PhD Proposal: Scalable Methods and Tools to Support Thermographic Data Collection and Analysis for Energy Auditing

Buildings account for 41% of primary energy consumption in the United States—more than any other sector—and contribute to an increasing portion of carbon dioxide emissions (33% in 1980 vs. 40% in 2009). To help address this problem, the U.S. Department of Energy recommends conducting energy audits to identify sources of inefficiencies that contribute to rising energy use. One effective tool used during energy audits is thermography. Thermographic-based energy audits use thermal cameras to identify, diagnose, and document energy efficiency issues in the built environment that are visible as anomalous patterns of electromagnetic radiation such as locations of air leakages, missing insulation, and moisture. This auditing technique is becoming more broadly used due to sensor improvements and falling costs, but one factor that impacts its scalability is that using a thermal camera during energy audits requires training and experience. While research has largely focused on automating data collection and modeling, there is a lack of specialized tools, algorithms, and methods to support human-oriented audit activities.

My dissertation focuses on scaling thermography along two dimensions: first, scaling in terms of who can perform thermographic audits by building and evaluating computer assisted thermographic tools to help with both capture and analysis; second, scaling in terms of time by building and evaluating new indoor, automated temporal data collection and analysis tools. To motivate and uncover key insights for my proposal, I have completed two preliminary studies. First, I investigated professional energy auditing practices and the role of thermography therein. Second, I investigated how minimally trained novice end-users approach thermographic energy auditing tasks using relatively new, commodity, smartphone-based thermal cameras. Findings from these studies and a survey of related literature indicate that thermographic energy auditing is largely a qualitative practice, that end-users desire increased support for data collection and analysis tasks (e.g., computer assisted issue analysis, smart emissivity adjustments), and that future systems will need to help end-users verify their findings.

Building on this preliminary work, I propose several threads of research that will help scale thermography along the who and time dimensions. With respect to scaling who can perform thermography, I will demonstrate how off-the-shelf machine learning and image processing techniques can identify anomalies in thermal images and how new, mixed-initiative interfaces can provide capture and analysis support. I will also investigate scaling in time by extending these techniques to enable retrospective analysis of temporal changes in buildings using thermal imagery captured by a new, easily deployable thermographic sensor system and new types of temporal, thermographic visualizations. To evaluate these two threads, I will perform both lab and field usability studies culminating with a multi-week end-user field evaluation similar to my preliminary work. Using pre- and post-questionnaires, device log data, and interviews, I will evaluate the proposed tools for accuracy, ease-of-use, and efficacy with respect to current thermographic tools.

In summary, the contributions of my dissertation will include formative evaluations of end-users of thermographic technologies, the evaluation of machine learning and image processing techniques for thermal anomaly classification, the design of new capture and analysis tools for thermographic energy auditing, and an exploration of new temporal visualization techniques.

Examining Committee:

Chair: Dr. Jon E. Froehlich

Dept rep: Dr. David W. Jacobs

Member: Dr. Andrea Wiggins