Abstract: In-situ measurements of thermal deformation of high-speed spacecraft are a hot and difficult research topic in the aerospace field. The aerodynamic heat generated during flight can cause serious thermal deformation. To improve the reliability and safety of spacecraft in service, research has been carried out on related structural deformation reconstruction methods. This study focuses on the structure of composite material control surfaces and conducts simulation analysis using ABAQUS software. By combining the advantages of Ko displacement theory and BP neural networks, a high-precision method for reconstructing spacecraft displacement is developed. The reconstruction of the deformation field using Ko displacement theory is performed under force load and steady-state temperature conditions, with satisfactory results. The reasons for the failure of Ko displacement theory under the combined action of force load and temperature gradient are analyzed, and BP neural networks are introduced to address the issue of low reconstruction accuracy. Sensitivity analysis is also utilized to determine the location of nodes that significantly impact the output, providing a reference for node selection. The effects of temperature information and the number of training set samples on BP neural networks are discussed. The research results indicate that temperature information plays a significant role in reconstruction accuracy; when the training sample proportion is below 10%, the relative error rises sharply; as the training sample proportion reaches 20%, the error stabilizes gradually, indicating that after a certain training set size, the sample quantity has a diminishing impact on reconstruction accuracy.