A data-driven intelligent learning algorithm for simultaneous prediction of aerodynamic heat and thermo-physical property parameters

It is of great importance and challenging to simultaneously determine time-varying aerodynamic heat and temperature-dependent thermo-physical property parameters with high accuracy, for the optimization of thermal protection systems of hypersonic vehicles. However, it is difficult to directly measure these parameters under high temperature conditions. It is an effective way to determine thermo-physical property parameters and aerodynamic heat by solving inverse problems, based on measurable or easily measured transient temperatures. However, the prediction error of these parameters may be too large, if the measurement error is large, due to the thermal inertia. To deal with this issue, an intelligent algorithm is proposed to simultaneously predict the aerodynamic heat and thermo-physical property parameters for the thermal protection systems of hypersonic vehicles, based on the temperature measurement information. It combines a genetic algorithm and a machine learning algorithm, and the genetic algorithm is employed to update the relevant parameters in the neural network. By training the neural network, the relationship among the predicted parameters and transient temperatures could be established. Thereafter, the aerodynamic heat subjected to the outer surface of the aircraft and the temperature-dependent non-linear thermo-physical property parameters could be predicted. Examples are given to verify the present algorithm. The results show that this work provides an accurate and efficient method for simultaneously determining the aerodynamic heat and thermo-physical property parameters for the thermal protection system of a hypersonic vehicle. The prediction errors of aerodynamic heat and thermo-physical property parameters are much smaller than the measurement errors, when there are relatively large measurement errors in the input data.