A heat flux distribution prediction method for hypersonic flight vehicle along trajectory based on POD and TSCN

Accurate and rapid determination of heat flux distribution along trajectories is essential for hypersonic flight vehicles. However, transient computational fluid dynamics (CFD) is time-consuming, which makes conjugate heat transfer (CHT) analysis prohibitively expensive. To address this issue, we propose a data-driven heat flux distribution prediction method using proper orthogonal decomposition (POD) and temporal-spatial convolutional network (TSCN) to replace CFD simulations in CHT analysis. This method derives the surface heat flux modes using POD to enhance the modeling accuracy. Subsequently, a TSCN model, capable of extracting temporal and spatial features from recent flight states and non-uniform wall temperatures affecting CFD, is developed to efficiently predict the low-dimensional mode coefficients, which can then be swiftly reconstructed into the heat flux distribution. The proposed method was employed to predict the heat flux distribution of the re-entry capsule along the return trajectory, achieving an average relative prediction error below 2 % by the TSCN model built on the samples from 15 possible trajectories. Above all, the heat flux distributions along a new trajectory can be obtained within an hour by the proposed method, with an efficiency increase of about 200 times in comparison with traditional CHT analysis.