Assessing chloride resistance and electrical conductivity of glass powder modified sustainable self-compacting concrete using hybrid predictive intelligence methods

This research introduces a Somersaulting Spider Optimized TabNet Feature Tokenizer Transformer (SSO-TabNet-FTTransformer) based structure for predicting the behavior of manufactured sand (M-sand) for self-compacting concrete (SCC) and waste glass powder (GP), along with carbon fibre as partial replacements for river sand and cement, respectively. To propose the SSO-TabNet-FTTransformer system, experimental and validated literature data were used. Experimentally, 51 cubes and 13 prism specimens of various SCC mix design grades, viz. M25, M30, and M40 were cast and tested to verify workability. The proposed SSO-TabNet-FTTransformer system was used to predict the highly significant fresh mechanical durability and materials optimization of SCC mixes. Computed results revealed that the workability, split tensile strength, Water penetration depth, flexural strength, and the Rapid Chloride Permeability Test (RCPT) values of SCC were accurately predicted. The SSO-TabNet-FTTransformer achieved 96.7% accuracy and was further cross-validated in MATLAB, yielding the lowest prediction error. Mineralogy studies, viz. X-ray diffraction (XRD), Fourier transform infrared spectroscopy (FTIR), and scanning electron microscopy (SEM) were used to examine the characteristics of calcium silicate hydrate (C-S-H) gel formation, pore connectivity, and the dense microstructure of GP-modified concrete in comparison with the conventional model. Based on experimental investigations, the proposed SSO-TabNet-FTTransformer system is established as a reliable, eco-friendly, highly accurate, sustainable, and intelligent system for high-performance SCC mix optimization and durability analysis. The proposed model achieved a desirability of 0.98.