Abstract:
In the Phase 1 of dissertation, a field survey was conducted to assess indoor thermal
comfort in dormitory and offices buildings followed by calculation of comfort
temperature (Tc). Afterwards, a comparative analysis of three Tc prediction
adaptive models (linear, cubic and logistic) was conducted. In the last part of Phase
1, multiple variables were input in logistic and a machine-learning algorithm for
prediction of thermal sensation. Furthermore, gender and seasonal differences were
considered during dormitories data analysis. However, different ventilation modes
were considered for analysis of offices data. Although thermal sensation votes of
both genders in dormitories were statistically different, no statistical difference in
indoor Tc between two genders were observed. Following Griffth’s method Tc in
dormitories were calculated as 26.8±1.5oC and 27.6±1.7oC during summer and
22.7±2.3oC and 22.3±2.0oC during winter for female and male occupants
respectively. Furthermore, in offices comparison of natural and central HVAC
system showed significance (p>0.05) in sensation and preference votes. Mean Tc
for offices under all five modes were 27.66, 27.18, 26.89, 19.15 and 19.73oC.
Percentage accuracies of three adaptive prediction methods under study showed
better performance of logistic regression. Besides, percentage accuracies of models
were improved when all variables were input in the model.
