一种基于GPU的实验变差函数计算优化算法
|
摘要
在地质统计学中,有效克里格插值的第一步是得到稳健的变差函数计算结果。目前主流的变差函数研究都是集中在对实验变差函数的拟合过程,而对实验变差函数计算过程的优化却非常少。然后在实际应用中,稳健的变差函数需要对大量的数据进行配对计算,并进行多次调试,三维环境下,计算量则更为庞大,使用传统的方法会将克里格插值过程中大部分时间花费在实验变差函数的计算过程中。论文在实验变差函数计算过程中,采用重新搜索空间样品对,GPU并行计算的方式优化计算过程,得到基于GPU的优化算法。最后使用了不同规格的数据进行算法分析,论文优化算法有着显著的性能提高,该算法现已初步应用到武汉地大坤迪公司储量估算软件平台中。
In geostatistics,the first stop to have kriging interpolation is getting steady result of variogram function. At present,the main stream of variogram study focuses on fitting process of experimental variogram while the optimization of computational process of experimental variogram is rarely involved. However,getting steady result of experimental variogram needs to conduct match-ing computation with large amounts of data with repeating debugging in practical application—not to mention the tremendous calculation in 3D environment. Experimental variogram's computing process costs most of the time in the process of kriging interpolationby using traditional method. This article discusses the optimization algorithm based on GPU by researching spatial sample pairsand GPU parallel computing in calculation process of experimental variogram. Then different data specifications are conducted to algorithm analysis. This article's optimization algorithm has significant improvement in performance,which has been initially applied to the reserve estimation software platform in the company of Kundi Data.
In geostatistics,the first stop to have kriging interpolation is getting steady result of variogram function. At present,the main stream of variogram study focuses on fitting process of experimental variogram while the optimization of computational process of experimental variogram is rarely involved. However,getting steady result of experimental variogram needs to conduct match-ing computation with large amounts of data with repeating debugging in practical application—not to mention the tremendous calculation in 3D environment. Experimental variogram's computing process costs most of the time in the process of kriging interpolationby using traditional method. This article discusses the optimization algorithm based on GPU by researching spatial sample pairsand GPU parallel computing in calculation process of experimental variogram. Then different data specifications are conducted to algorithm analysis. This article's optimization algorithm has significant improvement in performance,which has been initially applied to the reserve estimation software platform in the company of Kundi Data.
引文
[1]ZHU Jiacheng.A Three Dimensional Experimental Vario-gram Optimization Algorithm Based on the Method ofWhole-to-parts with GPU[J].International Journal ofEarth Sciences and Engineering,2015:2375-2385.
[2]朱家成.面向智慧矿山的智能型克里格储量估算法研究[D].武汉:中国地质大学,2016.ZHU Jiacheng.Research on Intelligent Kriging ReserveEstimate Method for Smart Mine[D].Wuhan:China Uni-versity of Geoscience,2016.
[3]王仁铎.线性地质统计学[M].北京:地质出版社,1988.WANG Renze.Linear Geostatistics[M].Beijing:GeologyPress,1988.
[4]侯景儒等.适用地质统计学(空间信息统计学)[M].北京:地质出版社,1998.HOU Jingru,et al.Applicable geostatistics[M].Beijing:Geology Press,1988.
[5]侯景儒.地质统计学及其在矿产储量计算中的应用[M].北京:地质出版社,1982.HOU Jingru.Geostatistics and Its Application in The Cal-culation of Mineral Reserves[M].Beijing:Geology Press,1988.
[6]於崇文.数学地质的方法与应用[M].北京:冶金工业出版社,1980.YU Chongwen.Methods and Applications of Mathematical Geology[M].Beijing:Metallurgical Industry Press,1980.
[7]潘国成,李秀峰.地质统计学储量估算方法在矿业中应用的若干实际问题(二)[J].国外金属矿山,1996(8):5-9.PAN Guocheng,LI Xiufeng.Some practical problems inthe application of geological statistical reserve estimationmethod in mining industry(Second)[J].Foreign MetalMining Magazine,1996(8):5-9.
[8]黄竞先,侯景儒.泛克立格法和指示克立格法在地球化学探矿中的应用[J].地球科学,1994(3):321-328.HUANG Jingxian,HOU Jingru.The application of TheUniversal kriging method and the kriging method in geo-chemical prospecting[J].Journal of Earth Science,1994(3):321-328.
[9]YAMAMOTO J K.On unbiased backtransform of lognor-mal kriging estimates[J].Computational Geosciences,2007,11(3):219-234.
