Skip to main content
Log in

A paralleled element-free Galerkin analysis for structures with cyclic symmetry

  • Original Article
  • Published:
Engineering with Computers Aims and scope Submit manuscript

Abstract

This paper presents a paralleled element-free Galerkin (EFG) algorithm to analyze the structure with cyclic symmetry. The global EFG equation system is proved to be block-circulant under a corresponding symmetry-adapted coordinate system, and is partitioned into a number of smaller independent sub-problems, which are solved simultaneously on a multiprocessor parallel computer in parallel virtual machine. Numerical examples are given to illustrate the correctness, speedup, and efficiency of the proposed algorithm.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8

Similar content being viewed by others

References

  1. Belytschko T, Lu YY, Gu L (1994) Element-free Galerkin methods. Int J Numer Methods Eng 37:229–256

    Article  MATH  MathSciNet  Google Scholar 

  2. Belytschko T, Krongauz Y, Organ D, Fleming M, Krysl P (1996) Meshless methods: an overview and recent developments. Comput Methods Appl Mech Eng 139:3–47

    Article  MATH  Google Scholar 

  3. Lu YY, Belytschko T, Tabbara M (1995) Element-free Galerkin method for wave propagation and dynamic fracture. Comput Methods Appl Mech Eng 126:131–153

    Article  MATH  MathSciNet  Google Scholar 

  4. Zhong WX, Qiu CH (1983) Analysis of symmetric or partially symmetric structures. Comput Methods Appl Mech Eng 38:1–18

    Article  MATH  Google Scholar 

  5. Liu JX, Wu GF (1994) Parallel analysis of rotationally periodic structures. Comput Struct 50:785–795

    Article  Google Scholar 

  6. Thomas DL (1979) Dynamics of rotationally periodic structures. Int J Numer Methods Eng 14:81–102

    Article  MATH  Google Scholar 

  7. Williams FW (1986) Exact eigenvalue calculation for structures with rotationally periodic substructure. Int J Numer Methods Eng 23:695–706

    Article  MATH  Google Scholar 

  8. Wu GF, Williams FW, Kennedy D (1997) Exact eigenvalue calculation for partially rotationally periodic structures. Comput Struct 64:275–284

    Article  MATH  Google Scholar 

  9. Wu GF, Yang HT (1994) The use of cyclic symmetry in two-dimensional elastic stress analysis by BEM. Int J Solids Struct 31:279–290

    Article  MATH  Google Scholar 

  10. Geist A, Beguelin A, Dongarra J, Jiang W, Manchek R, Sunderam V (1994) PVM: Parallel virtual machine: a users’ guide and tutorial for networked parallel computing. MIT Press, Cambridge

    MATH  Google Scholar 

  11. Wang XC (2001) Parallel computing method in structural optimization design. JLU Press, China

    Google Scholar 

Download references

Acknowledgment

The research leading to this paper is funded by the oversea returnee’s initiating fund [1999-363], the key project fund [99149], and backbone faculty fund [2000-65], all of which came from National Education Department of P. R. China. The research is also funded by NSF (10421002), NSF (10472019), NSF (10172024), NKBRSF [2005CB321704], and the fund of disciplines leaders of young and middle age faculty in colleges of Liaoning Province.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Ling Liu.

Appendix: Proof of block-circulant property

Appendix: Proof of block-circulant property

Two indices are employed to denote an individual point. The superscript represents which symmetric region the point locates, and the subscript indicates which orbit the point belongs to, as illustrated in Fig. 9 where x i j , x i+t j are nodal coordinates, and x p q , x p+t q are coordinates of integration points. x i+t j and x p+t q can be matched by a rotation of angle t*θ from points x i j and x p q , respectively. t denotes rotation order, and θ = 2π/N was defined in Sect. 2.

Fig. 9
figure 9

An illustration of coordinates transformation

Based on EFG theory [1], shape function Φ I in Eq. 7 can be described by

$$ \Phi _{I} ={\mathbf{p}}^{\rm T} ({\mathbf{x}}){\mathbf{A}}^{{ - 1}}{\mathbf{B}}_{I} $$
(26)

where

$$ {\mathbf{p}} (x) = {\left\{ {1,x,y} \right\}}^{\rm T}, \quad {\mathbf{A}} = {\sum\limits_I {w ({\mathbf{x}} - {\mathbf{x}}_{I}){\mathbf{p}} ({\mathbf{x}}_{I}){\mathbf{p}}^{\rm T} ({\mathbf{x}}_{I}),}}\quad {\mathbf{B}}_{I} = w ({\mathbf{x}} - {\mathbf{x}}_{I}){\mathbf{p}} ({\mathbf{x}}_{I}) $$
(27)
$$ w ({\mathbf{x}}^{p}_{q} - {\mathbf{x}}^{i}_{j}) = w ({\mathbf{x}}^{{p + t}}_{q} - {\mathbf{x}}^{{i + t}}_{j}) $$
(28)

