Predicting propellant heat flow calorimetry time series with machine learning
conference paper
We develop several machine learning models to predict Heat Flow Calorimetry (HFC) measurements of propellants. We train these models on an extensive HFC database built at TNO for over fifty years, containing thousands of unique propellants. Surprisingly, we find that most machine learning models do not outperform simple averaging of historical measurements. However, a convolutional neural network that makes slight adjustments to historical averages does outperform this baseline, indicating that there are hidden trends in heat flow calorimetry that may be utilized more effectively by more powerful models.
TNO Identifier
1015281
Source title
Artificial Intelligence, Machine Learning and Data Science in Energetic Materials Research. 54th International Annual Conference of the Fraunhofer ICT combined with Symposium "Advances in Metal Fuels and Reactive Mateterials Science and Technology", Karlsruhe, Germany, 24-27 June 2025
Collation
6 p.
Pages
Paper A26
Files
To receive the publication files, please send an e-mail request to TNO Repository.