Abstract
Pumps play a central role in water supply systems, especially for high-rise buildings. The inefficiency or failure of pumps leads to high energy cost or water supply disruption. Thus, it is crucial to monitor their health. While advanced pumps have in-built sensors that help to monitor the condition of the pump, most legacy systems are not equipped with such pumps. Thus, we adopt a more practical approach by installing non-invasive sensors to measure power, vibration, water pressure and level of a water pump system in a university building and collected a year of data. In this paper, we present our dataset and detail the interesting events and failures that were captured. We demonstrate the utility of our dataset by extracting specification rules from the data and showing that application of those rules allow us to detect those captured failures.