采用GPU加速復(fù)雜結(jié)構(gòu)角系數(shù)計算的方法研究
發(fā)布時間:2018-04-01 22:11
本文選題:角系數(shù) 切入點(diǎn):GPU 出處:《工程熱物理學(xué)報》2017年06期
【摘要】:采用離散傳遞法計算目標(biāo)的紅外輻射特征首先要計算角系數(shù),對于復(fù)雜目標(biāo),角系數(shù)計算會耗費(fèi)大量時間。針對角系數(shù)計算中遮擋判斷問題,提出了一種不需要求解微元面中心點(diǎn)連線與遮擋面交點(diǎn)的矢量法。開發(fā)了基于矢量法的GPU并行程序,并使用一個有理論解的模型驗(yàn)證了程序的正確性,最后針對航空發(fā)動機(jī)排氣系統(tǒng)復(fù)雜模型,在一臺配置有448個核心的GPU處理器和主頻為3.4GHz CPU處理器的工作站上,評估了基于矢量法的GPU程序相對單核CPU程序的加速效果。結(jié)果表明;在微元面數(shù)量為13670時,矢量法可獲得約73倍的加速比,是采用線面求交法獲得的加速比的1.4倍。使用離散傳遞法計算航空發(fā)動機(jī)排氣系統(tǒng)的紅外輻射特征時,僅將角系數(shù)模塊使用GPU并行計算,可減少45%的計算時間。
[Abstract]:In order to calculate the infrared radiation characteristics of the target by discrete transfer method, the angle coefficient is first calculated. For complex targets, the calculation of the angle coefficient will take a lot of time.In order to solve the occlusion judgment problem in the calculation of angle coefficient, a vector method is proposed, which does not need to solve the line of center point and the intersection point of shading surface.The GPU parallel program based on vector method is developed, and a theoretical model is used to verify the correctness of the program. Finally, the complex model of aero-engine exhaust system is used to verify the correctness of the program.On a 448 core GPU processor and a workstation with a main frequency of 3.4GHz CPU processor, the acceleration effect of a GPU program based on vector method compared with a single core CPU program is evaluated.The results show that when the number of differential surfaces is 13670, the speedup of vector method is about 73 times, which is 1.4 times of that obtained by line and surface intersection method.When the infrared radiation characteristics of aero-engine exhaust system are calculated by discrete transfer method, the calculation time can be reduced by 45% by using GPU in the angular coefficient module.
【作者單位】: 南京航空航天大學(xué)能源與動力學(xué)院;
【基金】:航空科學(xué)基金(No.20132752039)
【分類號】:O434.31;TP338.6
【相似文獻(xiàn)】
相關(guān)期刊論文 前1條
1 白心愛;;輻射換熱角系數(shù)的計算[J];紅外;2008年08期
,本文編號:1697511
本文鏈接:http://sikaile.net/kejilunwen/wulilw/1697511.html
最近更新
教材專著