Analysis of ice bending experiments with statistics and machine learning
Prior to modeling, decisions on which and how to include phenomena and parameters in physics-based models must be made. In other cases, the physics-based model is ready to use, but no measurements are available for necessary input parameters.
An example for both issues is the bending strength of ice. It is not itself a material property, but it reflects the maximum sustainable stress. It is critical for the performance of ice breaking ships or the break-up of ice due to waves. Moreover, it is seldom known beforehand.
One remedy is to analyze correlation of parameters with outcome in experiments. This helps ranking influential parameters as well as predict bending strength, if it is not known. Machine learning and statistical models can be used to establish such an importance ranking and make necessary predictions
Here, a database of ice bending experiments should be established. Statistics and machine learning tools are to be used to analyze the data. In the figure you can see how the algorithms makes predictions for a similar problem of compressive strength prediction (in MPa) based on different features. Each string represents one data point, the top bar the result/prediction.
- Python with Jupyter lab or Spyder/Anaconda. Easy to use and beginner-friendly programming language with a development environment of your choice
- MATLAB, for additional analyses and post-processing
- ANSYS, for optional, complementary FE analyses
Leon Kellner, email@example.com, Office C 4.004,
Franz von Bock und Polach, firstname.lastname@example.org, C 4.012,
Moritz Hartmann, email@example.com, Office C 4.002.
Don’t hesitate to contact us if you are interested or have any further questions.