NODAL ACCURACY IMPROVEMENT AND SUPER-CONVERGENT COMPUTATION IN FEM ANALYSIS OF FEMOL SECOND ORDER ODES
-
摘要: 采用
$ m $ 次单元对线法二阶常微分方程组(ODEs)进行有限元(FEM)求解,其单元内部位移为$ m + 1 $ 阶收敛,而端结点位移收敛阶可达$ 2m $ 阶。单元能量投影(EEP)超收敛计算恢复的单元内部位移精度一般为$\min $ $ (m + 2,2m)$ 阶,此收敛阶既受益于也受限于有限元端结点位移的精度。该文提出了一种修正EEP法(M-EEP),利用EEP超收敛解,先对端结点位移进行修正,再用其恢复单元内部位移。广泛的数值试验表明:对端结点位移修正后的收敛阶可达$ 2m + 2 $ 阶,再次修复的单元内部位移始终可达$ m + 2 $ 阶收敛,摆脱了$ 2m $ 阶收敛精度的限制。对于线性元,修正后结点位移的精度翻倍,单元内部M-EEP位移亦摆脱了原FEM解2阶收敛精度的限制,升到3阶收敛,基本达到二次元的收敛精度,效果显著。-
关键词:
- 有限元法 /
- 二阶常微分方程组 /
- 超收敛 /
- 单元能量投影(EEP) /
- 修正的EEP法(M-EEP) /
- 有限元线法(FEMOL)
Abstract: Elements with degree$ m $ is used in finite element method (FEM) to solve the second order ordinary differential equations (ODEs) derived from the FEM of lines (FEMOL). The interior displacement of elements generally has a convergence order of$ m + 1 $ , while the nodal displacements can achieve a convergence order of$ 2m $ . The super-convergence computation using the element energy projection (EEP) method usually has a convergence order of$ \min (m + 2,2m) $ , which benefits from the nodal displacements of a higher convergence order but also limits its accuracy by the nodal displacements of elements with lower degrees. In this paper, a modified EEP (M-EEP) method is proposed. With the EEP solution, the nodal displacement accuracy is improved first, and then the interior displacement of elements is recovered, which leads to a modified EEP solution. Numerical experiments show that improved nodal displacements can achieve a convergence order of$ 2m + 2 $ , and the interior displacements of elements always have a convergence order of$ m + 2 $ without the constraint of order$ 2m $ . For linear elements, the interior displacement of M-EEP solution does not have the limitation of second-order convergence from the traditional FEM solution and can achieve the remarkable third-order convergence, equivalent to the convergence order of quadratic elements. -
表 1 常系数问题端结点位移收敛阶
Table 1. Convergence orders of nodal displacement of constant coefficient problem
单元次数$ m $ 单元数$ {N_e} $ 有限元解 修正解 误差$ e_{i,{\kern 1pt} \max }^h $ 收敛阶ρ 误差$ \tilde e_{i,{\kern 1pt} \max }^h $ 收敛阶ρ 1 4 3.68×10−3 − 2.38×10−4 − 8 9.10×10−4 2.01 2.11×10−5 3.50 16 2.27×10−4 2.00 1.40×10−6 3.91 32 5.67×10−5 2.00 7.82×10−8 4.16 64 1.42×10−5 2.00 4.76×10−9 4.04 2 4 5.62×10−5 − 1.03×10−5 − 8 5.17×10−6 3.44 3.16×10−7 5.03 16 3.36×10−7 3.94 5.34×10−9 5.89 32 2.05×10−8 4.04 8.25×10−11 6.01 64 1.29×10−9 3.99 1.31×10−12 5.98 3 4 3.18×10−6 − 3.54×10−7 − 8 9.71×10−8 5.03 3.37×10−9 6.71 16 1.62×10−9 5.91 1.54×10−11 7.78 32 2.48×10−11 6.03 5.92×10−14 8.02 64 3.92×10−13 5.98 2.30×10−16 8.01 收敛阶 $ {{2}}{m} $ $ {{2}}{m}{{ + 2}} $ 表 2 常系数问题内部结点位移收敛阶
Table 2. Convergence orders of interior nodal displacement of constant coefficient problem
