Data-Driven Parameter Selection and Modeling for Concrete Carbonation | JoVE Visualize

Area of Science:

Civil Engineering
Materials Science
Environmental Science

Background:

Concrete carbonation is a stochastic process influenced by unconsidered parameters, leading to prediction uncertainties.
Accurate prediction of carbonation depth is vital for assessing concrete structure durability and service life.
Identifying and selecting critical influence factors is essential for robust carbonation prediction models.

Purpose of the Study:

To identify and quantitatively evaluate key parameters influencing concrete carbonation depth and its uncertainties.
To assess the impact of parameter selection on the performance of carbonation prediction models.
To develop and validate a simplified, accurate machine learning model for concrete carbonation depth prediction.

Main Methods:

Utilized statistical methods and machine learning techniques on 8204 datasets to analyze parameter correlations and influence.
Evaluated single parameters and parameter groups based on their correlation with carbonation depth and uncertainty reduction.
Developed and simplified machine learning models to propose a practical, concise prediction tool.

Main Results:

Compressive strength demonstrated the highest correlation with concrete carbonation depth.
The aggregate-cement ratio significantly reduced the dispersion of carbonation depth predictions.
Selected parameter groups enhanced machine learning model performance, enabling greater accuracy with fewer variables.
The developed practical model achieved high accuracy on both accelerated and natural carbonation datasets, with a mean absolute error of 1.56 mm for natural carbonation.

Conclusions:

Parameter selection is critical for reducing uncertainties in concrete carbonation prediction.
The aggregate-cement ratio is a highly effective parameter for improving prediction accuracy and reducing uncertainty.
A concise, practical model derived from machine learning shows significant potential for accurate concrete carbonation depth prediction in real-world applications.