A sparse reconstruction method of physical field via multi-source sensors for flight vehicle

The physical field reconstruction depending on sensors installed sparsely on the surface can provide more valuable and detailed information for design, control, and maintenance of flight vehicles. When various physical properties, such as pressure and heat flux, need to be measured simultaneously, different sensors will be installed for corresponding properties, which increases the number of sensors, leading to measuring complexity and cost. Given that different physical fields under the same flow conditions are correlated, a sparse reconstruction method utilizing information from different types of sensors is proposed to reduce the number of sensors and improve reconstruction accuracy. Meanwhile, a theoretical analysis is conducted to determine the minimum number of multi-source sensors required by the proposed method to achieve target reconstruction accuracy. The method is achieved by three steps. Firstly, with proper orthogonal decomposition algorithm, the modes of each physical field are recognized and extracted. Then, by using radial basis function neural network, the mapping from measurements of multi-source sensors to mode coefficients of multiple physical fields is created. Lastly, for given total number of sensors, the optimization model to determine the optimum locations and types of sensors is built, and solved by differential evolution algorithm. As a verification case, the pressure and heat flux fields of reentry capsule in hypersonic, are reconstructed. The results indicate that, compared with traditional single-source reconstruction method, the proposed method can decrease the total sensor number by about 50% with the same accuracy level, and with the same total number of sensors, the sparse reconstruction error can be reduced by 16% to 99%.