Chloride Permeability Coefficient Prediction of Rubber Concrete Based on the Improved Machine Learning Technical:...

Area of Science:

• Materials Science
• Civil Engineering
• Computational Intelligence

Background:

• Rubber concrete (RC) offers improved resistance to chloride ion attacks, crucial for coastal infrastructure.
• Accurate prediction of the chloride permeability coefficient (D) in RC is vital for its widespread application.
• Traditional methods for determining D in RC lack multi-factorial consideration and exhibit low prediction accuracy.

Purpose of the Study:

• To develop an optimized machine learning approach for accurately predicting the chloride permeability coefficient (D) of rubber concrete (RC).
• To enhance the performance of machine learning models by optimizing their parameters using a novel algorithm.
• To provide a more reliable and efficient method for assessing RC durability in challenging environments.

Main Methods:

• A novel Mixed Whale Optimization Algorithm (MWOA) was developed, incorporating Tent mapping and adaptive t-distribution variations to prevent local optima.
• Three machine learning models—Extreme Learning Machine (ELM), Random Forest (RF), and Elman Neural Network (ELMAN)—were trained using literature data.
• The MWOA was employed to optimize the parameters of the ELM, RF, and ELMAN models for predicting the chloride permeability coefficient (D) of RC.

Main Results:

• Optimized machine learning models using MWOA demonstrated significant improvements in prediction accuracy: ELM by 54.4%, RF by 62.9%, and ELMAN by 36.4%.
• The MWOA-ELM model emerged as the superior performer among the optimized models.
• The MWOA-optimized ML models significantly outperformed traditional multiple linear regression (87.15%) and mathematical models (85.03%) in accuracy.

Conclusions:

• The MWOA effectively optimizes machine learning models, substantially enhancing their predictive accuracy for the chloride permeability coefficient (D) of rubber concrete.
• The MWOA-ELM model presents a highly accurate and reliable tool for predicting RC's chloride permeability.
• This optimized machine learning approach offers a superior alternative to traditional methods for evaluating rubber concrete durability, particularly in coastal applications.