A computationally efficient method for fault diagnosis of fan-coil unit terminals in building Heating Ventilation and Air Conditioning systems. (January 2020)

• Record Type: Journal Article
• Authors: Ranade, Akshay; Provan, Gregory; El-Din Mady, Alie; O'Sullivan, Dominic
• Is Part Of: Journal of building engineering. Volume 27(2020)
• Publication Date: 2020-01
• Subjects: Fault detection and diagnosis (FDD) -- Fan-coil units -- Grey-box modelling -- Regression modelling; Building -- Periodicals; 690.05
• DOI: 10.1016/j.jobe.2019.100955
• Languages: English
• ISSNs: 2352-7102

Abstract

Abstract: Fan-coil units are widely used as terminal units in Heating, Ventilation and Air-Conditioning (HVAC) systems in buildings. Fault Detection and Diagnosis of HVAC systems has been an active area of research for several decades. However, the focus has mostly been on central units such as Air Handling Units, Chillers and Boilers, and Variable Air Volume (VAV) terminal units. In this work we propose a diagnosis scheme for fan-coil units based on a grey-box model based approach. The main contribution of this work is a systematic sub-system level diagnosis case study of the Fan Coil Unit. A systematic procedure to obtain a simplified model of a heat exchanger coil based on polynomial regression is described. The model is used to generate residuals. The results show that the residuals from this model facilitate accurate fault isolation by means of simple rules. The model is characterised by a small set of parameters and is computationally light-weight, thereby making it suitable for embedded diagnosis. For the control problem, the zone thermostat is sufficient. However, for facilitating diagnosis, additional sensors are required. We also examine the role played by different sensors in the fault detection and isolation. Highlights: Novel grey-box model for a heat exchanger coil. First systematic fault diagnosis for a Fan Coil Unit terminal at the subsystem level. The method and underlying model are accurate and also computationally efficient. Method suitable for embedded systems.