A structural damage identification model with finite thermomechanical sensors of the re-entry module

The re-entry module encounters extremely harsh aerodynamic pressure and heating conditions, and the high-precision identification of the structural damage state is crucial to the flight and reuse performance evaluation. The current techniques are mainly based on complex numerical simulations or indirect sensor measurements of finite nodes in time or space dimensions, respectively. This work developed a damage identification model that included numerical simulations, sensor measurements, and additional machine learnings to obtain the structural damage state of the module. First, the primary damage database was established by thermomechanical numerical simulations and a structural damage model, which was proposed based on the strain-equivalent-based stiffness reduction method with certain structural partition rules. Second, a database expansion method with higher accuracy based on the Kriging agent model was proposed, the damage database was expanded by 10 times with 7 % error. Third, the damage identification model was developed with inputs of the finite nodal temperature and stress and output of structural damage value based on the back propagation neural network, and a structural damage grade evaluation equation was finally formulated. The result shows that the model overfitting is fully suppressed and the identification error is reduced by 60 % compared with the original data without expansion, and great identification accuracy of 92.6 % with error threshold of 0.03 and good anti-interference ability of 1 % sensor noise are exhibited for the model. The model holds higher recognition efficiency and accuracy of structural residual capacity and indicates potentials for real-time safety assessment of re-entry module.