Controlling nano-silica cement rheology with a chitosan bio-surfactant using a gaussian process regression-bayesian desirability framework

This study presents an integrated experimental and computational framework for controlling the rheology and mechanical performance of cementitious systems modified with chitosan, a biodegradable bio-surfactant, and nano-silica (NS). Fourteen mixtures containing 0–1.0 % chitosan and 0–2.0 % NS were evaluated through setting time measurements, steady and oscillatory rheology, viscosity profiling, compressive strength testing, and mercury intrusion porosimetry (MIP). The combined modification produced pronounced synergistic effects: the optimal chitosan–NS formulation (1.0 % chitosan + 2.0 % NS) increased 28-day compressive strength from 45.1 MPa to 69.5 MPa (+54 %), tripled the static yield stress from 500 Pa to 1500 Pa, and reduced total porosity from 15.5 % to 8.0 %. To identify optimum proportions, Gaussian Process Regression (GPR) surrogate models were developed for static yield stress, high-shear stress (τ₁₀₀), and compressive strength, and integrated into an uncertainty-aware multi-objective desirability framework. This approach enabled statistically robust optimization despite the limited dataset, with experimental results validating the predicted optimum. The study demonstrates that dual modification using chitosan and nano-silica offers a sustainable and tunable route for enhancing fresh-state structuration, microstructural densification, and mechanical performance, supported by a predictive optimization methodology for mix design.