Development of new methodologies and tools for physiological measurements of indoor thermal comfort and well-being in buildings

The quality of indoor environments significantly affects occupants' well-being, with thermal conditions playing a vital role in health, work efficiency, and energy consumption. Traditional thermal comfort models rely on steady-state assumptions, failing to account for dynamic physiological responses to environmental changes. To address this gap, recent research has focused on developing Personal Comfort Models (PCMs) based on physiological data from wearable sensors. This thesis explores the application of wearable devices in multidomain comfort. By integrating physiological measurements and Machine Learning (ML), the research aims to advance PCM development. It also evaluates the effects of thermal discomfort in warm environments on work productivity, providing scientific evidence for the benefits of PCM adoption. The measurement chain includes physiological signals: Electroencephalogram (EEG), Heart Rate Variability (HRV), Electrodermal activity (EDA), and Skin Temperature (ST). Those signals were acquired via wearable devices, Interaxon MUSE, and Empatica e4. Experimental campaigns were conducted in NEXT.ROOM, a controlled environment. For each domain of investigation, appropriate protocols were followed, specifically fifty-two participants were exposed to three temperature settings (cold: 18°C ± 0.4, neutral: 25°C ± 0.2, warm: 32°C ± 0.5), while 24 participants underwent sessions combining temperature and lighting (Neutral: 4114 K, Red: 2010 K, blue Blue 178,000 K), resulting in a total of 416 tests. Physiological data were collected for 5 min after the acclimation period. Thirty-six participants were exposed to three different audio signals while EEG data was continuously recorded for 6 min in the test session. The data were pre-processed to eliminate artifacts, and customs procedures were employed for feature extraction. The identified features were used to create the PCM based on ML, and the uncertainty analysis of the wearable devices was applied to the measurement chain. The findings show significant physiological differences across thermal conditions, with Beta TP10, Gamma TP10, cvsd, and Tonic Perc25 demonstrating p-values < 0.05 in the contrast cold vs warm conditions. In neutral light conditions, the results showed higher EEG Approximate Entropy (ApEn) values in the temporoparietal region compared to blue (p = 0.000009) and red (p = 0.000011). In the acoustic domain, increased delta and theta activity was recorded during exposure to annoying stimuli (p < 0.001), while alpha waves rose with pleasant sounds (H-statistic = 11.1, p < 0.001). These results indicate that wearable EEG devices provide valuable insights into visual and acoustic perceptions. The study further examined the relationship between physiological parameters and thermal sensation (TS), testing various ML algorithms. The results of the uncertainty analysis revealed that Random Forest achieved an accuracy of 86 ± 0.04%. Additionally, it explored the impact of thermal comfort on cognitive and motor performance, revealing that participants in warmer environments had slower initiation times (F(1, 81) = 4.777, p = 0.032, η² = 0.056) but maintained accuracy, indicating a compensatory mechanism for cognitive strain. In conclusion, this thesis highlights the importance of integrating wearable technologies with PCMs for real-time monitoring of thermal comfort. Adjusting environmental conditions based on physiological responses can enhance individual well-being and improve energy efficiency in buildings.