.35 MPa, demonstrating its accuracy and efficiency. Results show that the ANN-G model outperforms conventional ANN models by reducing prediction errors and improving reliability. This approach streamlines the mix design process, reduces the need for extensive experimental testing, and enhances prediction accuracy. The ANN-G model offers a practical tool for designing geopolymer mortars in construction. Future work should focus on integrating larger datasets and exploring hybrid models to improve prediction stability and extend the model’s applicability in real-world construction scenarios.