Transient temperature fields of the tank vehicle with various parameters using deep learning method

Abstract

Calculation of transient temperature fields is widely used in engineering application and is also very crucial. Nevertheless, the existing methods for the prediction of complex transient temperature fields are difficult to meet the requirements of accuracy and real-time performance. In this work, a surrogate model for the fast calculation of transient temperature field under different thermal parameters is established based on deep learning. The proposed model combines Long Short-Term Memory (LSTM) network and Three Dimensional Generative Adversarial Neural Networks (3DGAN) to generate transient temperature fields satisfying the corresponding thermal parameters for different moments. To improve the ability of the model to generate details of the temperature field, the perceptual loss is fused in the network. The proposed LSTM-3DGAN model is validated by obtaining the transient temperature field dataset of the tank vehicle under different meteorological conditions through numerical simulation. The results show that the predicted transient temperature field has high accuracy and high-quality details, and the average error of typical temperature points is only 0.62 %. Furthermore, the predicted temperature field satisfies the physical laws of heat transfer. The results of ablation experiments show the prediction accuracy of the proposed model is much higher than that of the 3DGAN and LSTM-GAN models. This study is a successful application of deep learning in the field of complex heat transfer. The proposed model can efficiently generate temperature field datasets for further integration of deep learning with heat transfer.