Study On Thermal Model Optimization Method Of Alpha Magnetic Spectrometer Based On Surrogate Modeling

Alpha Magnetic Spectrometer(AMS)has been operating in orbit for more than ten years,and more than 100 billion cosmic rays have been collected through its highprecision detectors.Based on this,the origin of the universe and the important clues of dark matter and antimatter are obtained.In the face of extreme cosmic environment,thermal system is the key to ensure the long-term stable operation of AMS in orbit.Usually,the thermal network model is used for thermal analysis to determine the design parameters of thermal system.The lumped parameter idea that the heat transfer model is based on inevitably simplifies and assumes a large number of model parameters,resulting in the error that cannot be ignored between the AMS thermal model and the on-orbit temperature response data under specific conditions,which significantly affects the reliability of thermal analysis results.It is urgent to modify and optimize the uncertain parameters of the on-orbit thermal model of a class of aerospace instruments represented by AMS,so as to improve the reliability of on-orbit thermal analysis of aerospace instruments based on the heat transfer model.Therefore,this thesis studies the establishment method of AMS system-level thermal network model and its Kriging surrogate model,and establishes an optimization method combining surrogate model and genetic algorithm to optimize the parameters of AMS thermal model,so as to obtain an efficient system optimization method to improve the reliability of thermal model of aerospace instruments,and provide a method reference for the thermal analysis of AMS and other types of aerospace instruments.The main contents and conclusions are as follows:1.Firstly,the basic principles of AMS numerical modeling,surrogate modeling and optimization algorithm are theoretically deduced:the thermal environment and the main on-orbit external heat flux of AMS during on-orbit operation are theoretically analyzed,and the thermal equilibrium equation of AMS node thermal network is further established based on the analysis of each external heat source term of AMS during onorbit operation.The sampling principle and Gaussian process of surrogate modeling for AMS heat transfer numerical model are described in detail.Taking the surrogate model as the temperature solver of optimization iteration,the basic principle and related core functions of genetic algorithm in thermal model parameter optimization are theoretically deduced.It lays a theoretical foundation for the research of AMS thermal model optimization method based on surrogate model and genetic algorithm.2.The numerical modeling of AMS heat transfer model and its surrogate model is studied.According to the finite difference principle of lumped parameter idea,the system-level node thermal network model of AMS installed in the on-orbit operation of the international space station(ISS)is established.Taking the temperature measuring points on different components of AMS as the object,the numerical simulation results of the thermal network model are compared with the on-orbit experimental temperature to verify the reliability of the thermal network model.The results show that the thermal network model of AMS has certain error under specific working conditions,and it is necessary to modify the thermal model to improve the accuracy.The traditional optimization method executes the optimization iterative steps with the original numerical model,and the calculation cost is high.To solve this problem,the Kriging modeling method of the surrogate model of AMS thermal model is studied in this thesis.The input uncertainty parameters are needed to establish the surrogate model and the temperature measuring point of AMS where the output response is located are analyzed and determined,and the sensitivity analysis of the input parameters is further carried out to obtain the influence of various parameters on the temperature results of the thermal network model.3.The optimization method of AMS thermal model based on the combination of surrogate model and genetic algorithm is studied.Firstly,the Latin hypercube method is used to establish the parameter sample set of the surrogate model,and the surrogate models under different parameter sets are established according to the Kriging principle.The reliability of the established surrogate model is verified by additional sampling,and the surrogate model with the best consistency with the AMS thermal network model is selected as the iterative solver of genetic algorithm.Then,the temperature response of the selected AMS thermal model under the initial parameters and the error of the onorbit temperature data are taken as the optimization objectives,and the corresponding weight coefficients of each temperature measuring point in the objective function are determined when the local optimization and global optimization are carried out respectively.Then,the AMS thermal model is modified and optimized by the optimization method combined with the surrogate modeling and genetic algorithm.The results show that the average temperature error of AMS power distribution system(PDS)decreases from 1.99℃ to 1.23℃,and the average temperature error of WAKE main radiator decreases from 5.26℃ to 3.05℃ under local optimization.In the global optimization,the square root error of the selected temperature sensor will decrease from 3.95℃ to 2.28℃.The temperature prediction error level of AMS thermal model has been significantly reduced,which proves the reliability of the thermal model optimization method studied in this thesis.