Machine learning has proven to be very successfull in many disciplines and the general development is surely furthered by openly accesible algorithms and datasets for the whole community to work with. In geotechnics however, there are very few open datasets that could be used as benchmarks which possibly hinderes the development of ML in our field.
This is why I am currently looking for big open datasets related to geotechnics or engineering geology. Ideally the dataset contains input and output data (e.g. tunnel boring machine operational data + corresponding rockmass classification; or cone penetration tests + nearby borehole logs...) and there are no restrictions on using it.
TC304 Engineering Practice of Risk Assessment & Management of the International Society of Soil Mechanics and Geotechnical Engineering (ISSMGE) provides a list of open datasets that include data of soil and rock mechanics on its website.
Hi.,
Machine Learning in Geotechnics(MLG) is an Artificial Intelligence (AI) field of study that allows computers to learn from existing data without being explicitly programmed. In recent years, the application of ML in wide range of industries has grown rapidly
For More info:
Article An artificial neural network based model to predict spatial ...
Hello Mrs. Gomathy,
thank you for your answer and thank you for pointing me towards this article. However, this does not help me in my search for open datasets.
Georg
Some governments have portals with this type of data, and are often publicly accessible (unless they are confidential). Some examples:
- The British Geological Survey contains 400 datasets, including environmental monitoring data, digital databases, physical collections (borehole core, rocks, minerals and fossils), records and archives. https://www.bgs.ac.uk/data/search/search.cfc?method=selectData
- The US Geological Survey contains borehole-geophysical logs and related data files. https://webapps.usgs.gov/GeoLogLocator/#!/search
- Borehole Log Lithology dataset of British Columbia https://catalogue.data.gov.bc.ca/dataset/borehole-log-lithology
I hope this helps.
Hi Eduardo,
thank you and yes, that's right. There is also quite a lot of geo-data available via different data portals from the European Union (e.g. some maps and digital elevation models etc.: https://ec.europa.eu/eurostat/web/gisco/geodata/reference-data)
Do you happen to know any free datasets that have more of a rock - or soil mechanical background? Laboratory testing databases etc.?
FYI, by now I found one big online repository hosted by the geological survey of the Netherlands that provides a huge database of boreholes, Cone Penetration Tests etc.:
https://www.dinoloket.nl/en/subsurface-data
However, I was only able to download the data one by one, so it would be nice if anybodey would find a more efficient way ;)
Hey Georg H. Erharter
I found this article of Machine Learning data sets a few days back which contains more 70 of them.
You might get the data set you're looking for in it.
https://data-flair.training/blogs/machine-learning-datasets/
Hi Shivam Garg
thank you for the input, but these datasets are not related to engineering geology or geotechnics, as I have specified in my question.
Another update for this question: Google has just started a service called "Dataset Search" (https://datasetsearch.research.google.com/) where one can search for datasets like any other google search.
I would recommend a widely adopted dataset for TBM performance prediction which was proposed by Yagiz. This dataset was collected from the Queens tunnel, NYC, USA. This dataset might be helpful for you.
Hi Qianli Zhang
thank you very much for the hint!
Are you referring to the dataset from this paper: https://www.sciencedirect.com/science/article/abs/pii/S0886779807000508
?
Qianli Zhang I'm sorry, but do you have a copy of that dataset that you could share here? I don't have access to sciencedirect materials :c
Greetings from chile!
Dear Georg H. Erharter We have been developing an open-source site database for all the earthquake-induced strong-motion recording stations (total number: 1700+) in Japan. This database contains wave-velocities, depth to bedrcok, site resonant frequency, surface geology and topographic indicators (e.g., slope gradient). For more details, please refer to " A Site Database for all K-NET and KiK-net Stations" :
https://www.researchgate.net/project/A-Site-Database-for-all-K-NET-and-KiK-net-Stations
Enabling the application of machine learning techniques is one of our motivations to build such an open-source database. This database will be released soon. I hope it can benefit the community.
Chuanbin Zhu thank you for the input and this sounds like a very interesting project and I think it is great that you are working towards open source databases, which will certainly benefit the community. I am looking forward to seeing the data!
Recently, there is a number of papers regarding the application of machine learning in geoscience. The publishers of some paper (e.g. AUG journals) require the data statements in their acknowledgments. The researchers who are interested can try to explore along this way.
Just give an example. The GRL paper titled "Machine Learning Seismic Wave Discrimination: Application to Earthquake Early Warning. gives the data statements :
The Japanese waveform data can be downloaded from http://www.kik.bosai.go.jp/ (last accessed October 2017). For southern California we have used waveforms and parametric data from the Caltech/USGS Southern California Seismic Network (doi:10.7914/SN/CI) stored at the Southern California Earthquake Data Center (doi:10.7909/C3WD3xH1).
I wanted to inform you that I was just closely involved in the preparation and publication of a geotechnical datatset that is comprised of 1339 CPT tests and ~450 of them are also labelled with a soil classification over depth. My colleague Simon Oberhollenzer is leading this project and an accompanying publication is currently under submission in the "Data in Brief" Journal.
The dataset is however already online and you can download it from the following webpage: https://www.tugraz.at/en/institutes/ibg/research/computational-geotechnics-group/database/
We'd highly appreciate any feedback on this dataset.
TC304 Engineering Practice of Risk Assessment & Management of the International Society of Soil Mechanics and Geotechnical Engineering (ISSMGE) provides a list of open datasets that include data of soil and rock mechanics on its website.
- Badges
- Science topic