Estimation of concrete materials uniaxial compressive strength using soft computing techniques.

Area of Science:

• Civil Engineering
• Materials Science
• Artificial Intelligence

Background:

• Accurate concrete strength prediction is vital for construction safety and efficiency.
• Existing studies often use diverse machine learning algorithms on various concrete types.
• A gap exists in comparing advanced deep learning models specifically for mixed-design concrete strength.

Purpose of the Study:

• To comparatively analyze Convolutional Neural Networks (CNNs), Gated Recurrent Units (GRUs), and Long Short-Term Memory (LSTM) networks.
• To identify the optimal deep learning (DL) algorithm for predicting the uniaxial compressive strength of mixed-design concrete.
• To enhance material property prediction in the construction industry.

Main Methods:

• Development and optimization of deep learning models (CNN, GRU, LSTM) using experimental data.
• Focus on mixed-design concrete datasets.
• Application of hyperparameter tuning and regularization techniques for model improvement.

Main Results:

• The CNN model demonstrated superior performance in predicting concrete uniaxial compressive strength compared to GRU and LSTM models.
• Optimized DL algorithms showed enhanced predictive accuracy.
• Hyperparameter tuning and regularization further boosted model performance.

Conclusions:

• CNNs are the optimal deep learning algorithm for predicting mixed-design concrete strength.
• This research provides practical solutions for material property prediction, potentially improving construction efficiency and quality.
• The findings contribute to reducing resource burdens in the construction sector.