The asphalt test result database
Predict asphalt performance from real test data
AsphaltMine collects laboratory results from European partners in one organized database and uses them to train machine learning models that reduce the need for physical testing.
Latest, January 2026
First prediction tool published
A physics based machine learning model for Marshall test results that combines a data driven approach with physical constraints for better reliability and interpretability. The methodology is meant to extend to other asphalt tests.
2,600+
Marshall test results
6
Participating countries
2
Prediction models
What you can do with AsphaltMine
Contribute your results, explore what others have shared, and use the prediction tools.
Contribute data
Enter records with a web form or upload many at once from an Excel template. You decide who can see your data.
Explore and visualize
Search records, download results, and see live statistics across tests, mixtures and years.
Predict performance
Estimate Marshall stability and flow, and bulk density and air voids, from mixture properties.
Partners from Switzerland, Austria, Belgium, Serbia, Latvia and Lithuania already contribute. Developed at the Empa Concrete and Asphalt Laboratory and funded by the Swiss National Science Foundation.
Join the AsphaltMine community
We can not do this alone. Contribute your asphalt test results and get access to the prediction tools.