In a symmetric system, points belonging to the same orbit (see Fig. 9) follow

$$ {\mathbf{x}}^{{i + t}}_{j} ={\mathbf{T}}_{t} \cdot{\mathbf{x}}^{i}_{j},\quad {\mathbf{T}}_{t} =\left[ \begin{array}{*{20}c} {{\cos t\theta}}&{{ - \sin t\theta}}\\ {{\sin t\theta}}&{{\cos t\theta}} \end{array} \right] $$
(29)

and

$$ {\mathbf{T}}_{{i + k - 1}} ={\mathbf{T}}_{i} \cdot{\mathbf{T}}_{k} $$
(30)

Substituting Eq. 29 for 27 then yields

$${\mathbf{p}} ({\mathbf{x}}^{{i + t}}_{j}) = \overline{{\mathbf{\alpha}}} \cdot{\mathbf{p}} ({\mathbf{x}}^{i}_{j}),\quad {\mathbf{p}} ({\mathbf{x}}^{{p + t}}_{q}) = \overline{{\mathbf{\alpha}}} \cdot{\mathbf{p}} ({\mathbf{x}}^{p}_{q}),\quad \overline{{\mathbf{\alpha}}} =\left[{\begin{array}{*{20}c} {1}&{0} \\ {0}&{{{\mathbf{T}}_{t}}} \end{array}}\right] $$
(31)

Utilizing Eqs. 2629 and 31, one can obtain

$$\Phi ^{i}_{j} ({\mathbf{x}}^{p}_{q}) ={\mathbf{p}}^{\rm T} ({\mathbf{x}}^{p}_{q}){\mathbf{A}}^{{ - 1}}{\mathbf{B}}^{i}_{j} $$
(32)
$$ \begin{aligned} \Phi ^{{i + t}}_{j} ({\mathbf{x}}^{{p + t}}_{q}) &={\mathbf{p}}^{\rm T} ({\mathbf{x}}^{{p + t}}_{q}){\mathbf{A}}^{{ - 1}}{\mathbf{B}}^{{i + t}}_{j} \\ &={\mathbf{p}}^{\rm T} ({\mathbf{x}}^{{p + t}}_{q})\left[{\sum\limits_i{{\sum\limits_j{w ({\mathbf{x}}^{{p + t}}_{q} -{\mathbf{x}}^{{i + t}}_{j}){\mathbf{p}} ({\mathbf{x}}^{{i + t}}_{j}){\mathbf{p}}^{\rm T} ({\mathbf{x}}^{{i + t}}_{j})}}}}\right]^{{ - 1}} w ({\mathbf{x}}^{{p + t}}_{q} -{\mathbf{x}}^{{i + t}}_{j}){\mathbf{p}} ({\mathbf{x}}^{{i + t}}_{j}) \\ &={\mathbf{p}}^{\rm T} ({\mathbf{x}}^{p}_{q})\overline{{\mathbf{\alpha}}} ^{\rm T} \left[{\sum\limits_i{{\sum\limits_j{w ({\mathbf{x}}^{p}_{q} -{\mathbf{x}}^{i}_{j})\overline{{\mathbf{\alpha}}}{\mathbf{p}} ({\mathbf{ x}}^{i}_{j}){\mathbf{p}}^{\rm T} ({\mathbf{x}}^{i}_{j})}}}}\overline{{\mathbf {\alpha}}} ^{\rm T} \right]^{{ - 1}} w ({\mathbf{x}}^{p}_{q} -{\mathbf{x}}^{i}_{j})\overline{{\mathbf{\alpha}}}{\mathbf{p}} ({\mathbf{ x}}^{i}_{j}) \\ &={\mathbf{p}}^{\rm T} ({\mathbf{x}}^{p}_{q})\overline{{\mathbf{\alpha}}} ^{\rm T} (\overline{{\mathbf{\alpha}}} ^{\rm T})^{{ - 1}} \left[{\sum\limits_i{{\sum\limits_j{w ({\mathbf{x}}^{p}_{q} -{\mathbf{x}}^{i}_{j}){\mathbf{p}} ({\mathbf{x}}^{i}_{j}){\mathbf{p}}^{\rm T} ({\mathbf{x}}^{i}_{j})}}}}\right]^{{ - 1}} \overline{{\mathbf{\alpha}}} ^{{ - 1}} \overline{{\mathbf{\alpha}}} w ({\mathbf{x}}^{p}_{q} -{\mathbf{x}}^{i}_{j}){\mathbf{p}} ({\mathbf{x}}^{i}_{j}) \\ &={\mathbf{p}}^{\rm T} ({\mathbf{x}}^{p}_{q})\left[{\sum\limits_i{{\sum\limits_j{w ( {\mathbf{x}}^{p}_{q} -{\mathbf{x}}^{i}_{j}){\mathbf{p}} ({\mathbf{x}}^{i}_{j}){\mathbf{p}}^{\rm T} ({\mathbf{x}}^{i}_{j})}}}}\right]^{{ - 1}} w ({\mathbf{x}}^{p}_{q} -{\mathbf{x}}^{i}_{j}){\mathbf{p}} ({\mathbf{x}}^{i}_{j}) \\ &={\mathbf{p}}^{\rm T} ({\mathbf{x}}^{p}_{q}){\mathbf{A}}^{{ - 1}}{\mathbf{B}}^{i}_{j} \\ \end{aligned} $$
(33)

