University of Tehran Civil Engineering Infrastructures Journal 2322-2093 59 1 2026 06 01 Thermal Performance Prediction for Alkali-Activated Concrete Using GGBFS, NaOH and Sodium Silicate 241 253 98949 10.22059/ceij.2024.369661.1996 EN Pramod Kumar Assistant Professor, Department of Civil Engineering, Mohan Babu University (SVEC), Tirupati, Andhra Pradesh, India. 0000-0002-1411-9374 Sanjay Sharma Research Scholar, Department of Civil Engineering, National Institute of Technology, Jamshedpur, Jharkhand, India. 0009-0005-3129-6080 P Siva Kumar Assistant Professor, Department of Civil Engineering, Mohan Babu University (SVEC), Tirupati, Andhra Pradesh, India. 0009-0008-9864-3165 M. S. Yuvaraj Assistant Professor, Department of Civil Engineering, Mohan Babu University (SVEC), Tirupati, Andhra Pradesh, India. 0000-0003-4431-053X D V Purushotham Assistant Professor, Department of Civil Engineering, Mohan Babu University (SVEC), Tirupati, Andhra Pradesh, India. 0009-0008-9864-3165 Saurabh Kumar Assistant Professor, Department of Civil Engineering, IIMT University, Meerut, Uttar Pradesh, India. 0000-0002-1044-5658 Kapil Kumar Vashistha Assistant Professor, Department of Civil Engineering, IIMT University, Meerut, Uttar Pradesh, India. 0000-0002-1044-5658 Amit Kumar Assistant Professor, Department of Civil Engineering, IIMT University, Meerut, Uttar Pradesh, India. 0000-0002-8475-2538 Abhilash Gogineni Research Scholar, Department of Civil Engineering, National Institute of Technology, Jamshedpur, Jharkhand, India. Journal Article 2023 12 15 In fire safety, understanding the behaviour of concrete exposed to high temperatures is essential. This study experimentally explored the mechanical properties of Alkali-Activated Concrete (AAC) and utilized Recurrent Neural Network (RNN)-based Long Short-Term Memory (LSTM) techniques to predict the mechanical properties of AAC at elevated temperatures. The LSTM models accurately predicted compressive, flexural, and split tensile strengths, with coefficients of determination (R²) exceeding 0.9 for training and testing datasets. Specifically, R² values were 0.9838 and 0.9134 for compressive strength, 0.9965 and 0.9861 for flexural strength, and 0.9743 and 0.9852 for split tensile strength in training and testing, respectively. The models also yielded low Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values, further underscoring their predictive reliability. Error analysis across all mechanical properties affirmed the LSTM models' robustness in capturing AAC's complex behaviour under thermal stress. These results suggest that LSTM networks are highly effective tools for predicting material properties crucial for structural fire safety and sustainable construction, offering a promising approach for improving the resilience and safety of AAC structures in extreme conditions. https://ceij.ut.ac.ir/article_98949_562b7248bf501ee9275924a11569aefd.pdf