Research
Machine learning for ultra-high-performance concrete
Kasetsart University, Thailand · IAESTE research internship · 2026
Supervised by Dr. Vatwong Greepala
As a research intern, I handled the technical side of a machine learning research programme on ultra-high-performance concrete, building the pipeline behind a planned series of four international journal papers.
- Machine learning pipeline
- Group-aware benchmarking
- Conformal prediction
- SHAP
About the research
Ultra-high-performance concrete (UHPC) is an advanced cement-based material known for exceptional strength and durability. Its mechanical properties depend on many interacting mix ingredients, which makes them difficult to predict without costly laboratory testing. Machine learning can learn these relationships from experimental data.
The research benchmarks machine learning models for predicting UHPC's mechanical properties in a group-aware way, so that evaluation reflects how well models generalise to genuinely new data. It also uses conformal prediction to give every prediction a range with a guaranteed coverage rate, rather than a single number.
My role: I handled the technical side end to end, the machine learning work and the pipeline behind it.
Papers
- ML01Submitted
Group-Aware Machine Learning Benchmark for Mechanical Properties of Ultra-High-Performance Concrete with Conformal Uncertainty Quantification
- ML02In progress
Title to be announced on publication
- ML03In progress
Title to be announced on publication
- ML04In preparation
Title to be announced on publication
Papers will be available here once published.