Prediction of the Strength of the Concrete-Filled Tubular Steel Columns Using the Artificial Intelligence

Modern Trends in Construction Urban and Territorial Planning, Год журнала: 2024, Номер 3(3), С. 40 - 48

Опубликована: Окт. 2, 2024

Introduction . The machine learning algorithms are highly promising for predicting the load-bearing capacity of building structures. paper aims at predictive models calculating strength concrete-filled steel tubular (CFST) columns to enable a accurate prediction ultimate loads entire possible range parameters affecting eccentrically compressed columns. Materials and Methods article studies short circular cross-section. Model input parameters: column outer diameter, pipe wall thickness, yield steel, compressive concrete, relative eccentricity. Output without taking into account random eccentricities. were trained on synthetic data generated based theoretical principles limit equilibrium method. Two built. When training first model, determined given eccentricity longitudinal force additional second was taken account. effect features model predictions assessed using Feature Importance function. Optuna method used select hyperparameters. implemented in Jupiter Notebook environment Gradient Boosting total volume sample 179 025 samples. Results importance most values have been determined. For both models, diameter proved be important features, which is consistent with existing experience designing such Optimisation hyperparameters Grid Search enabled getting improved results. high accuracy has ascertained by low regression metrics: MSE = 9.024; MAE 9.250; MAPE 0.004 — built eccentricity; 8.673; Discussion Conclusion developed cross-section, eccentricities, demonstrated stability prediction, they can applied assessing during design construction, will reduce time resources involved physical testing. In future, it planned expand including other materials, different cross-section geometries slenderness parameter, may improve generalization ability model.

Язык: Английский

Soft Computing for Comprehensive Concrete Strength Prediction – A Comparative Study

Journal of Soft Computing Paradigm, Год журнала: 2023, Номер 5(4), С. 417 - 432

Опубликована: Дек. 1, 2023

The evolution of concrete strength prediction methodologies has transitioned from empirical formulas based on experimental data to contemporary soft computing approaches. Initially, the mix design was reliant simple relationships between proportions and compressive strength; later, early techniques evolved include statistical models incorporating material properties, curing conditions, environmental variables. advent computational tools artificial intelligence marked a paradigm shift, with accurate crucial for influencing structural integrity, safety, cost-effectiveness in construction. article explores analytical before reviewing application approaches such as fuzzy logic, genetic algorithms, neural networks. integration these hybrid is discussed this research study by highlighting their effectiveness handling complex within parameters. A comparative analysis various methods applied non-structural elements carried out demonstrate diverse applications advantages optimizing designs, enhancing performance, contributing cost time efficiency construction processes.

Язык: Английский