Research On Multi-fidelity Model-driven Solid Rocket Motor Optimization

Based on the demand to improve the optimization efficiency of solid rocket motor(SRM),this paper adopts the optimization method based on surrogate model to carry out the research on the multi-fidelity model-driven optimization design method and its application,so as to provide a technical basis for improving the efficiency of SRM optimization design.The multi-fidelity simulation model of SRM was established.Combined with the relationship between different fidelity simulation models,the construction method of multi-fidelity surrogate model based on smooth training of scale function was proposed,the influence of different fidelity sample points on the surrogate model was analyzed,and the model feature-driven biased design of experiment method and adaptive parallel sampling method of multi-fidelity model were established.The overall optimization design of SRM by multi-fidelity models was completed to verify the feasibility of the optimization method.The multi-fidelity simulation model of SRM was established,including the calculation of grain burning surface,internal ballistic,specific impulse and mass.A fast calculation method of burning surface based on nearest neighbor search was proposed,which solves the problem of fast calculation of minimum distance function;Combined with one-dimensional quasi steady internal ballistic calculation,the coupling simulation of internal ballistic calculation and burning surface displacement was realized,and the problem of accurate and rapid prediction of internal ballistic performance under the influence of erosion was solved;Based on the secondary development of commercial software,the parametric model of each component of SRM is established to realize the accurate calculation of SRM mass characteristics;A low-precision modeling method based on historical data was proposed.Through the mapping of geometric and performance parameters under different design requirements,the reuse of historical data was realized,and the data application level was greatly improved.A multi-fidelity surrogate model construction method based on smooth training of scaling function was proposed.The relationship between different precision models was analyzed,and the scale function smoothing training method was proposed to improve the construction effect of low-fidelity data on the multi-fidelity surrogate model,and significantly improved the prediction accuracy of the multi-fidelity surrogate model.The multi-fidelity surrogate model construction method proposed in this paper showed better performance and robustness than other multi-fidelity surrogate models in test examples with different dimensions and different correlations.An adaptive parallel sampling method was proposed.The optimized Latin hypercube design method was used to realize the spatial uniform sampling of the lowfidelity model.Using the prior information provided by the low-fidelity model,the biased sampling of the high-fidelity simulation model was carried out based on the weighted clustering method,which improves the pertinence of the high-fidelity simulation calculation.The influence of different fidelity sample points on the multi fidelity surrogate model was analyzed,and an adaptive parallel sampling method for dynamic configuration of multi fidelity model was proposed to realize the rapid optimization of multi-fidelity surrogate model.The efficiency of the method was verified by the comparative analysis of several numerical examples.The optimization design method of SRM based on multi-fidelity model was completed,and the overall design optimization of SRM is carried out.The calculation results were compared with the traditional optimization algorithm based on single fidelity model.Through the comparison of design results and efficiency,the effectiveness and efficiency of this method were verified,which lays a good foundation for the popularization and application of the method in engineering practice.