Performance of hybrid ANFIS and numerical simulation to predict the effect of steel fibre on compressive strength of concrete

Abstract

The growing demand for high-performance concrete with enhanced mechanical properties calls for innovative mix designs and accurate predictive tools. This study uses varying formulations to investigate the mechanical performance, particularly compressive strength, of high-performance steel fiber-reinforced concrete (HP-SFRC). Key parameters include water-to-binder ratios (w/b) of 0.35, 0.40, and 0.45; silica fume as a partial cement replacement at 10% and 15%; and steel fiber volume fractions (Vf) of 0%, 0.5%, 1.0%, and 1.5%, with fiber aspect ratios of 80 and 40. Results demonstrate that incorporating silica fume and steel fibers enhances compressive strength, with the most notable gains at Vf = 1.5% attributed to improved stress transfer within the matrix. An adaptive neuro-fuzzy inference system (ANFIS) model was developed and trained on experimental data to improve prediction accuracy. The ANFIS model exhibited superior performance over conventional models, providing more accurate and reliable predictions. Multiple linear regression (MLR) models were also evaluated for predictive ability. While MLR models offered reasonable estimates, they were consistently outperformed by the ANFIS model. This study contributes to optimizing HP-SFRC formulations by highlighting the synergistic effects of silica fume and steel fibers on compressive strength. It underscores the potential of machine learning tools like ANFIS in predictive modeling for advanced construction materials.