Prediction of pit depth growth over time of buried steel pipes in soil with machine learning

Please check whether enrolling in this project is possible by directly contacting the responsible supervisor(s) indicated below as soon as possible. Note that although we announce many topics, we won't be able to offer all of them simultaneously due to limited lab capacity.

romanov data — Predicted mean corrosion rates compared to the measured corrosion rates from the Romanoff [1] (from [2])

Melvin Romanoff has collected a data set with the corrosion data from steel pipes buried in different soils. He reported for different exposure times the weight loss and the maximum penetration depth for the pipes as well as related soil parameters such as pH or the composition of the water near the samples. In a previous bachelor thesis, the mean corrosion rate (calculated from the metal loss in grams) was analyzed with several machine learning algorithms. For several models, the importance of each parameter was calculated.

In this thesis, we aim to analyze the penetration depth instead of the mass loss in dependence of the time. Furthermore, we like to investigate the importance of each parameter in the model. If possible, the predictions by the student can be compared to existing models derived from the same data set.

soil parameter influence — Analysis of the influence of the soil parameters to the mean corrosion rate, where “high” is the specimens which are the 50% highest mean corrosion rates from the data set [1] and “low” are the other 50% (from [2])

References:

[1] M. Romanoff, Underground Corrosion (1957)

[2] E. Hou, Data analysis of Romanoff's "Underground Corrosion" to predict corrosion behavior in soil (2020), Bachelor Thesis (unpublished)