Optimization and metamodelization based on machine learning of a new neuro human thermal model

Nowadays, due to climate change, heatwaves become stronger in terms of frequency and intensity. This phenomenon can have serious impact on the indoor environments, indoor thermal comfort and on public health. These situations of high indoor thermal conditions can expose the occupants to health risks such as hyperthermia, dehydration, and heat strokes. Then, the estimation of these risks is crucial. The currently used indices to estimate health risks such as WBGT, HSI and PHS are generally dedicated to outdoor environments and for subjects exerting heavy activities. In addition, these indices do not consider the thermophysiological responses of the human body. In 2020, a human thermoregulation model, called NHTM, was developed to mimic the thermal behavior of the human body in indoor non-uniform and transient conditions. The outputs of the NHTM are the central temperature and the water loss that can be used to assess the health risks. This model considers the interindividual differences between different populations by adjusting its parameters using genetic algorithm and choosing the values that correspond to the studied population. The present study aims to 1) optimize the NHTM model using genetic algorithm on Stolwijk and Hardy study and 2) simplify the thermophysiological calculation by developing metamodels that reduce the calculation time and the complexity of a non-uniform calculation using the NHTM model.