Predictive Modeling of Compressive Strength and Slump in High-Performance Concrete Utilizing Machine Learning
Compressive strength is important in HPC, for it actually reflects the ability to bear stresses without disintegration. It ensures structural stability and durability and, hence, resistance to various types of external loads, which is critical for infrastructure serviceability over a long period of...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Electronic Journals for Science and Engineering - International
2025-06-01
|
| Series: | Electronic Journal of Structural Engineering |
| Subjects: | |
| Online Access: | http://10.0.0.97/EJSE/article/view/713 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Compressive strength is important in HPC, for it actually reflects the ability to bear stresses without disintegration. It ensures structural stability and durability and, hence, resistance to various types of external loads, which is critical for infrastructure serviceability over a long period of time. Whereas slump is indicative of the uniformity and workability of HPC, it affects the ease of placing and consolidation, and construction quality and efficiency. Mix design optimization, through proper balancing between compressive strength and slump, will enhance the capability of HPC to meet the stringent operational standard for heavy applications like bridges, high-rise buildings, and nuclear facilities concerning safety and longevity with cost-effectiveness while constructing the projects. The research estimates the compressive strength and slump of the HPC by advanced machine learning regression frameworks such as ADAR, SVR, and three optimizers: GOA and CBOA. Combining these frameworks with an optimizer results in a novel hybrid framework that not only gives enhanced precision and functionality. Results show that the ADGA model is also optimum in the CS target during training, with an RMSE of 2.451 and a value of 0.992 for R2, followed by the ADCB model outperforming the base model ADA with an RMSE of 3.618 and 0.982 for R2. Notably, the SVR model and its hybrid forms exhibit poorer operation compared to the ADA model in the CS target. These results then emphasize the capability of hybrid ML frameworks in predicting the characteristics of concrete with a good degree of accuracy.
|
|---|---|
| ISSN: | 1443-9255 |