Mechanical Engineering ETDs


Hongbo He

Publication Date



SHW systems are generally expected to last for at least 20 years with little or no maintenance. However, in many cases failures occur far sooner due to a variety of problems, many of which are undetected or detected long after the system has failed because the backup heater silently assumes the heating load. Some of the failures may cause the system to run inefficiently or even damage other system components, such as when a system loses fluid in the solar loop and the pump runs dry, eventually destroying itself. In recent years there has been an increasing demand for SHW systems to become economic and reliable. Fault Detection and Diagnosis (FDD) in SHW systems is an important part of maintaining proper performance, reducing power consumption and unnecessary peak electricity demand. The aim of the current work is to develop anomaly detection system that can reliably detect both anticipated and unforeseen faults and can be implemented in commercial SHW systems without any additional sensors to the ones commonly needed for ordinary system control. Adaptive Resonance Theory (ART)-based neural networks are chosen to perform this task, because the ART-based neural networks are fast, efficient learners and retain memory while learning new patterns. In particular, the ART networks can be incorporated into SHW system controller without any extra sensors and have the capability of an early detection of performance degradation faults. Other benefits of ART-based neural networks are on-line fault detection for its high computational efficiency and no involvement of faulty data for the training process. A testbed for SHW system reliability is developed for the purposes of investigating the fault detection system. The input patterns of the fault detection system are generated from two sensors: collector plate temperature and water tank heat exchanger outlet temperature, which are normally installed in residential SHW systems installed by commercial operators. One of the strengths of the system is that only few data points are needed, meaning that it will not be necessary to instrument SHW systems with additional sensors, something which would not be acceptable in an aggressively competitive industry where reducing costs is paramount. The training data for the fault detection system are generated from a verified SHW system TRNSYS (Transient Systems Simulation) model. The simulation and experimental results show that the ART-based anomaly detection has the capability to accurately and efficiently detect degradation and failure. Faults are detected at various levels depending on their severity. The ART-based anomaly detection can be used for SHW real-time reliability monitoring, as well as, eventually, in larger, more complex systems such as commercial building HVAC systems or subsystems.


Solar water heaters--Automatic control, Neural networks (Computer science)

Degree Name

Mechanical Engineering

Level of Degree


Department Name

Mechanical Engineering

First Advisor

Mammoli, Andrea A.

First Committee Member (Chair)

Mammoli, Andrea A.

Second Committee Member

Razani, Arsalan

Third Committee Member

Vorobieff, Peter

Fourth Committee Member

Caudell, Thomas P.


Sandia National Laboratories.

Document Type