錢江,王凡
(1.河海大學(xué)理學(xué)院,江蘇 南京 211100;2.海岸災(zāi)害及防護(hù)教育部重點(diǎn)實(shí)驗(yàn)室(河海大學(xué)),江蘇 南京 210098;3.南京農(nóng)業(yè)大學(xué)理學(xué)院,江蘇 南京 210095)
非均勻二型三角剖分二元二次樣條的數(shù)值積分公式
錢江1,2,王凡3
(1.河海大學(xué)理學(xué)院,江蘇 南京 211100;2.海岸災(zāi)害及防護(hù)教育部重點(diǎn)實(shí)驗(yàn)室(河海大學(xué)),江蘇 南京 210098;3.南京農(nóng)業(yè)大學(xué)理學(xué)院,江蘇 南京 210095)
給出了構(gòu)成矩形域的4個(gè)三角形子區(qū)域的二元樣條擬插值算子的等價(jià)形式,對(duì)這4個(gè)三角形子區(qū)域分別建立了數(shù)值積分公式,相加后得到一般矩形域上的數(shù)值積分公式,同時(shí)給出了構(gòu)造數(shù)值積分公式所需的結(jié)點(diǎn)處的函數(shù)值與相應(yīng)的求積系數(shù)。進(jìn)一步,利用算子范數(shù)、連續(xù)模及擬插值算子的保多項(xiàng)式性,針對(duì)具有不同連續(xù)性的被積函數(shù),得到了相應(yīng)的數(shù)值積分公式的求積余項(xiàng)。研究表明,提出的數(shù)值積分公式不僅具有較高的計(jì)算精度,而且計(jì)算量約為二元張量積型求積公式的1/5。數(shù)值算例進(jìn)一步說(shuō)明了數(shù)值積分公式的有效性。
多元樣條;光滑余因子協(xié)調(diào)法;二元數(shù)值積分;樣條擬插值; B網(wǎng)
多元樣條是研究多元數(shù)值逼近、計(jì)算幾何、有限元等的重要工具。多元樣條的研究方法除經(jīng)典的張量積型B樣條[1-2]、B網(wǎng)[3-4]外,光滑余因子協(xié)調(diào)法[5-6]值得關(guān)注。光滑余因子協(xié)調(diào)法可以確定多元樣條函數(shù)空間維數(shù),計(jì)算樣條基函數(shù)[6-8],其中表示基于區(qū)域的任意三角剖分上具有階光滑度的次多元樣條函數(shù)空間。
事實(shí)上,借助光滑余因子協(xié)調(diào)法,多元樣條的任何問(wèn)題都能轉(zhuǎn)化為求解等價(jià)的代數(shù)方程組。如文獻(xiàn)[9-10]提出了具有最小局部支集的三次樣條基函數(shù)。文獻(xiàn)[11]提出了二型非均勻三角剖分上的樣條函數(shù)空間。文獻(xiàn)[12-13]建立了均勻二型三角剖分上的二元三次、四次樣條函數(shù)空間及相應(yīng)的樣條擬插值算子。文獻(xiàn)[14-16]研究了基于非均勻二型三角剖分的二元三次異度樣條、樣條擬插值及其導(dǎo)數(shù)逼近。文獻(xiàn)[17]借助三角Gauss型求積公式,建立了具有圓形邊界的凸多邊形元上的低次代數(shù)數(shù)值積分。文獻(xiàn)[18]利用四邊形有限元給出了新的求積公式,并用B網(wǎng)方法計(jì)算得到凸四邊形區(qū)域上的積分元素,結(jié)果表明其具有高精度。文獻(xiàn)[19]根據(jù)帶重節(jié)點(diǎn)在二型三角剖分上的二元二次B樣條函數(shù),推導(dǎo)了樣條擬插值分層逼近方法,研究表明,其具有保多項(xiàng)式性與最優(yōu)逼近性。文獻(xiàn)[20-23]利用樣條擬插值算子在具體算例中計(jì)算了數(shù)值積分。
主要研究?jī)?nèi)容如下:第1節(jié)給出子區(qū)域上樣條擬插值算子的等價(jià)表達(dá)式;第2節(jié)利用樣條基函數(shù)的B網(wǎng)系數(shù)構(gòu)造數(shù)值求積公式;第3節(jié)基于樣條擬插值算子范數(shù)與連續(xù)模,分別推導(dǎo)被積函數(shù)具有連續(xù)性的求積余項(xiàng);第4節(jié)給出數(shù)值算例。
計(jì)算得到,八邊形支集4個(gè)拐角處的三角域上樣條函數(shù)滿足:
用虛線將4個(gè)拐角處的三角域一分為二,如圖1所示,相應(yīng)的曲面片記為,。這樣在4個(gè)拐角處的三角域上曲面片分別具有統(tǒng)一的表達(dá)式[6]。
圖1 的最小八邊形支集Fig.1 Minimal octagonal support of
的二型三角剖分。采用文獻(xiàn)[24-25]構(gòu)造的樣條擬插值算子:
