High-Performance Numerical Simulation Software for Life Extension of High-Speed Trains: Developments and Applications
摘要
For the whole-vehicle-level life evaluation and optimization of high-speed trains (HSTs), we develop the theory and model for life extension optimization (LEO) of HSTs for exascale computing, thus improving the scale and efficiency of calculations. We realize a million-core parallel computing for aerodynamics/structure coupling analysis by breaking through the massively parallel algorithms for aerodynamic solving and overset grids that are applicable to domestic heterogeneous supercomputing platforms. From the perspective of system mechanics, we propose a hierarchical and collaborative LEO method for HSTs, which organically links the multi-physics analysis and optimization with the high-performance computing (HPC). For the needs of full-process simulations of LEO, we independently develop the core functional modules used for aerodynamic solving, structural parametric modeling, structural finite element analysis, fatigue life calculation and structural multi-objective optimization, thus integrating the high-performance numerical simulation software for LEO of HSTs (“Numerical HST Version 1.0”). We have completed the software deployment and testing on typical HPC platforms like “Sunway Taihulight” and “SunRising-1”, as well as in the computing environment of CRRC Sifang Co., Ltd. Moreover, the developed software has been applied in the tasks for the safety and reliability design of HSTs, such as aerodynamic surface modification design for 400 km/h HSTs, prediction of the respective contributions of load components on the bogie frame to its structural fatigue life, and LEO for the key components of HSTs.