Three-Dimensional Temperature Field Prediction in Double-Wall Cooling Structure Using Deep Learning Method

The double-walled cooling structure can achieve high cooling efficiency but has a higher risk of failure compared to traditional structures. In order to optimize the structural variables of the structure, a fast and accurate temperature field prediction method is urgently proposed This study establishes a deep learning model using MLPs and SRCNN modules to predict the 3D temperature field of the outer wall of a double-walled cooling structure (DWCS) unit. The model takes in geometric structure variables and working condition variables as inputs. To train the model, a temperature field dataset is generated by CFD numerical simulation. The results demonstrate that the deep learning method can accurately predict the 3D temperature field of the DWCS unit at multiple scales, and the model training can be convergent with the proper design of the model architecture and training strategies. Compared to numerical simulation, the deep learning model can predict the temperature field quickly and be combined with machine learning optimization algorithms for the optimization of DWCS variables.