Fault Detection and Diagnosis for Central Heating System Using Equipment Emulators and Vibration Monitoring Techniques

Abstract

<p>The presented research provides a novel approach for the development of Fault Detection and Diagnosis (FDD) algorithms for equipment commonly found within legacy buildings. A physics based non-condensing boiler model – capable of emulating 3 types of faults – was designed as the basis for the boiler FDD. The model outputs were then classified using machine learning algorithms. The pump FDD was performed using time-series analysis of experimental vibration data of common bearing faults as inputs to machine learning algorithms. Both boiler and pump models were highly accurate, averaging a classification accuracy of 94% and 99%, respectively. This research provides operators with the ability to leverage Building Automation Systems (BAS) and vibration data to detect and track equipment degradation or poor operating characteristics. The FDD tool developed in this research enables HVAC equipment issues to be rapidly corrected and maintain a high level of efficiency – maximizing service life while reducing cost and energy use.</p>