therefore

$$ \Phi ^{i}_{j} (x^{p}_{q}) = \Phi ^{{i + t}}_{j} (x^{{p + t}}_{q}) $$
(34)
$$ \frac{\partial}{{\partial x^{i}}}\Phi ^{i}_{j} \left|{_{{x^{t}_{q}}}} \right. = \frac{\partial}{{\partial x^{{i + t}}}}\left.{\Phi ^{{i + t}}_{j}} \right|_{{{\mathbf{x}}^{{p + t}}_{q}}} = \left(\cos (t\theta)\frac{{\partial}}{{\partial x^{i}}} + \sin (t\theta)\frac{\partial}{{\partial y^{i}}}\right)\left.{\Phi ^{{i + t}}_{j}} \right|_{{{\mathbf{x}}^{{p + t}}_{q}}} $$
(35)
$$ \frac{\partial}{{\partial y^{i}}}\Phi ^{i}_{j} \left|{_{{x^{p}_{q}}}} \right. = \frac{\partial}{{\partial y^{{i + t}}}}\left.{\Phi ^{{i + t}}_{j}} \right|_{{{\mathbf{x}}^{{p + t}}_{q}}} = \left( - \sin (t\theta)\frac{\partial}{{\partial x^{i}}} + \cos (t\theta)\frac{\partial}{{\partial y^{i}}}\right)\left.{\Phi ^{{i + t}}_{j}} \right|_{{{\mathbf{x}}^{{p + t}}_{q}}} $$
(36)
$$ \frac{\partial}{{\partial x^{i}}}\left.{\Phi ^{{i + t}}_{j}} \right|_{{{\mathbf{x}}^{{p + t}}_{q}}} = \left(\cos (t\theta)\frac{\partial}{{\partial x^{i}}} - \sin (t\theta)\frac{\partial}{{\partial y^{i}}}\right)\left.{\Phi ^{i}_{j}} \right|_{{{\mathbf{x}}^{p}_{q}}} $$
(37)
$$ \frac{\partial}{{\partial y^{i}}}\left.{\Phi ^{{i + t}}_{j}} \right|_{{{\mathbf{x}}^{{p + t}}_{q}}} = \left(\sin (t\theta)\frac{\partial}{{\partial x^{i}}} + \cos (t\theta)\frac{\partial}{{\partial y^{i}}}\right)\left.{\Phi ^{i}_{j}} \right|_{{{\mathbf{x}}^{p}_{q}}} $$
(38)