得到
其中,
注1式(3)也可寫為
定理1[20-21]對(duì)任意的,有
定理2對(duì)任意的樣條擬插值算子可表示為
定理3對(duì)任意的樣條擬插值算子可表示為
定理4對(duì)任意的樣條擬插值算子可表示為
定理5對(duì)任意的樣條擬插值算子可表示為
借助于面積坐標(biāo)系,二元二次多項(xiàng)式在三角域上的數(shù)值積分可轉(zhuǎn)化為B網(wǎng)系數(shù)之和與三角形面積的乘積[6],有
引理1設(shè)面積坐標(biāo)系下的二元二次多項(xiàng)式為
則有
定理6三角域上的數(shù)值積分公式為
事實(shí)上,由式(9),可得二重積分:
表1 與樣條函數(shù)對(duì)應(yīng)的和
Table 1andcorresponding to the splines
定理7三角域上的求積公式為
表2 與樣條函數(shù)對(duì)應(yīng)的和
Table 2andcorresponding to the splines
定理8三角域上的求積公式為
表3 與樣條函數(shù)對(duì)應(yīng)的和
Table 3andcorresponding to the splines
定理9三角域上的求積公式為
表4 與樣條函數(shù)對(duì)應(yīng)的和
Table 4andcorresponding to the splines
結(jié)合表1~表4,整理得到
定理10諸矩形子區(qū)域上的求積公式為
表5及其對(duì)應(yīng)的
Table 5and its corresponding
注2由于式(6)具有保多項(xiàng)式性,因此,式(25)對(duì)任意的精確成立。
在計(jì)算量上,式(21)~式(24)均需9次乘法和8次加減法,式(25)需11次乘法和8次加減法,因此,矩形域上的數(shù)值積分共需次乘法和次加減法。為便于說(shuō)明,繪制了矩形域在4個(gè)拐角處的非零B樣條的最小八邊形支集,如圖2所示,將八邊形支集自左至右、自上至下平移,得到每個(gè)三角形子區(qū)域上的所有非零B樣條基函數(shù)。另外,如果采用矩形域上張量積型二元二次樣條函數(shù)構(gòu)造數(shù)值積分公式,則需利用代數(shù)方法精確確定求積系數(shù),計(jì)算量較大。如對(duì)于張量積型B樣條,其數(shù)值積分需次乘法和次加減法。可見,本文構(gòu)造的矩形域數(shù)值積分公式的總計(jì)算量約為基于張量積型B樣條求積公式計(jì)算量的1/5。
圖2 矩形域上八邊形支集的平移過(guò)程Fig.2 Translation process of the octagonal support over rectangular domain
定理11(i)若函數(shù),則對(duì)充分大的正整數(shù),有
證明 對(duì)于情形(i),根據(jù)B樣條基函數(shù)的單位分解性,對(duì)任意的有
有
由此,結(jié)合算子范數(shù),推得
因此,由定理11,對(duì)被積函數(shù)具有不同光滑性的數(shù)值積分,推導(dǎo)相應(yīng)的求積余項(xiàng)。
定理12設(shè)表示矩形域的面積,當(dāng)充分大時(shí),有:
對(duì)所有i,j求和,可得矩形域上的求積余項(xiàng):
對(duì)所有i,j求和,可得矩形域上的求積余項(xiàng):
證畢。
算例1設(shè)函數(shù)非均勻剖分上x方向的節(jié)點(diǎn)為y方向的節(jié)點(diǎn)為矩形胞腔中點(diǎn)處的橫坐標(biāo)分別為0.03,0.075,0.15,0.30,0.50,0.65,0.80,縱坐標(biāo)分別為0.04,0.08,0.20,0.35,0.45,0.525,0.625,由此計(jì)算49個(gè)中點(diǎn)處的函數(shù)值。
由定理10可知,計(jì)算每個(gè)小矩形胞腔上的數(shù)值積分需要與此胞腔相鄰的9個(gè)胞腔中點(diǎn)處的函數(shù)值,因此需計(jì)算矩形域上每個(gè)小矩形胞腔上的數(shù)值積分。將中點(diǎn)縱坐標(biāo)為的5個(gè)矩形胞腔分別稱為第層,得到數(shù)值積分近似值,按MATLAB二重積分程序dblquad計(jì)算真值,結(jié)果如表6所示。
表6 算例1數(shù)值積分的真值、近似值與誤差Table 6 The real values,approximating values and error estimation of the numerical cubature in example 1
