Eccentricity effects in the finite element modelling of composite beams

https://doi.org/10.1016/j.advengsoft.2012.06.004Get rights and content

Abstract

When modelling composite or built up beams using finite element software, analysts find it often convenient to connect two standard Euler–Bernoulli beam elements at the nodes by using a rigid bar or use master–slave type kinematic constraints to express the degrees-of-freedoms of one of the members in terms of the other. However, this type of modelling leads to eccentricity related numerical errors and special solutions that avoid eccentricity related issues may not be available for a design engineer due to the limitations of the software. In this study, a simple correction technique is introduced in the application of master–slave type constraints. It is shown that the eccentricity related numerical errors in the stiffness matrix can be completely corrected by using extra fictitious elements and springs. The correction terms are obtained by using the exact homogenous solution of the composite beam problem as the interpolation functions which impose the zero-slip constraint between the two components in the point-wise sense. The effects of the eccentricity related errors are demonstrated in numerical examples.

Highlights

► A correction technique is introduced to avoid eccentricity related numerical errors in the modelling of composite beams. ► It is shown that the error can be significant in practical applications. ► Effect of loading type on the eccentricity related numerical errors is illustrated. ► Effect of relative layer stiffness on the numerical error is illustrated.

Introduction

Composite beams and laminated plates that consist of components juxtaposed with a shear connection find widespread applications in structural engineering. A mathematical model for composite beams with flexible shear connectors was introduced by Newmark et al. [1], in which two Euler–Bernoulli beams are connected by assuming that vertical separation does not occur between the components. Subsequently, several displacement-based finite element formulations were developed based on Newmark’s model, which include the works of Arizumi et al. [2], Daniels and Crisinel [3], Ranzi et al. [4], [5] and Dall’Asta and Zona [6]. However, in many practical cases the interlayer connections are very stiff between the two components such that the interlayer slip is negligible. In such cases, displacement-based finite element formulations based on flexible shear connectors may suffer from the so-called slip-locking phenomenon because of the coupling between the displacement fields in the discretized form of Newmark’s model [7]. Erkmen and Bradford [8] used the kinematic interpolatory strategy and assumed strain-mixed formulations to alleviate locking behaviour for stiff connections. An exact finite element formulation is also presented in [8]. A mesh free approach that eliminates slip locking was recently developed by Erkmen and Bradford [9].

Special solutions, however, are not always available for a design engineer in which case the modelling options are limited to standard conventional beam type finite elements. Therefore, a convenient practice in the modelling of composite beams especially for the cases with stiff connections is to connect the two conventional beam type finite element components by using a rigid bar to connect the end nodes of the two components or use master–slave type kinematic constraints to express the nodal degrees-of-freedom of one of the members in terms of the other. However, this type of modelling leads to eccentricity related numerical errors as initially shown by Gupta and Ma [10] in a composite cantilever beam problem. A similar type of error in multiple-point constraint applications for built-up plates and shells was pointed out by Crisfield [11]. Recently, Erkmen et al. [12] adopted variational multiscale approach to correct numerical errors in the applications of master–slave type constraints when forming composite beams.

In this study, a simple correction technique is introduced for the applications of master–slave type constraints to form composite beams and it is shown that the eccentricity related numerical errors in the stiffness matrix can be completely corrected by using extra fictitious elements and springs. The correction terms are obtained by selecting interpolation functions that do not violate the zero-slip constraint between the components in the point-wise sense. Since numerical errors are avoided the developed solution provides identical results to those that can be obtained using the elementary solution for composite sections, e.g. transformed area method in classical mechanics of solids books [13].

Section snippets

Displacements and strains

The composite member is made up of a top and a bottom Euler–Bernoulli beam elements, which are referred to as members 1 and 2, respectively. Deformations of each component can be expressed in terms of the axial displacements wi of the centroids and the vertical displacements vi of the members, where i=1,2 throughout the document. Axial strain in each component of the composite beam can be determined by using the Euler–Bernoulli beam kinematics, hence in terms of the axial displacement gradients

Finite element formulation

A displacement based finite element formulation can be developed by employing the total potential energy functional, i.e.Π=12i=12LAiEiεi2dAdz-Πextwhere the first integral is the elastic bending energies of the two components, in which Ai is the cross-sectional area of the component and L is the span of the composite member, and Πext is the work done by the external forces. From Eqs. (1), (2), the total potential energy functional in Eq. (2) can be written asΠ=12i=12LwiviEiAi00EiIiwividz

Numerical errors due to eccentricity

In order to illustrate the numerical behaviour, a 2 m span composite cantilever beam is analysed herein. As shown in Fig. 1, the top component of the composite beam has the modulus of elasticity of E1 = 200 × 103 MPa, cross-sectional area of A1 = 6 × 103 mm2 and the moment of inertia of I1 = 112.2 × 103 mm4. The bottom component has the modulus of elasticity of E2 = 26 × 103 MPa, cross-sectional area of A2 = 7.1 × 103 mm2 and the moment of inertia of I2 = 124 × 106 mm4. The kinematic constraint conditions of no vertical

Conclusion

In this study, a simple correction technique is introduced to avoid eccentricity related numerical errors in the modelling of composite beams with the conventional beam type finite elements. From the exact homogenous solution of the problem the interpolation functions that satisfy the kinematic constraints in the point-wise sense are obtained and the correct stiffness matrix is developed. The resulting stiffness matrix is contrasted with that of the conventional finite element and the missing

References (13)

There are more references available in the full text version of this article.

Cited by (0)

View full text