CFD driven prediction of mean radiant temperature inside an automobile cabin using machine learning

The present work describes the effect of various climatic conditions on the thermal environment inside an automobile cabin soaked under direct sunlight. Three-dimensional soaking simulations of the in-cabin flow and heat interactions have been carried out using the commercial solver ANSYS Fluent 18.1. The influence of external climatic conditions has been incorporated into the computations through three parameters, viz. ambient temperature, solar flux, and wind speed. The effects of each of these parameters have been investigated by considering a rich parametric space consisting of different values and combinations of these parameters. The influence of these variables on the thermal environment inside the cabin is described with the help of temperature contours. The MRT at the driver’s location has been evaluated, and its dependence on each of the external climatic parameters is reported. The results from the CFD simulations have been further used to train three supervised machine learning algorithms, viz. linear regression with stochastic gradient descent (LR), random forests (RF), and artificial neural network (ANN) to predict MRT at the driver’s location. The MRT predictions made by these models have also been compared based on the performance metrics such as mean absolute error and Wilcoxon signed-rank test. The machine learning model’s performance has been tested using climatic data of different cities. These results indicated that the machine learning models make predictions above 99% accuracy. This methodology enables MRT estimation without relying on experiments or CFD simulations and subsequently allows better control and automation of automobile air-conditioning systems.