Development of Time Series Models and Algorithms: Creep Prediction for Low-Carbon Concrete Materials

Area of Science:

• Civil Engineering
• Materials Science
• Computational Science

Background:

• Low-carbon concrete is crucial for sustainable development, but its creep behavior requires accurate prediction.
• Traditional creep characterization methods are slow and do not fully capture multi-factor interactions.
• Accurate creep prediction is vital for the reliable application of sustainable concrete materials.

Purpose of the Study:

• To develop and validate a novel temporal modeling framework for predicting creep in low-carbon concrete.
• To enhance prediction accuracy by optimizing Artificial Neural Network (ANN), Random Forest (RF), and Long Short-Term Memory (LSTM) models using an advanced algorithm.
• To identify the most effective model for creep prediction in low-carbon concrete materials.

Main Methods:

• Construction of a time-series database for low-carbon concrete behavioral characteristics.
• Retraining of ANN, RF, and LSTM models for creep prediction.
• Implementation and validation of an enhanced Adaptive Crowned Porcupine Optimization algorithm (ACCPO) for model optimization.

Main Results:

• ACCPO significantly improved the performance of ANN, RF, and LSTM models.
• Single-metric accuracies reached 95.9% (ANN), 93.9% (RF), and 97.8% (LSTM) post-optimization.
• Error reductions of 22.6% (ANN), 7.9% (RF), and 8% (LSTM) were achieved, confirming model effectiveness.

Conclusions:

• The proposed temporal modeling framework is rational and effective for creep prediction.
• The ACCPO algorithm demonstrates significant improvements in optimizing predictive models.
• The ACCPO-LSTM time series model is identified as the superior choice for low-carbon concrete creep prediction.