Durability Characterization and Prediction of Bio-Cement Concrete Using RCPT and Machine Learning Approaches

Abstract

Concrete durability is significantly compromised by chloride ingress, which accelerates reinforcement corrosion and reduces service life. This study investigates the use of bio-cement incorporation (1%) to improve the resistance of M25, M40, and M60 concretes against chloride penetration at 28, 56, and 90 days, assessed through the Rapid Chloride Permeability Test (RCPT). In addition to experimental evaluation, soft computing approaches were applied for predictive modeling and optimization. Response Surface Methodology (RSM) provided statistical adequacy with an R² of 0.9999, adjusted R² of 0.9983, predicted R² of 0.9207, and adequate precision of 82.77, confirming excellent model reliability. Artificial Neural Network (ANN) analysis further validated these findings, yielding regression slopes close to unity and residual distributions within acceptable limits, indicating superior nonlinear predictive capability. The combined RSM–ANN framework not only ensures accurate prediction of RCPT behavior but also establishes a robust optimization strategy for bio-cement concretes. The results confirm that bio-cement, supported by advanced computational modeling, is a promising and sustainable solution to enhance chloride resistance and expand the service life of structures made by concrete.