基于GPU并行加速的高速飞行目标实时红外仿真技术研究
发布时间：2019-02-12 18:15
【摘要】：红外仿真在军事、工农业生产、资源勘探等领域都有广泛应用,传统的红外仿真的计算量很大,需要预先求解不同时刻的温度场数据,在所有时刻的温度场求解完成后才可生成红外仿真结果。这种仿真方法中温度场的求解和红外仿真结果输出不具有实时性,不适用于运行工况需要经常变化且需要同步输出红外仿真结果的高速飞行目标问题。由于GPU具有出色的浮点运算能力,可以有效提高红外仿真的实时性,因此对基于GPU并行加速的实时红外仿真的研究具有重要的意义。本文针对典型高速飞行目标的实时红外仿真进行了研究,编写了三维非稳态温度场求解程序,对目标非稳态温度场进行模拟。然后将温度场求解的部分数值运算移植到GPU中,提高温度场求解的速度,达到实时仿真的目的。本文首先介绍了三维非稳态温度场求解的有限体积法,采用附加热源法对不同类型的边界条件进行推导,最终得到一个统一的边界条件处理方式。通过对比不同类型的边界条件下FLUENT计算结果和程序计算结果,验证了三维非稳态温度场求解程序的正确性。使用FLUENT计算出不同飞行高度和飞行速度下的绝热壁温和定壁温热流,将这些数据通过距离反比插值方法插值到目标蒙皮网格上,建立目标气动加热热流数据库,以此为基础建立目标气动加热参数化模型,以附加热流的形式加入非稳态温度场求解程序中,对目标沿指定轨迹飞行过程的非稳态温度场进行求解分析,并由已知的温度场求解出目标不同波长范围的辐射力分布。提出了一种基于GPU并行计算的三维非稳态温度场求解的CUDA实现方法,对该方法能实现的加速比进行分析,发现在网格数目较少的情况下,采用这种GPU并行算法不能起到加速效果,当网格数目增大时,加速比随网数目的增加而增加,说明对大网格数目的模型采用GPU并行运算加速效果较好。最后利用Open GL实时显示该典型高速飞行目标的蒙皮温度场图像。
[Abstract]:Infrared simulation is widely used in military, industrial and agricultural production, resource exploration and so on. The traditional infrared simulation has a large amount of calculation, so it is necessary to solve the temperature field data at different times in advance. The infrared simulation results can only be generated after the solution of the temperature field is completed at all times. In this method, the solution of temperature field and the output of infrared simulation results are not real-time, so they are not suitable for the high-speed flight target which needs to change frequently and needs to output the infrared simulation results synchronously. Because GPU has excellent floating-point computing ability and can effectively improve the real-time performance of infrared simulation, it is of great significance to study real-time infrared simulation based on parallel acceleration of GPU. In this paper, the real-time infrared simulation of a typical high-speed flying target is studied, and a three-dimensional unsteady temperature field solution program is developed to simulate the unsteady temperature field of the target. Then the partial numerical operation of temperature field solution is transplanted into GPU to improve the speed of temperature field solution and achieve the purpose of real-time simulation. In this paper, the finite volume method for solving the three-dimensional unsteady temperature field is introduced, and the additional heat source method is used to deduce the boundary conditions of different types. Finally, a unified boundary condition treatment method is obtained. The correctness of the three-dimensional unsteady temperature field solution program is verified by comparing the results of FLUENT calculation and program calculation under different boundary conditions. Using FLUENT to calculate the adiabatic wall and constant wall heat flow at different flight altitudes and velocities, the data are interpolated to the target skin mesh by the inverse distance interpolation method, and the target aerodynamic heating heat flow database is established. Based on this, a parameterized model of target aerodynamic heating is established. The unsteady temperature field of the target flying along the specified trajectory is solved and analyzed by adding the unsteady temperature field in the form of additional heat flux. The radiation force distribution in different wavelength range of the target is obtained from the known temperature field. In this paper, a CUDA implementation method for solving 3D unsteady temperature field based on GPU parallel computing is proposed. The speedup ratio of this method is analyzed, and it is found that the number of meshes is small. Using this GPU parallel algorithm can not accelerate the result. When the number of mesh increases, the speedup ratio increases with the increase of the number of grids, which shows that the acceleration effect of GPU parallel operation is better for the model with large mesh number. Finally, Open GL is used to display the skin temperature field image of the typical high speed flying target in real time.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN219
本文编号：2420663
[Abstract]:Infrared simulation is widely used in military, industrial and agricultural production, resource exploration and so on. The traditional infrared simulation has a large amount of calculation, so it is necessary to solve the temperature field data at different times in advance. The infrared simulation results can only be generated after the solution of the temperature field is completed at all times. In this method, the solution of temperature field and the output of infrared simulation results are not real-time, so they are not suitable for the high-speed flight target which needs to change frequently and needs to output the infrared simulation results synchronously. Because GPU has excellent floating-point computing ability and can effectively improve the real-time performance of infrared simulation, it is of great significance to study real-time infrared simulation based on parallel acceleration of GPU. In this paper, the real-time infrared simulation of a typical high-speed flying target is studied, and a three-dimensional unsteady temperature field solution program is developed to simulate the unsteady temperature field of the target. Then the partial numerical operation of temperature field solution is transplanted into GPU to improve the speed of temperature field solution and achieve the purpose of real-time simulation. In this paper, the finite volume method for solving the three-dimensional unsteady temperature field is introduced, and the additional heat source method is used to deduce the boundary conditions of different types. Finally, a unified boundary condition treatment method is obtained. The correctness of the three-dimensional unsteady temperature field solution program is verified by comparing the results of FLUENT calculation and program calculation under different boundary conditions. Using FLUENT to calculate the adiabatic wall and constant wall heat flow at different flight altitudes and velocities, the data are interpolated to the target skin mesh by the inverse distance interpolation method, and the target aerodynamic heating heat flow database is established. Based on this, a parameterized model of target aerodynamic heating is established. The unsteady temperature field of the target flying along the specified trajectory is solved and analyzed by adding the unsteady temperature field in the form of additional heat flux. The radiation force distribution in different wavelength range of the target is obtained from the known temperature field. In this paper, a CUDA implementation method for solving 3D unsteady temperature field based on GPU parallel computing is proposed. The speedup ratio of this method is analyzed, and it is found that the number of meshes is small. Using this GPU parallel algorithm can not accelerate the result. When the number of mesh increases, the speedup ratio increases with the increase of the number of grids, which shows that the acceleration effect of GPU parallel operation is better for the model with large mesh number. Finally, Open GL is used to display the skin temperature field image of the typical high speed flying target in real time.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN219
【参考文献】
相关期刊论文 前3条
1 吴晗平;;掠海红外动态点目标图像仿真系统[J];光学精密工程;2007年08期
2 张明明;傅忠谦;;基于OpenGL的导弹红外仿真研究[J];红外技术;2011年06期
3 黄涛;明德烈;揭斐然;雷波;;一种实时红外场景仿真方法[J];计算机与数字工程;2013年01期
，本文编号：2420663
本文链接：https://www.wllwen.com/kejilunwen/dianzigongchenglunwen/2420663.html