算例2設(shè)函數(shù)定義域與矩形剖分同算例1,同理計(jì)算近似值、真值及誤差,結(jié)果如表7所示。
表7 算例2數(shù)值積分的真值、近似值與誤差Table 7 The real values,approximating values and error estimation of the numerical cubature in example 2
本文所構(gòu)造的二元二次樣條的數(shù)值積分公式不僅具有良好的保多項(xiàng)式性,而且計(jì)算量小。另外,由于非均勻三角剖分二元二次樣條表達(dá)式的復(fù)雜性,直接用傳統(tǒng)的樣條函數(shù)計(jì)算擬插值,進(jìn)而計(jì)算數(shù)值積分的計(jì)算量非常大,且計(jì)算過(guò)程冗長(zhǎng)。采用B網(wǎng)系數(shù)可大大簡(jiǎn)化計(jì)算過(guò)程。進(jìn)一步可將二元數(shù)值積分公式應(yīng)用于淺水動(dòng)力學(xué)模型的求解,研究非均勻三角剖分的三次樣條數(shù)值積分問(wèn)題。
感謝特拉華州立大學(xué)Shi Xiquan教授與匿名審稿專家提出的寶貴意見和建議。
[1]SCHUMAKER L L. Spline Functions:Basic Theory[M]. Malabar,F(xiàn)L: Krieger Publishing Company,1993.
[2]王國(guó)瑾,汪國(guó)昭,鄭建民. 計(jì)算機(jī)輔助幾何設(shè)計(jì)[M]. 北京:高等教育出版社,2001.
WANG G J, WANG G Z,ZHENG J M. Computer Aided Geometric Design[M]. Beijing: Higher Education Press,2001.
[3]FARIN G. Triangular Bernstein-Bézier patches[J]. Computer Aided Geometric Design, 1986,3(2): 83-127. DOI:10.1016/0167-8396(86)90016-6
[4]FARIN G. Curves and Surfaces for Computer-Aided Geometric Design: A Practical Guide[M]. San Diego: Academic Press Professional,1993.
[5]王仁宏. 多元齒的結(jié)構(gòu)與插值[J]. 數(shù)學(xué)學(xué)報(bào), 1975,18(2): 91-106.
WANG R H. Structure and interpolation of multiple teeth[J]. Acta Mathematica Sinica, 1975,18(2): 91-106.
[6]WANG R H. Multivariate Spline Functions and Their Applications[M]. Beijing: Science Press/ Dordrecht:Kluwer Academic Publishers, 2001.
[7]SHI X Q. The singularity of Morgan-Scott triangulation[J]. Computer Aided Geometric Design, 1991,8(3): 201-206. DOI:10.1016/0167-8396(91)90002-S
[8]XU Z Q, WANG R H. The instability degree in the dimension of spaces of bivariate spline[J]. Analysis in Theory and Applications, 2002,18(1): 68-80. DOI:10.1007/BF02837049
[9]WANG S M. Spline interpolations over type-2 triangulations[J]. Applied Mathematics and Computation, 1992,49(2/3):299-313. DOI:10. 1016/0096-3003(92)90031-U
[10]WANG S M, WANG C L. Smooth interpolations on some triangulations[J]. Utilitas Mathematica, 1992, 41:309-317.
