ANALYZING FLEXURAL STRENGTH DATA OF ICE: HOW USEFUL IS EXPLAINABLE MACHINE LEARNING?

Kurzfassung

The climate crisis results in a rapid sea ice decline, making shipping routes accessible for longer durations throughout the year and therefore increasing maritime traffic. At the same time, ice-structure interaction is known to cause damage to ships and structures. In this context, the flexural strength is a key property of the ice. It is also an important factor in the process of the formation of ice ridges which then act as obstacles to marine transit. Due to the complexity of natural materials, whose properties are often determined by many influencing variables, the development of a suitable material model for ice remains a major challenge. Many experimental studies on flexural strength have been done whose results can be used to draw conclusions about the material properties. However, a major drawback is that these experiments differ significantly regarding, e.g., test method, ice conditions, measured variables and testing boundary conditions, which makes a comparison as well as the derivation of general laws for material properties of ice challenging. Moreover, most studies investigate univariate relationships whereas in reality the behavior of ice is influenced by numerous factors. In this paper an explainable machine learning approach to analyze data from various bending tests and to identify relevant features regarding the flexural strength of ice is presented. Using an approach similar to Kellner et al. [1,2], a database of flexural strength tests is established. The data is used to create machine learning models, whose predictions are interpreted with the explainable AI (XAI) method Shapley Additive exPlanaionts (SHAP). The goal is to show a new approach to investigate the flexural strength of ice and to get a better understanding of how suitable the use of XAI for this problem is