Artificial Neural Network Algorithms to Predict the Compressive Strength of Fly Ash-Based Geopolymers
This study explores the application of Artificial Neural Networks (ANNs), specifically Self-Organizing Maps (SOM) and Hierarchical Self-Organizing Maps (HSOM), to predict the compressive strength of geopolymer concrete (GPC), activated using sodium hydroxide (NaOH) and sodium silicate (Na2SiO3). GPC...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | en en en |
| Published: |
Semarak Ilmu Journal of Advanced Research in Applied Mechanics
2025
|
| Subjects: | |
| Online Access: | http://ir.unimas.my/id/eprint/49704/3/dsva_Teo%20Siaw%20Hui.pdf http://ir.unimas.my/id/eprint/49704/4/Thesis%20MEng_Teo%20Siaw%20Hui.pdf http://ir.unimas.my/id/eprint/49704/5/Thesis%20MEng_Teo%20Siaw%20Hui_24%20pages.pdf http://ir.unimas.my/id/eprint/49704/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study explores the application of Artificial Neural Networks (ANNs), specifically Self-Organizing Maps (SOM) and Hierarchical Self-Organizing Maps (HSOM), to predict the compressive strength of geopolymer concrete (GPC), activated using sodium hydroxide (NaOH) and sodium silicate (Na2SiO3). GPC samples were cast and cured at 80°C. A total of eighteen distinct mix proportions were prepared, and the resulting dataset was used to train the ANN models based on the SOM and HSOM architectures. The study aimed to correlate the effects of the water-to-binder ratio and the molarity of the alkaline solution with the compressive strength of GPC. SOM modeling offers a cost-effective, time-efficient alternative to conventional experimental methods, which are often limited by high material usage, time-consuming procedures, and challenges in capturing complex variable interactions. Model performance was assessed using five-fold cross-validation, yielding consistent results with low Root Mean Square Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Error (MAE), along with a coefficient of determination (R²) value of 0.7. A Hierarchical SOM (HSOM) approach was also evaluated to enhance clustering and interpretability. Comparative analysis confirmed the effectiveness of both SOM and HSOM in accurately predicting GPC strength. Despite promising outcomes, challenges remain due to limited data diversity and the need to improve the practical interpretability of machine learning outputs. Overall, the findings highlight the potential of SOM, HSOM, and other AI-based techniques as powerful tools for optimizing GPC mix designs and advancing sustainable construction materials.
Keywords: Geopolymer concrete (GPC); compressive strength; Artificial Neural Network (ANN); Self Organizing Map (SOM); Hierarchical SOM (HSOM) |
|---|