The first challenge, presented by Aquafin, is a case on the detection of failures in pump stations. Being specialists in data science and asset management, this is exactly our cup of tea.

The second challenge is a case presented by Fluvius on a reporting tool for in-the-field survey taking. We have a lot of experience with this type of work from our projects at Infrabel.

Getting Started

For data science projects such as the Aquafin pump failure detection case, we use our internal ‘Python for data science’ cookiecutter templates. Not only do they help us to quickly jumpstart a project, they also make it easy to collaborate. We all know the tooling, and it is trivial to distribute the project with virtual environments and all.

With our dev environment all set up, we start a Jupyter Notebook where we can build a prototype prior to moving code to modules. We downloaded the data and now we get started by writing some simple tooling to make handling them easy. Since there is some variation in the files, it is important to get everything in a consistent format.

Now that we have made it easy to work with the data, we move to our next step of building more tools to visualise and otherwise explore the data set. At this stage, we just want to get a feel for what we are looking at. Are there any interesting artefacts or trends that immediately stand out?