Machine Learning Method to Explore the Correlation between Fly Ash Content and Chloride Resistance.

Area of Science:

• Materials Science
• Civil Engineering
• Environmental Science

Background:

• Chloride ion corrosion is a primary cause of reinforced concrete structure degradation in marine and deicing salt environments.
• Understanding factors influencing concrete durability against chloride attack is crucial for infrastructure longevity.

Purpose of the Study:

• To investigate the relationship between fly ash content and concrete durability, specifically its resistance to chloride ion erosion.
• To identify the optimal fly ash admixture ratio for mitigating chloride-induced corrosion in concrete.

Main Methods:

• Analysis of factor correlations for chloride ion content, highlighting sampling depth, cement dosage, and fly ash dosage.
• Application of three machine learning algorithms (Random Forest, Gradient Boosting Regression, Decision Tree) to predict total chloride content in fly ash concrete.
• Evaluation of machine learning model performance using R², MSE, RMSE, and MAE metrics.

Main Results:

• Factor correlation analysis indicated sampling depth > cement dosage > fly ash dosage.
• The Random Forest model predicted an optimal fly ash admixture threshold of 30-40% for chloride resistance.
• Replacing cement with fly ash below this threshold alters concrete's phase and pore structure, improving permeability and reducing free chloride ions.

Conclusions:

• Fly ash concrete exhibits enhanced resistance to chloride ion erosion within the 30-40% admixture range.
• Machine learning models accurately predict concrete properties, reducing the need for extensive physical testing.
• Developing machine learning for concrete property prediction is vital for engineering decarbonization and intelligent infrastructure.