[10]Emery X.On Some Consistency Conditions for Geostatis-tical Change-of-Support Models[J].Mathematical Geol-ogy,2007,39(2):205-223.
[11]李章林,王平,李冬梅.实验变差函数计算方法的研究与运用[J].国土资源信息化,2008(2):10-14.LI Zhanglin,WANG Ping,LI Dongmei.The study andapplication of the experimental variogram[J].Land andresources informatization,2008(2):10-14.
[12]Agterberg F P.Practical geostatistics[J].Earth ScienceReviews,1979,16(4):385-386.
[13]Huang T,Lu D T,Li X,et al.GPU-based SNESIM im-plementation for multiple-point statistical simulation[J].Computers&Geosciences,2013,54(4):75-87.
[14]General Chair-Molnar S,General Chair-Pfister H,Pro-gram Chair-Doggett M,et al.Proceedings of the ACMSIGGRAPH/EUROGRAPHICS conference on Graphicshardware[C]//ACM Siggraph/eurographics Conferenceon Graphics Hardware.Eurographics Association,2003:105-106.
[15]Jacob R,Carson J.SU-E-QI-14:Quantitative Vario-gram Detection of Mild,Unilateral Disease in Elas-tase-Treated Rats[J].Medical Physics,2014,41(6Part22):379-380.
[2]朱家成.面向智慧矿山的智能型克里格储量估算法研究[D].武汉:中国地质大学,2016.ZHU Jiacheng.Research on Intelligent Kriging ReserveEstimate Method for Smart Mine[D].Wuhan:China Uni-versity of Geoscience,2016.
[3]王仁铎.线性地质统计学[M].北京:地质出版社,1988.WANG Renze.Linear Geostatistics[M].Beijing:GeologyPress,1988.
[4]侯景儒等.适用地质统计学(空间信息统计学)[M].北京:地质出版社,1998.HOU Jingru,et al.Applicable geostatistics[M].Beijing:Geology Press,1988.
[5]侯景儒.地质统计学及其在矿产储量计算中的应用[M].北京:地质出版社,1982.HOU Jingru.Geostatistics and Its Application in The Cal-culation of Mineral Reserves[M].Beijing:Geology Press,1988.
[6]於崇文.数学地质的方法与应用[M].北京:冶金工业出版社,1980.YU Chongwen.Methods and Applications of Mathematical Geology[M].Beijing:Metallurgical Industry Press,1980.
[7]潘国成,李秀峰.地质统计学储量估算方法在矿业中应用的若干实际问题(二)[J].国外金属矿山,1996(8):5-9.PAN Guocheng,LI Xiufeng.Some practical problems inthe application of geological statistical reserve estimationmethod in mining industry(Second)[J].Foreign MetalMining Magazine,1996(8):5-9.
[8]黄竞先,侯景儒.泛克立格法和指示克立格法在地球化学探矿中的应用[J].地球科学,1994(3):321-328.HUANG Jingxian,HOU Jingru.The application of TheUniversal kriging method and the kriging method in geo-chemical prospecting[J].Journal of Earth Science,1994(3):321-328.
[9]YAMAMOTO J K.On unbiased backtransform of lognor-mal kriging estimates[J].Computational Geosciences,2007,11(3):219-234.
[10]Emery X.On Some Consistency Conditions for Geostatis-tical Change-of-Support Models[J].Mathematical Geol-ogy,2007,39(2):205-223.
[11]李章林,王平,李冬梅.实验变差函数计算方法的研究与运用[J].国土资源信息化,2008(2):10-14.LI Zhanglin,WANG Ping,LI Dongmei.The study andapplication of the experimental variogram[J].Land andresources informatization,2008(2):10-14.
[12]Agterberg F P.Practical geostatistics[J].Earth ScienceReviews,1979,16(4):385-386.
[13]Huang T,Lu D T,Li X,et al.GPU-based SNESIM im-plementation for multiple-point statistical simulation[J].Computers&Geosciences,2013,54(4):75-87.
[14]General Chair-Molnar S,General Chair-Pfister H,Pro-gram Chair-Doggett M,et al.Proceedings of the ACMSIGGRAPH/EUROGRAPHICS conference on Graphicshardware[C]//ACM Siggraph/eurographics Conferenceon Graphics Hardware.Eurographics Association,2003:105-106.
[15]Jacob R,Carson J.SU-E-QI-14:Quantitative Vario-gram Detection of Mild,Unilateral Disease in Elas-tase-Treated Rats[J].Medical Physics,2014,41(6Part22):379-380.