Enterprise AI Analysis
Predicting self-healing efficiency in recycled aggregate concrete using optimized machine learning models
Authors: Kunpeng Cao, Dunwen Liu, Kian Hau Kong, Wanmao Zhang, Yu Tang & Yinghua Jian
Journal: Scientific Reports | Publication Date: 2025
Unlocking Sustainable Construction with AI-Powered Self-Healing Concrete
This groundbreaking research leverages AI and machine learning to predict and optimize the self-healing efficiency of recycled aggregate concrete (RCA), offering a path to more durable, cost-effective, and environmentally friendly construction. By minimizing microcracks and extending material lifespan, this innovation addresses critical challenges in sustainable infrastructure.
0
Peak Prediction Accuracy (R²)
0
Lowest Prediction Error (RMSE)
0
Minimal Absolute Error (MAE)
0
Optimized Runtime (s)
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Decision Tree (DT)
DT predicts outcomes by generating a tree-like model through successive data splits. For regression, it minimizes Mean Squared Error (MSE) at each node. Prediction value is typically the average of target values within a leaf node. Formula 1: Prediction value at leaf node.
Random Forest (RF)
RF is an ensemble method enhancing accuracy by constructing multiple decision trees and averaging their predictions. It provides a more stable and accurate prediction for regression tasks. Formula 2: Final prediction value of RF.
Support Vector Regression (SVR)
SVR finds a balanced hyperplane that ensures prediction errors fall within a tolerance range, maximizing the margin to enhance generalization. Regression prediction is a linear or nonlinear kernel function. Formulas 3 & 4: Regression prediction and loss function.
Artificial Neural Network (ANN)
ANN mimics biological neural networks with interconnected nodes (neurons) for complex pattern recognition and regression. It learns nonlinear relationships by adjusting weights and biases. Formula 5: Forward propagation and loss function.
XGBoost
XGBoost is a decision tree-based boosting algorithm designed for computational speed and model performance. It builds strong learners by integrating multiple weak learners, minimizing an objective function that includes loss and regularization. Formulas 7, 8 & 9: XGBoost prediction and objective function.
Optimization (GS/PSO/NRBO)
These algorithms optimize hyperparameters (max_depth, eta, num_trees) for XGBoost. Grid Search (GS) systematically evaluates all combinations. Particle Swarm Optimization (PSO) is a population-based algorithm simulating bird flocking. N-Raphson Based Optimization (NRBO) uses gradient and Hessian matrix information for efficient local optimal solution. Formulas 10-15: Optimization steps.
0.9569
R² of NRBO-XGBoost
The NRBO-XGBoost model achieved the highest prediction accuracy with an R² value, significantly outperforming other models in forecasting concrete's self-healing performance.
Self-Healing Prediction Model Workflow
Data Collection & Preprocessing
→
Variable Selection & Correlation Analysis
→
Hyperparameter Optimization (NRBO-XGBoost)
→
Model Training & Validation (5-Fold CV)
→
Performance Evaluation & Comparison
→
Sensitivity Analysis (SHAP)
→
Self-Healing Performance Prediction
| Model | Advantages | Disadvantages |
|---|---|---|
| NRBO-XGBoost |
|
|
| XGBoost (Base) |
|
|
| SVR / ANN |
|
|
The Economic Advantage of RCA in Self-Healing Concrete
The study highlights that using Recycled Coarse Aggregate (RCA) significantly reduces material costs without compromising self-healing performance. For instance, RCA is on average 27 yuan cheaper per ton than crushed stone in Shanghai. This cost reduction, coupled with enhanced durability from self-healing, makes RCA concrete a highly attractive option for sustainable construction projects aiming for both economic viability and environmental benefits. RCA also provides a conducive environment for bacterial survival, potentially reducing the required bacterial dosage for bio-healing concrete.
"Combining RCA with self-healing concrete not only reduces the manufacturing cost but also enhances the survival rate of bacteria in bacterial-based self-healing concrete, thereby reducing the required bacterial dosage."
Source: Conclusion section of the paper
Advanced ROI Calculator
Estimate the potential ROI for integrating AI-powered material prediction into your enterprise operations. Adjust the parameters below to see tailored savings.
Estimated Annual Savings
$0
Hours Reclaimed Annually
0
AI Integration Roadmap
A phased approach to integrating AI-powered material prediction ensures smooth adoption and maximum impact within your enterprise.
Phase 1: Data Audit & Strategy Alignment
Assess existing material data, identify key integration points, and align AI strategy with business objectives. This phase involves defining success metrics and establishing data governance protocols.
Phase 2: Model Customization & Training
Leverage our NRBO-XGBoost framework to custom-train models on your specific material datasets (e.g., concrete mixes, aggregate types, environmental conditions). Focus on achieving optimal prediction accuracy for your unique applications.
Phase 3: Pilot Deployment & Validation
Implement the AI prediction module in a controlled pilot environment. Conduct rigorous validation against real-world material performance and iterate on model refinements based on feedback and results.
Phase 4: Full-Scale Integration & Monitoring
Integrate the validated AI model into your existing material design and quality control workflows. Establish continuous monitoring systems to track performance, detect anomalies, and ensure ongoing accuracy and efficiency.
Ready to Transform Your Enterprise?
Connect with our AI specialists to explore how optimized machine learning can revolutionize your material science and construction projects. Let's build a more sustainable future, together.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat