摘要:

AbstractA forming model based on a viscoplastic flow formulation is derived which includes the effects of small elastic strains. A significant feature of the formulation is its reliance on the dominant inelastic material characteristics in the formation of the stiffness matrix for large strain problems. The resultant non‐linear system of equations is solved by an adaptive descent method which combines the rapid convergence of Newton's method near the solution with the robustness of a method of successive approximations. The use of the adaptive descent method effectively extends the viscoplastic flow formulations into the nearly rate‐insensitive range of behaviours exhibited, for example, by metals at low temperature, where slow convergence of the non‐linear solution algorithm has traditionally hampered thei

ISSN:0029-5981    

DOI:10.1002/nme.1620310602

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractBoundary integral formulations for the 2D Helmholtz equation involve kernels in the form of modified Bessel functions. Accurate schemes for evaluating integrals of the kernels and their derivatives are presented. Special attention is paid to integrals involving singular and near singular kernels. Both boundary and domain integrals are considered. It is shown that, with the use of series expansion functions for the modified Bessel functions, the boundary integrals can be evaluated analytically in the neighbourhood of the singularity. For domain integrals, the behaviour of the kernels in the vicinity of the singularity is used to construct accurate numerical quadrature schemes. A transient heat conduction problem is formulated as a Helmholtz equation, solved, and compared against analytic solution to demonstrate the effectiveness of these schemes in relation to traditional methods. References are made to previous work to advocate the utility of the boundary integral method for non‐linear and time‐transient probl

ISSN:0029-5981    

DOI:10.1002/nme.1620310603

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractA variational higher‐order theory involving all transverse strain and stress components is proposed for the analysis of laminated composite plates. Derived from three‐dimensional elasticity with emphasis on developing a viable computational methodology, the theory is well suited for finite element approximations as it incorporates bothC0andC−1continuous kinematic fields and Poisson boundary conditions. From the theory, a simple three‐node stretching‐bending finite element is developed and applied to the problem of cylindrical bending of a symmetric carbon/epoxy laminate for which an exact solution is available. Both the analytic and finite element results were found to be in excellent agreement with the exact solution for a wide range of the length‐to‐thickness ratio. The proposed higher‐order theory has the same computational advantages as first‐order shear‐deformable theories. The present methodology, however, provides greater predictive capabilities, especially, for thick

ISSN:0029-5981    

DOI:10.1002/nme.1620310604

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractThere is consensus in the literature that the eigenvalues of smallest absolute value from linear initial stability analyses of thin, elastic shells by the finite element method (FEM) do not posses bounding properties with respect to corresponding stability limites from geometrically non‐linear stability analyses. A ‘linear initial stability analysis’ by the FEM representes the first step of an ‘accompanying linear stability analysis’ by this method. In this paper, two modes of such stability analyses of thin, elastic shells will be presented. It will be proved that, for mode 1, in contrast to mode 2, bounding properties of eigenvalues of smallest absolute value with respect to corresponding stability limits from geometrically non‐linear stability analyses, in fact, do exist. Moreover, bounding properties of such eigenvalues from mode 1 relative to corresponding eigenvalues from mode 2 will be shown to exist. The existence of these properties is important from the standpoint of engineeri

ISSN:0029-5981    

DOI:10.1002/nme.1620310605

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractA procedure was implemented to determine some optimum values for the parameters of the four‐level‐time scheme which can be utilized for the numerical solution of a system of ordinary differential equations. A multi‐variable search method was employed to estimate these optimum parameters. Two different optimization studies were performed. The resultant optimum four‐level‐time schemes produced very accurate results, satisfied physical reality, and did not produce any oscillations for the specific problems that were studied. The optimum values for the three parameters α, β and γ in the four‐level‐time scheme were α = 16.72, β = 6.26 and γ = 2.46. The optimum value of θ for the Wilson‐θ method was θ = 1.42. The optimum schemes were compared to some existing four‐level‐time schemes. The outcome demonstrated that the optimum schemes were superior. The Houbolt method and the Galerkin weighting procedure were shown to have greater inaccurac