[11]LIU H W, HONG D,CAO D Q. Bivariatecubic spline space over a nonuniform type-2 triangulation and its subspaces with boundary conditions[J]. Computers and Mathematics with Applications, 2005,49(11/12):1853-1865. DOI:10.1016/j.camwa.2004.08.014
[12]LI C J, WANG R H. Bivariate cubic spline space and bivariate cubic NURBS surfaces[C]// Proceedings of Geometric Modeling and Processing 2004. Beijing: IEEE,2004:115-123. DOI:10.1109/GMAP.2004. 1290033
[13]WANG R H, LI C J. Bivariate quartic spline spaces and quasi-interpolation operators[J]. Journal of Computational and Applied Mathematics, 2006,190(1): 325-338. DOI:10.1016/j.cam.2004.11.052
[14]QIAN J, WANG F. On the approximation of the derivatives of spline quasi-interpolation in cubic spline space[J]. Numerical Mathematics Theory Methods and Applications, 2014,7(1): 1-22. DOI:10.4208/nmtma.2014.y12035
[15]QIAN J, WANG R H,LI C J. The bases of the non-uniform cubic spline space[J]. Numerical Mathematics: Theory,Methods and Applications, 2012,5(4): 635-652. DOI:10.1017/S10048979 00001094
(下轉(zhuǎn)第頁(yè))
QIAN J, WANG R H,ZHU C G, et al. On spline quasi-interpolation in cubic spline space[J]. SCIENTIA SINICA Mathematica, 2014, 44(7):769-778. DOI:10.1360/N012013-00140
[17]ARTIOLI E, SOMMARIVA A,VIANELLO M. Algebraic cubature on polygonal elements with a circular edge[J]. Computers and Mathematics with Applications, 2020,79(7): 2057-2066. DOI:10. 1016/j.camwa.2019.10.022
[18]HU Q Y, XIA Y,HU P, et al. A concave-admissible quadrilateral quasi-conforming plane element using B-net method[J]. European Journal of Mechanics-A/Solids, 2016,57: 34-44. DOI:10. 1016/j.euromechsol.2015.12.001
[19]LAMBERTI P, SAPONARO A. Multilevel quadratic spline quasi-interpolation[J]. Applied Mathematics and Computation, 2020,373: 125047. DOI:10.1016/j.amc.2020.125047
[20]DAGNINO C, LAMBERTI P. On the approximation power of bivariate quadraticsplines[J]. Journal of Computational and Applied Mathematics, 2001,131(1): 321-332. DOI:10.1016/S0377-0427(00)00265-X
[21]DAGNINO C, LAMBERTI P. Some performances of local bivariate quadraticquasi-interpolating splines on nonuniform type-2 triangulations[J]. Journal of Computational and Applied Mathematics, 2005,173(1): 21-37. DOI:10.1016/j.cam.2004.02.017
[22]DAGNINO C, REMOGNA S. Quasi-interpolation based on the ZP-element for the numerical solution of integral equations on surfaces in[J]. BIT Numerical Mathematics, 2017, 57:329-350. DOI:10.1007/s10543-016-0633-x
[23]LAMBERTI P. Numerical integration based on bivariate quadratic spline quasi-interpolants on bounded domains[J]. BIT Numerical Mathematics, 2009, 49:565-588. DOI:10.1007/s10543-009-0237-9
[24]WANG R H, LU Y. Quasi-interpolating operators and their applications in hypersingular integrals[J]. Journal of Computational Mathematics, 1998,16(4): 337-344.
WANG R H, LU Y. Quasi-interpolating operators inon non-uniform type-2 triangulations[J]. Numerical Mathematics: A Journal of Chinese Universities,1999(2): 97-103.
[26]WANG G, LIANG Q H,ZHENG J H. A new multilayer nonhydrostatic formulation for surface water waves[J]. Journal of Coastal Research, 2019,35(3): 693-710. DOI:10.2112/JCOASTRES-D-18-00022.1
Numerical integration formulas of bivariate quadratic splines upon non-uniform type-2 triangulation
QIAN Jiang1,2, WANG Fan3
(1. College of Science,Hohai University,Nanjing211100,China;2. Key Laboratory of Coastal Disaster and Defence(Hohai University),Ministry of Education,Nanjing210098,China;3. College of Science,Nanjing Agricultural University,Nanjing210095,China)
In the paper, equivalent representations of the spline quasi-interpolation are presented over four triangular sub-domains contained in a general rectangular domain. Moreover, the direct bivariate numerical integration formulas are constructed over the four triangular sub-domains, while summing them yields the integration formula over the general rectangular domain. For illustration, the necessary function values and the corresponding integration coefficients are listed in several tables. Furthermore, based on the norm of the operator, the module of continuity and the reproduction of bivariate polynomials, error estimations of the numerical integration are derived for continuously differential functions with different orders. The computational cost of the proposed method is approximately 1/5 of that based on tensor-product-type quadratic splines. Numerical examples show the validity of the proposed numerical integration approach.
multivariate spline; conformality of smoothing cofactor method; bivariate numerical integration; spline quasi-interpolation; B-net
O 241.5
A
1008?9497(2022)05?555?09
2021?04?06.
江蘇省自然科學(xué)基金青年基金項(xiàng)目(BK20160853);河海大學(xué)中央高?;究蒲袠I(yè)務(wù)費(fèi)項(xiàng)目(2019B19414);海岸災(zāi)害與防護(hù)教育部重點(diǎn)實(shí)驗(yàn)室開放基金項(xiàng)目(河海大學(xué)202011);國(guó)家留學(xué)基金資助出國(guó)留學(xué)項(xiàng)目(訪問(wèn)學(xué)者201806715010).
錢江(1981—),ORCID:https://orcid.org/0000-0002-0526-5660,男,博士,副教授,主要從事數(shù)值逼近與計(jì)算幾何研究,E-mail:qianjianghhu@sina.com.
10.3785/j.issn.1008-9497.2022.05.006