Anomaly Detection for Industrial Water Pumps
Water pumps are a critical part of our modern infrastructure as they enable power plants, fresh & waste water supply as well as district heating and other use cases. Detecting and predicting outages in such critical parts of our modern society is of high value as this can reduce potential downtimes, lead to shorter repair durations as well as make the maintenance predictable. Using IoT sensors, the goal of this project is to detect anomalies in running water pumps by measuring vibrational profiles.
A variety of different machine learning techniques are investigated. The sensor network is strongly constrained by several properties like battery and mobile network usage. In addition, each water pump is build-wise unique and is applied in a different environment. This introduces further challenges to create generalized models. The collaboration aims at developing a scalable system which is able to detect anomalies within the different pumps and to transfer knowledge about anomalies among the water pumps.
This project is a joint work with the company KSB SE & Co. KGaA , which develops individual water pumps for its customers.