This content is not included in your SAE MOBILUS subscription, or you are not logged in.
Prediction of Vehicle Cabin Occupant Thermal Comfort Using Deep Learning and Computational Fluid Dynamics
- Alok Warey - General Motors Global Research and Development, USA ,
- Shailendra Kaushik - General Motors Global Research and Development, USA ,
- Bahram Khalighi - General Motors Global Research and Development, USA ,
- Michael Cruse - Siemens Product Lifecycle Management Software Inc., USA ,
- Ganesh Venkatesan - Siemens Product Lifecycle Management Software Inc., USA
ISSN: 2574-0741, e-ISSN: 2574-075X
Published August 19, 2021 by SAE International in United States
Citation: Warey, A., Kaushik, S., Khalighi, B., Cruse, M. et al., "Prediction of Vehicle Cabin Occupant Thermal Comfort Using Deep Learning and Computational Fluid Dynamics," SAE Intl. J CAV 4(3):269-278, 2021, https://doi.org/10.4271/12-04-03-0022.
Heating, ventilation, and air-conditioning (HVAC) systems can have a significant impact on the driving range of battery electric vehicles (BEVs). In our previous work, high-fidelity Computational fluid dynamics (CFD) simulations, validated against climatic wind tunnel measurements, were coupled with machine learning (ML) algorithms to predict vehicle occupant thermal comfort for any combination of glazing properties for any window surface, environmental conditions, and HVAC settings (flow rate and discharge air temperature). In the present study, the input feature space was expanded significantly to include climate seats (heated/cooled), heated steering wheel, radiant heating pads, and airflow direction from the air-conditioning (A/C) vents. The modified vehicle cabin CFD model, which included these additional features, was used to generate steady-state training and test data. Feedforward artificial neural networks (ANN) were applied to the simulation data to predict the equivalent homogeneous temperature (EHT) for each occupant. The EHT is a recognized measure of the total heat loss from the human body that can be used to characterize highly nonuniform thermal environments. The prediction performance of the trained deep learning models was evaluated on an unseen test dataset. An ensemble of five neural network models was able to achieve a mean absolute error (MAE) of 2°C or less in predicting the EHT for all occupants in the vehicle, which is acceptable for rapid evaluation of thermal comfort technologies under steady-state conditions.