• Computational Structural Engineering
• /
• v.2 no.4
• /
• pp.15-20
• /
• 1989

이글에서는 2가지 예제를 통해 h-version과 p-version의 비교를 살펴보면서 p-version 해석이 h-version에 비해 상대적으로 많은 장점들을 가지고 있으며, 신뢰도, 정확도, 효율성, 경제성, 용장성 등 측면에서 우월함을 증명해 보였다. 특히 응력집중(stress concentration)이 일어나는 crack-tips, cut-outs, reentrant corners, presence of stiffners, mixed boundary conditions 등 많은 특이성(singularity) 문제에 더욱 적합함을 본 예제 외의 발표된 많은 논문들을 통해 알 수 있으며, 모델링의 단순성에 기인하여 사용이 매우 쉽다는 것도 무엇보다 큰 이점이라 하겠다. p-version은 h-version의 비효율성을 차수 p를 1, 2 또는 3으로 줄인 후 이 값을 고정시키고 다시 요소분할을 통해 진해(true solution)에 접근시키는 방식을 위하면 다시 종래의 h-version으로 환원되는 호환성을 갖고 있다는 것이다. 고로 구조해석에서 h-p version이 가장 이상적인 유한요소해석 방법이라 할 수 있겠는데, 다시 말하면 균열문제의 경우 균열선단(crack-tip)에서는 p-level을 높이고 (p=8, 9 or 10) 비교적 응력집중이 낮은 영역에서는 p-level을 낮춤으로써 (p=3, 4 or 5) 그 효율성을 극대화할 수 있겠다.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.14 no.4
• /
• pp.729-740
• /
• 1994

In the shape optimal design based on h-version of FEM, the ideal mesh for the initial geometry most probably will not be suitable for the final analysis. Thus, it is necessary to remesh the geometry of the model at each stage of optimization. However, the p-version of FEM appears to be a very attractive alternative for use in shape optimization. The main advantages are as follows; firstly, the elements are not sensitive to distortion for interpolation polynomials of order $p{\geq}3$; secondly, even singular problems can be solved more efficiently with p-version than with the h-version by proper mesh design; thirdly, the initial mesh design are identical. The 2-D p-version model for shape optimization is presented on the basis of Bezier's curve fitting, gradient projection method, and integrals of Legendre polynomials. The numerical results are performed by p-version software RASNA.

• Computational Structural Engineering
• /
• v.5 no.1
• /
• pp.75-82
• /
• 1992

In the h-version of F.E.M., all piecewisely smooth curved boundaries can be approximated by a sufficient number of straight-sided elements. However, in the p-version the size of the element is usually large and hence the probability of distortions is more. An attempt has been made to generate a curved boundary by using a transfinite interpolation technique to avoid the discretization errors. In the following sections, it will be shown how to construct transfinite interpolants both in h-version and in p-version over polygonal and nonpolygonal regions. Three numerical tests are shown to validate the applicability and superior capability of transfinite interpolation technique.

• The Pure and Applied Mathematics
• /
• v.12 no.2 s.28
• /
• pp.153-159
• /
• 2005

We investigate the error estimates of the h and p versions of the finite element method for an elliptic problems. We present theoretical results showing the p version gives results which are not worse than those obtained by the h version in the finite element method.

• Korean Journal of Mathematics
• /
• v.8 no.1
• /
• pp.47-52
• /
• 2000

We describe some results for the $h$ version which pertain to the questions on numerical quadrature. We also present an example that illustrates the rate of convergence predicted for linear elements under certain quadrature schemes in [2], [4], [5].

• Computational Structural Engineering
• /
• v.3 no.3
• /
• pp.101-111
• /
• 1990

The elastic analysis of floor slabs using the p-version of finite element method encounters stress singularities at certain types of reentrant corners, openings and cut-outs. Results obtained using the computer code based on C.deg. - hierarchic plate element formulated by Reissner-Mindlin theory are compared with theoretical predictions and with computational results reported in the literature. The convergence rate of h-, p- and hp-version can be estimated on the basis of the energy norm in global sense. If accuracy in terms of the number of degree-of-freedom is used as a criterion, the solutions presented here are the most efficient that have been published up to date. Examples are the rhombic plate with the obtuse angle of 150.deg. and the square plate with cut-outs.

• Korean Journal of Mathematics
• /
• v.6 no.2
• /
• pp.195-203
• /
• 1998

In [3], we showed that overintegration may be needed to obtain the optimal $H^1$ error rate for the $p$ version. In this paper, we examine the convergence of the $p$ version in practice, and comment on the implementation of the $p$ version in commercial codes. Also, we give an example of a problem with extremely rough coefficients, for which overintegration is necessary to obtain the optimal $H^1$ convergence rate.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.13 no.2
• /
• pp.151-160
• /
• 1993

This paper is concerned with formulations of the hierarchical $C^{\circ}$-plate element on the basis of Reissner-Mindlin plate theory. On reason for the development of the aforementioned element based on Integrals of Legendre shape functions is that it is still difficult to construct elements based on h-version concepts which are accurate and stable against the shear locking effects. An adaptive mesh refinement and selective p-distribution of the polynomial degree using hp-version of the finite element method are proposed to verify the superior convergence and algorithmic efficiency with the help of the simply supported L-shaped plate problems.

• Computational Structural Engineering
• /
• v.6 no.4
• /
• pp.57-66
• /
• 1993

The p-version crack model based on integrals of Legendre polynomial and virtual crack extension method is proposed with its potential for application to stress intensity factor computations in linear elastic fracture mechanics. The main advantage of this model is that the data preparation effort is minimal because only a small number of elements are used and high accuracy and the rapid convergence can be achieved in the vicinity of crack tip. There are two important findings from this study. Firstly, the limit value, the strain energy of the exact solution, can be estimated with successive three p-version approximations by ascertaining that the approximations enter the asymptotic range. Secondly, the rate of convergence of p-version model is almost twice that of h-version model on the basis of uniform or quasiuniform mesh refinement for the cracked panel problem subjected to tension.

• Journal of applied mathematics & informatics
• /
• v.9 no.1
• /
• pp.311-320
• /
• 2002

For second order linear elliptic problems over smooth domains, it is well known that the rate of convergence of the error in the $L_2$norm is one order higher than that in the $H^1$norm. For polygonal domains with reentrant corners, it has been shown in [15] that this extra order cannot be fully recovered when the h-version is used. We present theoretical and computational examples showing the sharpness of our results.