Determining equivalent rectangular stress block parameters in geopolymer concrete using multi-gene genetic algorithm-supported models and ML techniques


ÖZBAYRAK A.

Structures, vol.73, 2025 (SCI-Expanded) identifier

  • Publication Type: Article / Article
  • Volume: 73
  • Publication Date: 2025
  • Doi Number: 10.1016/j.istruc.2025.108323
  • Journal Name: Structures
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Keywords: Artificial intelligence, Balanced reinforcement ratio, Equivalent stress block parameters, Geopolymer concrete, Varying activator solution ratios
  • Erciyes University Affiliated: Yes

Abstract

Extensive research is currently being conducted in various fields regarding geopolymer concrete. New information about this concrete's advantages and disadvantages is being obtained daily. However, for this concrete to be widely used in practice, the applicability of traditional reinforced concrete design principles to geopolymer concrete must be investigated. This study calculated equivalent rectangular stress block parameters for geopolymer concrete mixtures with different activator solution ratios and balanced reinforcement ratios. Experimental, analytical, and multi-gene genetic algorithm-supported models were used to plot stress-strain curves. The calculated stress block parameters from the curves were evaluated according to ACI318 and TS500 specifications. It was found that geopolymers with lower elastic modulus exhibited less deformation and sudden failure after peak load compared to traditional concrete, significantly affecting the stress block parameters. A t-Test was performed between the experimental data and ACI318 specifications for the k1 parameter and balanced reinforcement ratio, revealing a significant correlation and similarity. For the k3 parameter, this relationship can be observed under a compressive strength of 40 MPa. However, analytically calculated k3 parameters were found to be higher than those specified in ACI318, as reported in some studies. While the balanced reinforcement ratio was lower than the ACI318 code according to experimental data, analytical and genetic algorithm data yielded much higher values. The reasons for these different findings were discussed, and it was deemed necessary further to investigate these parameters through structural column and beam experiments.