Based on Eqs. 37, 38, one can obtain

$$ \left.({\mathbf{L}}^{\rm T} \Phi ^{i}_{j}){\mathbf{D}} ({\mathbf{L}}\Phi ^{k}_{l})\right|_{{{\mathbf{x}}^{p}_{q}}} =\left[ \begin{array}{*{20}c} {{xi}}&{0}&{{yi}} \\ {0}&{{yi}}&{{xi}} \end{array} \right] \cdot{\mathbf{D}} \left[{ \begin{array}{*{20}c} {{xk}}&{0} \\ {0}&{{xk}}\\ {{yk}}&{{xk}} \end{array}} \right] $$
(39)
$$ \begin{aligned} \left.{\mathbf{\alpha}}^{\rm T} ({\mathbf{L}}^{\rm T} \Phi ^{{i + t}}_{j}){\mathbf{D}} ({\mathbf{L}}\Phi ^{{k + t}}_{l}){\mathbf{\alpha}}\right|_{{{\mathbf{x}}^{{p + 1}}_{q}}} & =\left[ \begin{array}{*{20}c} {{\cos (t\theta)}}&{{\sin (t\theta)}} \\ {{- \sin (t\theta)}}&{{\cos (t\theta)}} \end{array}\right] \\ & \quad \times \left[ \begin{array}{*{20}c} {{\cos (t\theta) \cdot xi -\sin (t\theta) \cdot yi}}&{0}&{{\sin (t\theta) \cdot xi + \cos (t\theta) \cdot yi}} \\ {0}&{{\sin (t\theta) \cdot xi + \cos (t\theta) \cdot yi}}&{{\cos (t\theta) \cdot xi - \sin (t\theta) \cdot yi}} \end{array}\right] \\ & \quad \times{\mathbf{D}} \left[ \begin{array}{*{20}c} {{\cos (t\theta) \cdot xk -\sin (t\theta) \cdot yk}}&{0} \\ {0}&{{\sin (t\theta) \cdot xk +\cos (t\theta) \cdot yk}} \\ {{\sin (t\theta) \cdot xk + \cos (t\theta)\cdot yk}}&{{\cos (t\theta) \cdot xk - \sin (t\theta) \cdot yk}} \end{array} \right] \left[ \begin{array}{*{20}c} {{\cos (t\theta)}}&{{- \sin (t\theta)}} \\ {{\sin (t\theta)}}&{{\cos (t\theta)}} \end{array} \right] \\ \end{aligned} $$
(40)

where

$$ xi = \frac{\partial}{{\partial x^{i}}}\left.{\Phi ^{i}_{j}} \right|_{{{\mathbf{x}}^{p}_{q}}} ,\quad yi = \frac{\partial}{{\partial y^{i}}}\left.{\Phi ^{i}_{j}} \right|_{{{\mathbf{x}}^{p}_{q}},}\quad{\mathbf{\alpha}} ={\mathbf{T}}_{t} $$
(41)

Comparing Eq. 39 with Eq. 40 in details then gives

$$ \left.{ ({\mathbf{L}}^{\rm T} \Phi ^{i}_{j}){\mathbf{D}} ({\mathbf{L}}\Phi ^{k}_{l})} \right|_{{{\mathbf{x}}^{p}_{q}}} = \left.{{\mathbf{\alpha}}^{\rm T} ({\mathbf{L}}^{\rm T} \Phi ^{{i + t}}_{j}){\mathbf{D}} ({\mathbf{L}}\Phi ^{{k + t}}_{l}){\mathbf{\alpha}}} \right|_{{{\mathbf{x}}^{{p + t}}_{q}}} $$
(42)

therefore

$$ {\mathbf{T}}^{\rm T}_{p} \cdot{\mathbf{K}}^{{pk}}_{{ql}} \cdot{\mathbf{T}}_{k} ={\mathbf{T}}^{\rm T}_{{p + t}} \cdot{\mathbf{K}}^{{p + t,k + t}}_{{ql}} \cdot{\mathbf{T}}_{{k + t}} $$
(43)
$$ \overline{{\mathbf{T}}} ^{\rm T}_{p} \cdot{\mathbf{K}}^{{pk}} \cdot \overline{{\mathbf{T}}} _{k} = \overline{{\mathbf{T}}} ^{\rm T}_{{p + t}} \cdot{\mathbf{K}}^{{p + t,k + t}} \cdot \overline{{\mathbf{T}}} _{{k + t}} $$
(44)
$$ \overline{{\mathbf{K}}} ^{{ik}} = \overline{{\mathbf{K}}} ^{{i +t,k + t}} $$
(45)

Similarly, it can be proved that

$$ \overline{{\mathbf{G}}} ^{{ik}} = \overline{{\mathbf{G}}} ^{{i + t,k + t}} $$
(46)

Rights and permissions

Reprints and permissions

About this article

Cite this article

Liu, L., Yang, H. A paralleled element-free Galerkin analysis for structures with cyclic symmetry. Engineering with Computers 23, 137–144 (2007). https://doi.org/10.1007/s00366-006-0050-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00366-006-0050-x

Keywords

Navigation