单元次数$ m $ 单元数$ {N_e} $ 有限元解 EPP解 M-EEP解 误差$ e_{\max }^h $ 收敛阶ρ 误差$ e_{\max }^{\text{*}} $ 收敛阶ρ 误差$ \tilde e_{\max }^* $ 收敛阶ρ 1 4 1.30×10−2 − 3.68×10−3 − 1.08×10−3 − 8 3.54×10−3 1.88 9.11×10−4 2.02 1.08×10−4 3.31 16 9.47×10−4 1.90 2.27×10−4 2.00 9.24×10−6 3.55 32 2.57×10−4 1.88 5.67×10−5 2.00 7.07×10−7 3.71 64 6.69×10−5 1.94 1.42×10−5 2.00 5.21×10−8 3.76 2 4 6.07×10−4 − 7.39×10−5 − 4.61×10−5 − 8 1.02×10−4 2.58 6.63×10−6 3.48 3.04×10−6 3.92 16 1.58×10−5 2.69 4.04×10−7 4.04 1.32×10−7 4.52 32 2.16×10−6 2.87 2.28×10−8 4.15 5.02×10−9 4.72 64 2.82×10−7 2.93 1.36×10−9 4.06 1.72×10−10 4.87 3 4 4.72×10−5 − 1.02×10−5 − 4.12×10−6 − 8 4.65×10−6 3.34 4.21×10−7 4.60 1.44×10−7 4.84 16 3.73×10−7 3.64 1.09×10−8 5.27 3.42×10−9 5.39 32 2.83×10−8 3.72 2.20×10−10 5.64 6.58×10−11 5.70 64 1.97×10−9 3.85 3.90×10−12 5.82 1.14×10−12 5.85 收敛阶 $ {m}{{ + 1}} $ ${ {{\rm{min}}(2} }{m},{m}{ { + 3)} }$ $ {m}{{ + 3}} $ 表 3 变系数问题端结点位移收敛阶
Table 3. Convergence orders of nodal displacement of variable coefficient problem
单元次数$ m $ 单元数$ {N_e} $ 有限元解 修正解 误差$ e_{i,{\kern 1pt} \max }^h $ 收敛阶ρ 误差$ \tilde e_{i,{\kern 1pt} \max }^h $ 收敛阶ρ 1 32 1.54×10−4 − 8.76×10−7 − 64 3.86×10−5 2.00 5.37×10−8 4.03 128 9.66×10−6 2.00 3.32×10−9 4.01 256 2.42×10−6 2.00 2.08×10−10 4.00 512 6.04×10−7 2.00 1.30×10−11 4.00 2 32 2.04×10−7 − 1.59×10−9 − 64 1.26×10−8 4.02 2.46×10−11 6.01 128 7.83×10−10 4.00 3.84×10−13 6.00 256 4.90×10−11 4.00 6.01×10−15 6.00 512 3.06×10−12 4.00 9.39×10−17 6.00 3 32 4.69×10−10 − 1.71×10−12 − 64 7.28×10−12 6.01 6.64×10−15 8.01 128 1.14×10−13 5.99 2.62×10−17 7.99 256 1.78×10−15 6.00 1.02×10−19 8.00 512 2.79×10−17 6.00 3.99×10−22 8.00 收敛阶 $ {{2}}{m} $ $ {{2}}{m}{{ + 2}} $ 表 4 变系数问题内部结点位移收敛阶
Table 4. Convergence orders of interior nodal displacement of variable coefficient problem
单元次数$ m $ 单元数$ {N_e} $ 有限元解 EEP解 M-EEP解 误差$ e_{\max }^h $ 收敛阶ρ 误差$ e_{\max }^{\text{*}} $ 收敛阶ρ 误差$ \tilde e_{\max }^*$ 收敛阶ρ 1 32 7.45×10−4 − 1.54×10−4 − 7.27×10−6 − 64 1.98×10−4 1.91 3.86×10−5 2.00 5.14×10−7 3.82 128 5.14×10−5 1.95 9.66×10−6 2.00 4.88×10−8 3.40 256 1.31×10−5 1.97 2.42×10−6 2.00 5.97×10−9 3.03 512 3.31×10−6 1.99 6.04×10−7 2.00 7.38×10−10 3.02 2 32 1.12×10−5 − 2.10×10−7 − 5.68×10−8 − 64 1.51×10−6 2.89 1.26×10−8 4.06 2.06×10−9 4.78 128 1.96×10−7 2.94 7.83×10−10 4.00 6.97×10−11 4.89 256 2.36×10−8 3.05 4.90×10−11 4.00 3.89×10−12 4.16 512 2.98×10−9 2.99 3.06×10−12 4.00 2.44×10−13 4.00 3 32 2.58×10−7 − 3.51×10−9 − 1.10×10−9 − 64 1.80×10−8 3.84 6.51×10−11 5.75 1.97×10−11 5.80 128 1.19×10−9 3.92 1.11×10−12 5.88 3.31×10−13 5.90 256 6.73×10−11 4.14 2.32×10−14 5.58 6.95×10−15 5.57 512 4.26×10−12 3.98 7.32×10−16 4.99 2.19×10−16 4.98 收敛阶 $ {m}{{ + 1}} $ $ {{{\rm{min}}(2}}{m},{m}{{ + 2)}} $ $ {m}{{ + 2}} $ 表 5 变系数问题端结点应力函数收敛阶
Table 5. Convergence orders of nodal stress function of variable coefficient problem
单元次数$ m $ 单元数$ {N_e} $ 有限元解 修正解 误差$ e_{i,{\kern 1pt} \max }^h $ 收敛阶ρ 误差$ \tilde e_{i,{\kern 1pt} \max }^h $ 收敛阶ρ 1 4 1.44×10−2 − 6.21×10−3 − 8 3.50×10−3 2.04 1.37×10−3 2.18 16 1.22×10−3 1.52 2.71×10−4 2.33 32 3.87×10−4 1.66 3.37×10−5 3.01 64 9.27×10−5 2.06 2.58×10−6 3.70 表 6 变系数问题内部结点应力函数收敛阶
Table 6. Convergence orders of interior nodal stress function of variable coefficient problem
单元次数$ m $ 单元数$ {N_e} $ 有限元解 EEP解 M-EEP解 误差$ e_{\max }^h $ 收敛阶ρ 误差$ e_{\max }^{\text{*}} $ 收敛阶ρ 误差$ \tilde e_{\max }^* $ 收敛阶ρ 1 4 4.44×10−2 − 3.59×10−2 − 3.07×10−2 − 8 2.17×10−2 1.03 8.81×10−3 2.03 7.60×10−3 2.01 16 1.04×10−2 1.07 1.88×10−3 2.23 1.66×10−3 2.19 32 4.19×10−3 1.31 4.33×10−4 2.12 2.85×10−4 2.54 64 1.42×10−3 1.56 9.83×10−5 2.14 3.85×10−5 2.89 -
