Pump failure prediction for smarter predictive maintenance
Aquafin is responsible for the transport and treatment of household wastewater in Flanders. We manage more than 2500 pumping stations as part of our operations. These installations pump raw untreated sewage from one location to another, e.g. under under a road. In case of pump failure, risks are flooding of streets and discharge of raw untreated waste water to rivers, lakes, …
About the challenge
Aquafin wants to know whether there is any indicative value in existing data that warns for upcoming pump failures.The successful application is an algorithm or data pipeline to detect upcoming pump failures or a visualisation that focusses the attention of asset engineers on possibly malfunctioning pumps. If solution is succesful, a better maintenance would be possible and a reduced risk of sewage discharge to clean surface water. Hence, as part of the Bemas hackathon, you are invited to work on a dataset of four pumping stations.
The dataset comprises logs of relevant online measurements as well as a description of the maintenance performed on the assets during a period of 6 to 12 months. The dataset, shared in csv as well as in the parquet file format, can be downloaded below.
Data scientists, A.I. Machine Learning, asset management specialists.