ISSN:0029-5981    

DOI:10.1002/nme.1620310606

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractGiven a list of points defining a domain boundary, a three‐stage process is often used to triangulate a domain. First, an appropriate distribution of interior points is generated. Next the points are connected to form triangles. And, finally, the connectivity data are used to reposition the interior points using the Laplacian smoothing technique, thereby usually improving the shapes of the shapes of the triangles. This paper describes a new techniques for mesh improvement—adjusting the connection structure during the second stage of this process. The new scheme, which we callmesh relaxation, consists of a procedure for iteratively making the mesh topology more regular by edge swapping. For each interior edge, arelaxation indexis computed that depends on thedegreesof its end points and adjacent points. Any edge for which this index exceed a prescribedthresholdwill be swapped, i.e. replaced by a new edge connecting the adjacent points of the original edge. After all edge swaps are completed, Laplacian smoothing is applied to the mesh. Example show that, when the mesh point density varies smoothly and due care is taken in the vicinity of the boundary, mesh relaxation can dramatically increase the regularity of the mesh and produce improve triangle sha

ISSN:0029-5981    

DOI:10.1002/nme.1620310607

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractA new Petrov‐Galerkin method for the incompressible Navier‐Stokes equations is presented. The use of the so‐called ‘optimal upwind’ parameter in multidimensions is justified by a time‐scale analysis of the relevant physical processes. The resulting procedure circumvents the Babuška‐Brezzi condition and allows equal order interpolation for velocity and press

ISSN:0029-5981    

DOI:10.1002/nme.1620310608

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractThe investigation of coupled fluid‐structure systems, subjected to dynamic loads, calls for the use of discrete techniques such as the finite element method (FEM) and the boundary element method (BEM). This paper reports on the development of a FEM‐BEM coupling procedure, where finite elements are used to model the linear elastic structure, while the adjacent fluid is represented by boundary elements. The new approach is formulated entirely in the time domain to be extensible to non‐linear problems.Various numerical calculations are presented which clearly demonstrate the applicability and the merits of the hybrid method. The obtained results are in excellent agreement with solutions stemming from a boundary element analysis which uses subdomain techn

ISSN:0029-5981    

DOI:10.1002/nme.1620310609

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractIn response to the need for an advanced computational model for wave propagation in jointed‐rock media a new finite element for jointed media with a single set of regularly spaced joints is developed. The element is a numerical implementation of the higher‐order homogenization model recently proposed by Murakami and Hegemier. Due to the dispersive effects induced by regularly spaced joints, wave phenomena in jointed media are altered significantly. Therefore, in order to improve the interpretation of seismograms for accurate source identification, it is necessary to develop a higher‐order continuum element. The accuracy and efficiency of the new element is investigated by applying it to wave‐guide and wave‐normal problems of a jointed half‐space and by comparing the wave response with that of DYNA2D. The analyses by DYNA2D discretize explicitly the details of the joint microstructure, and are adopted as numerically exact measures for the assessment of the proposed finite element; good correlations were obtained. The validation study also confirmed the importance of wave dispersion for non‐linear as well as linear joint responses. Finally, as a more practical application of the proposed element, the problem of a jointed full‐space with a cylindrical cavity pressurized by step and pulse loadings was solved. Velocities at several observation points were compared with the numerically exact results of DYNA2D. Similar analyses carried out for elastic isotropic media predicted totally different velocity responses and confirmed the need for the p

ISSN:0029-5981    

DOI:10.1002/nme.1620310610

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY

摘要:

AbstractA wholly new concept of intermediate structure is developed in this paper. Based on this, a new component mode iteration (CMI), or subdomain subspace iteration method, is proposed.In the new CMI method, the ordinarily required step of ensuring interface compatibility is eliminated. An outstanding achievement of this method is that it retains the merit of the standard subspace iteration (SSI) technique concerning the number of vectors required in computation. In the present method iteraton is carried out at subdomain levels. On machines capable of parallel processing, this work can be performed simultaneously.Numerical examples clearly demonstrate the excellence of this new method.

ISSN:0029-5981    

DOI:10.1002/nme.1620310611

出版商:John Wiley&Sons, Ltd    

年代:1991

数据来源: WILEY