Assessment of the use of artificial neural networks to detect and diagnose some soft faults in heat pumps

This paper aims to explore the use of artificial neural networks for soft fault detection and diagnosis in a water-to-water heat pump. Unfaulty and faulty operational data are collected from a dedicated experimental campaign. The artificial neural networks are first trained o unfaulty conditions to allow them to predict some of the operational parameters that are usually measured in a heat pump during normal operation. Then, their potentiality in detecting and identifying faults is assessed by comparing the parameters measured under faulty conditions with those predicted by the trained artificial neural networks.