Digital twins for decoding human-building interaction in multi-domain test-rooms for environmental comfort and energy saving via graph representation

Human comfort studies are a complex topic given their interdisciplinary and multi-stimuli nature. The integration of different sources of information from multi-domain experiments for occupants' environ-mental perception investigation can change the way that researchers perceived the interactions among building agents. This study aims to establish a new method for digital twin development from a test room for human-comfort and energy behavior analysis, through a Graph Neural Network. An as-monitored BIM model was integrated with monitoring information, i.e., environmental data from the facility, physiolog-ical signals, and survey answers from occupants, and the relationships between the building agents were outlined, based on the experiment purposes. A real-world experiment was adopted to demonstrate the procedure and results, connecting the information from experiments and a test room facility employed to the study. The original approach filled a recurrent gap in human-comfort studies: the integration of all the sources of information in a single visual platform through Graph Neural Networks, allowing the recognition of patterns, connections, and the influence of the diverse stimuli in the human perception of the surrounding environment. In complex building systems, with large sensing structures, high occu-pancy and multi-source environmental stimuli, new conclusions can be drawn to a full understanding of the human-building interactions.(c) 2022 Elsevier B.V. All rights reserved.