Getting and Keeping the Right Temperature
Heating, ventilation and air conditioning systems are an often unnoticed, yet integral, part of buildings. Behind the scenes, they make rooms comfortable for the people using them. In doing so, they also typically use half of the energy consumed by the buildings they service.
Students and faculty from the Civil and Environmental Engineering department are working on a number of projects to improve the energy efficiency of these systems. Part one of this two-part series detailed the research of PhD students Jerry Lei and Jingkun Gao, who are improving software used in HVAC systems and naming sensors in the systems, respectively.
This article explores the work of PhD student Irem Velibeyoglu as well as a collaboration between IBM and Carnegie Mellon, through which members of the CEE department are assisting Facilities Management Services (FMS) – which maintains all aspects of CMU’s buildings – in deploying an energy-saving analytics system.
Detecting and Diagnosing Faults
Most problems in HVAC systems are detected only when they create noticeable outcomes, like air that is too hot or cold. Because the systems are large and complex, it is often hard to find where the issues stem from even after they are detected.
Velibeyoglu, a PhD student advised by professors Hae Young Noh and Matteo Pozzi, is using a system she is creating in order to more easily pinpoint and diagnose system problems.
“We are trying to localize faults,” Velibeyoglu said. “Earlier … we would only know about a fault if there was a complaint from people inside a building.”
Through the model, building managers may one day be able to monitor HVAC systems to detect faults at all times. Velibeyoglu aims to develop an online system that will show, in real-time, issues and their locations within systems as they arise.
“If we can find anomalies in the behavior of the system, we can tell the … facility managers there is something happening that they may want to check,” said Noh.
When her model is complete, Velibeyoglu also aims to have the ability to diagnose multiple issues occurring at the same time. When faults occur simultaneously, it is often hard to differentiate those issues from one another.
If Velibeyoglu is successful, though, it will be possible to analyze and separate the data being collected about multiple issues so that each fault can be precisely pinpointed. The model would enable people maintaining a system to find and fix issues without needing to first inspect the entire system. With Velibeyoglu’s model, faults in HVAC systems will be found long before they become noticeable to a building’s occupants.