Heat Flux Sensing for Machine-Learning-Based Personal Thermal Comfort Modeling

Area of Science:

• Building Science
• Human-Computer Interaction
• Physiological Computing

Background:

• Personalized thermal comfort is crucial for adaptive Heating, Ventilation, and Air-Conditioning (HVAC) systems.
• Physiological sensing offers a promising avenue for enhancing Human-In-The-Loop (HITL) control.
• Existing models often rely on less direct physiological indicators.

Purpose of the Study:

• To investigate the efficacy of heat flux sensing for inferring personal thermal comfort.
• To explore the relationship between skin heat exchange rates and thermal preferences under transient conditions.
• To evaluate the performance of machine learning models using heat flux data compared to traditional methods.

Main Methods:

• Experimental study involving 18 human subjects under varying ambient conditions.
• Measurement of heat exchange rates from facial and wrist skin.
• Coupling heat flux data with skin and ambient temperatures for machine learning model development.
• Evaluation of well-known classifiers with different learning scenarios.

Main Results:

• Significant variations in personal thermal preferences were observed and correlated with physiological variables.
• Heat exchange rate measurements provided valuable insights into individual thermal perception.
• Models incorporating heat exchange rates demonstrated superior performance in personal thermal comfort inference compared to those using only skin temperature.

Conclusions:

• Heat flux sensing is a highly effective method for improving the accuracy of personal thermal comfort inference.
• Facial and wrist skin heat exchange rates are key indicators for adaptive HVAC control.
• This approach enhances the flexibility and human-centeredness of distributed control systems.