Vision-based building energy diagnostics and retrofit analysis using 3D thermography and Building Information Modeling

The emerging energy crisis in the building sector and the legislative measures on improving energy efficiency are steering the construction industry towards adopting new energy efficient design concepts and construction methods that decrease the overall energy loads. However, the problems of energy efficiency are not only limited to the design and construction of new buildings. Today, a significant amount of input energy in existing buildings is still being wasted during the operational phase. One primary source of the energy waste is attributed to unnecessary heat flows through building envelopes during hot and cold seasons. This inefficiency increases the operational frequency of heating and cooling systems to keep the desired thermal comfort of building occupants, and ultimately results in excessive energy use. Improving thermal performance of building envelopes can reduce the energy consumption required for space conditioning and in turn provide building occupants with an optimal thermal comfort at a lower energy cost. In this sense, energy diagnostics and retrofit analysis for existing building envelopes are key enablers for improving energy efficiency. Since proper retrofit decisions of existing buildings directly translate into energy cost saving in the future, building practitioners are increasingly interested in methods for reliable identification of potential performance problems so that they can take timely corrective actions. However, sensing what and where energy problems are emerging or are likely to emerge and then analyzing how the problems influence the energy consumption are not trivial tasks. The overarching goal of this dissertation focuses on understanding the gaps in knowledge in methods for building energy diagnostics and retrofit analysis, and filling these gaps by devising a new method for multi-modal visual sensing and analytics using thermography and Building Information Modeling (BIM). First, to address the challenges in scaling and localization issues of 2D thermal image-based inspection, a new computer vision-based method is presented for automated 3D spatio-thermal modeling of building environments from images and localizing the thermal images into the 3D reconstructed scenes, which helps better characterize the as-is condition of existing buildings in 3D. By using these models, auditors can conduct virtual walk-through in buildings and explore the as-is condition of building geometry and the associated thermal conditions in 3D. Second, to address the challenges in qualitative and subjective interpretation of visual data, a new model-based method is presented to convert the 3D thermal profiles of building environments into their associated energy performance metrics. More specifically, the Energy Performance Augmented Reality (EPAR) models are formed which integrate the actual 3D spatio-thermal models (‘as-is’) with energy performance benchmarks (‘as-designed’) in 3D. In the EPAR models, the presence and location of potential energy problems in building environments are inferred based on performance deviations. The as-is thermal resistances of the building assemblies are also calculated at the level of mesh vertex in 3D. Then, based on the historical weather data reflecting energy load for space conditioning, the amount of heat transfer that can be saved by improving the as-is thermal resistances of the defective areas to the recommended level is calculated, and the equivalent energy cost for this saving is estimated. The outcome provides building practitioners with unique information that can facilitate energy efficient retrofit decision-makings. This is a major departure from offhand calculations that are based on historical cost data of industry best practices. Finally, to improve the reliability of BIM-based energy performance modeling and analysis for existing buildings, a new model-based automated method is presented to map actual thermal resistance measurements at the level of 3D vertexes to the associated BIM elements and update their corresponding thermal properties in the gbXML schema. By reflecting the as-is building condition in the BIM-based energy modeling process, this method bridges over the gap between the architectural information in the as-designed BIM and the as-is building condition for accurate energy performance analysis. The performance of each method was validated on ten case studies from interiors and exteriors of existing residential and instructional buildings in IL and VA. The extensive experimental results show the promise of the proposed methods in addressing the fundamental challenges of (1) visual sensing: scaling 2D visual assessments to real-world building environments and localizing energy problems; (2) analytics: subjective and qualitative assessments; and (3) BIM-based building energy analysis: a lack of procedures for reflecting the as-is building condition in the energy modeling process. Beyond the technical contributions, the domain expert surveys conducted in this dissertation show that the proposed methods have potential to improve the quality of thermographic inspection processes and complement the current building energy analysis tools