[1] 陈传淼. 有限元超收敛构造理论[M]. 长沙: 湖南科学技术出版社, 2001.Chen Chuanmiao. Structure theory of superconvergence of finite elements [M]. Changsha: Hunan Science & Technology Press, 2001. (in Chinese) [2] Douglas J. Galerkin approximations for the two point boundary problem using continuous, piecewise polynomial spaces [J]. Numerische Mathematik, 1973, 22(2): 99 − 109. [3] Zienkiewicz C, Zhu J Z. The superconvergent patch recovery (SPR) and a posterior error estimates. Part 1: The recovery technique [J]. International Journal for Numerical Methods in Engineering, 1992, 33: 1331 − 1364. doi: 10.1002/nme.1620330702 [4] Zienkiewicz C, Zhu J Z. The superconvergent patch recovery (SPR) and a posterior error estimates. Part 2: Error estimates and adaptivity [J]. International Journal for Numerical Methods in Engineering, 1992, 33: 1365 − 1382. [5] Babuška I, Rheinboldt W C. Reliable error estimation and mesh adaptation for the finite element method [J]. Computational Methods in Nonlinear Mechanics, 1980: 67 − 108. [6] Bank R E, Weiser A. Some a posteriori error estimators for elliptic partial differential equations [J]. Math Comput, 1985, 44: 283 − 301. [7] 袁驷, 邢沁妍, 袁全. 基于EEP技术的一维有限元结点位移误差计算[J]. 工程力学, 2020, 37(9): 1 − 7, 29. doi: 10.6052/j.issn.1000-4750.2020.07.0446Yuan Si, Xing Qinyan, Yuan Quan. Calculation of errors of nodal displacements in one-dimensional finite element methods using element energy projection technique [J]. Engineering Mechanics, 2020, 37(9): 1 − 7, 29. (in Chinese) doi: 10.6052/j.issn.1000-4750.2020.07.0446 [8] Yuan Si. The finite element method of lines [M]. Beijing-New York: Science Press, 1993. [9] Strang G, Fix G J. An analysis of the finite element method (Second Edition) [M]. Wellesley MA: Wellesley- Cambridge Press, 2008. [10] 袁驷, 王枚. 一维有限元后处理超收敛解答计算的 EEP 法[J]. 工程力学, 2004, 21(2): 1 − 9. doi: 10.3969/j.issn.1000-4750.2004.02.001Yuan Si, Wang Mei. An element-energy-projection method for post-computation of super-convergent solutions in one-dimensional FEM [J]. Engineering Mechanics, 2004, 21(2): 1 − 9. (in Chinese) doi: 10.3969/j.issn.1000-4750.2004.02.001 [11] 袁驷, 肖嘉, 叶康生. 线法二阶常微分方程组有限元分析的EEP超收敛计算[J]. 工程力学, 2009, 26(11): 1 − 9, 22.Yuan Si, Xiao Jia, Ye Kangsheng. EEP super-convergent computation in FEM analysis of FEMOL second order ODEs [J]. Engineering Mechanics, 2009, 26(11): 1 − 9, 22. (in Chinese) [12] 袁驷, 王枚, 王旭. 二维有限元线法超收敛解答计算的EEP法[J]. 工程力学, 2007, 24(1): 1 − 10. doi: 10.3969/j.issn.1000-4750.2007.01.001Yuan Si, Wang Mei, Wang Xu. An element-energy projection method for super-convergence solutions in two-dimensional finite element method of lines [J]. Engineering Mechanics, 2007, 24(1): 1 − 10. (in Chinese) doi: 10.3969/j.issn.1000-4750.2007.01.001 [13] Yuan S, Wu Y, Xu J J, et al. A super-convergence strategy for two-dimensional FEM based on element energy projection technique [J]. Journal of Nanoelectronics and Optoelectronics, 2017, 12(11): 1284 − 1294. doi: 10.1166/jno.2017.2272 [14] Yuan S, Wu Y, Xing Q Y. Recursive super-convergence computation for multi-dimensional problems via one-dimensional element energy projection technique [J]. Applied Mathematics and Mechanics, 2018, 39(7): 1031 − 1044. doi: 10.1007/s10483-018-2345-7 -