Optimizing the utilization of Metakaolin in pre-cured geopolymer concrete using ensemble and symbolic regressions

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Фев. 26, 2025

The optimization of metakaolin (MK) in pre-cured geopolymer concrete involves developing predictive models to capture the interplay various influencing factors and guide mix design for improved compressive strength sustainability. Ensemble methods symbolic regression are promising approaches this task due their complementary strengths solving challenges associated with repeated experiments laboratory. Choosing machine learning predictions over repeated, expensive, time-consuming research projects, such as optimizing utilization concrete, presents a paradigm shift how data-driven insights can revolutionize material development. integration ensemble enables researchers derive valuable optimize critical performance parameters efficiently. In work, 235 records were collected from extensive literature search different mixing ratios metakaolin-based at ages. Each record contains MK: content (kg/m3), SHS: Sodium hydroxide solution SHSM: molarity (Mole), SSS: silicate W: Extra water (not including alkaline solutions) W/S: Water Solid ratio (Total / part activator solutions + MK), Na2O/Al2O3: oxide aluminium ratio, SiO2/Al2O3: Silicon H2O/Na2O: CA/FA: Coarse Fine aggregate CAg: coarse aggregates SP: super-plasticizer PCC: 0 no pre-curing, 1 pre-curing 60 °C, 2 80 CT: Curing temperature (°C), Age: age testing (days) CS: Compressive (MPa). portioned into training set (180 records≈75%) validation (55 records≈ 25%) modeled methods. At end model metrics used evaluate models' ability Hoffman Gardener's sensitivity analysis was impact variables on mixed metakaolin. GB KNN became decisive excellent which outclassed others indicated that SHSM, SSS, W/S, Na2O/Al2O3 most influential predicted strength.

Язык: Английский

Modeling the compressive strength behavior of concrete reinforced with basalt fiber

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 3, 2025

This research investigates the compressive strength behavior of basalt fiber-reinforced concrete (BFRC) using machine learning models to optimize predictions and enhance its practical applications. The study incorporates various modeling techniques, including Artificial Neural Networks (ANN), k-Nearest Neighbors (KNN), Support Vector Machines (SVM), Decision Trees, Random Forest (RF), evaluate their predictive capabilities. Basalt Fiber Reinforced Concrete is a composite material that fibers into traditional mechanical durability properties. use fibers, derived from natural volcanic rocks, aligns with sustainability goals due eco-friendliness, cost-effectiveness, high performance. BFRC combines structural excellence sustainability, making it an ideal for modern construction practices. Its ability performance, reduce environmental impact, ensure long-term positions as pivotal solution sustainable infrastructure development. developed were used predict fiber (Cs_bf) mixture contents, age, dimensions. All created "Orange Data Mining" software version 3.36. A total three hundred nine (309) records collected literature different mixing ratios at ages. Each record contains following data: C-Cement content (Kg/m3), FA-Fly ash W-Water SP-Super-plasticizer CAg-Coarse aggregates FAg-Fine Age-The age testing (days), L_b-length (mm), d_bf-Diameter (µm), V_bf-Volume (%) Cs_bf-Compressive fibre (MPa). divided training set (249 records≈80%) validation (60 records≈ 20%). At end process, can be shown present work outclassed other ML techniques applied in previous paper, which reported utilization same size data entries reinforced constituents. Taylor chart measured predicted ANN, KNN, SVM, Tree RF presented comparing performance by illustrating key statistical measures simultaneously: correlation coefficient (R), normalized standard deviation (σ), root-mean-square error (RMSE). Finally, deduced after considering indices selected ensemble classification utilized this all modes have almost excellent level accuracy 95%, but SVR produced R2 0.98 each KNN producing MAE 1.4 MPa, MSE 2.5 MPa outperform ANN 1.55 MPa/MSE 4.1 1.6 3.85 respectively. Three estimate impact input on strength, namely matrix, sensitivity analysis relative importance chart.

Язык: Английский

Advanced predictive techniques for estimating compressive strength in recycled aggregate concrete: exploring interaction, quadratic models, ANN, and M5P across strength classes

Multiscale and Multidisciplinary Modeling Experiments and Design, Год журнала: 2024, Номер 8(1)

Опубликована: Дек. 29, 2024

Язык: Английский