# On the coupling of local 3D solutions and global 2D shell theory in structural mechanics

- 206 Downloads

**Part of the following topical collections:**

## Abstract

Most of mechanical systems and complex structures exhibit plate and shell components. Therefore, 2D simulation, based on plate and shell theory, appears as an appealing choice in structural analysis as it allows reducing the computational complexity. Nevertheless, this 2D framework fails for capturing rich physics compromising the usual hypotheses considered when deriving standard plate and shell theories. To circumvent, or at least alleviate this issue, authors proposed in their former works an in-plane-out-of-plane separated representation able to capture rich 3D behaviors while keeping the computational complexity of 2D simulations. However, that procedure it was revealed to be too intrusive for being introduced into existing commercial softwares. Moreover, experience indicated that such enriched descriptions are only compulsory locally, in some regions or structure components. In the present paper we propose an enrichment procedure able to address 3D local behaviors, preserving the direct minimally-invasive coupling with existing plate and shell discretizations. The proposed strategy will be extended to inelastic behaviors and structural dynamics.

## Keywords

Plate and shells theories In-plane-out-of-plane separated representations PGD Dynamics## Introduction

Many mechanical systems and complex structures involve plate and shell parts or components whose main particularity is having a characteristic dimension (the one related to the thickness) much lower than the other ones (in-plane dimensions). In order to analyse such structures, beam, plate and shell theories have been developed in solid mechanics [1, 2] and extended later to many other physics, like flows in narrow gaps, thermal or electromagnetic problems in laminates, among many others. In these theories the introduction of appropriate kinematic and mechanic hypotheses on the evolution of the solution through the thickness of the plate (shell) allows the reduction of the general 3D mechanical problem to a 2D one involving the in-plane coordinates.

However, in many cases, when addressing complex coupled physics, inelastic behaviors or any other exhibiting localization, the validity of hypotheses able to reduce models from 3D to 2D becomes doubtful and consequently in order to ensure accurate results 3D discretizations seem compulsory. However mesh-based solutions of models defined in such degenerated domains is a challenging issue because the resulting meshes usually involve too many degrees of freedom, where the mesh size is almost determined by the domain thickness and the material and/or solution details to be represented. In order to alleviate the associated computational complexity in [3] authors proposed computing the fully 3D solution employing an in-plane-out-of-plane separated representation whose computational complexity remains the one characteristic of 2D plate or shell simulations, using the proper generalized decomposition-PGD-method [4].

However the proposed in-plane-out-of-plane separated representation appears to be too intrusive to be implemented in structural mechanics commercial software that generally propose different plate and shell finite elements, even in the case of multilayered composites plates or shells. For this reason in this work we propose a minimally-intrusive method which allows integrating fully 3D local descriptions in plate or shell models implemented in any software, without affecting its computational complexity that remains the one related to standard 2D analyses. For that purpose, two different enrichment routes will be considered, the first based on the use of that separated representation and the second on a simple condensation. The former methodology allows for very fine 3D enriched descriptions while the last is particularly well adapted to address inelastic and dynamical behaviors.

## Elastostatic problem definition

In this type of domains, plate theory is usually used in order to reduce the general 3D mechanical problem to a 2D one involving the in-plane coordinates only. Two kinds of theories exist: the thin plate theory proposed by Kirchoff [5] which establishes that the normal remains straight and orthogonal to the middle plane after deformation and the thick plate theory proposed by Reissner [6] and Mindlin [7] which assumes that normals remain straight, but not necessarily orthogonal to the middle plane after deformation.

In both theories the middle plane is taken as the reference plane \((z = 0)\) for deriving the plane kinematic equations. In this work, we consider the Reissner–Mindlin theory whose fundamental hypotheses are the following: (i) on the middle plane \((z=0)\) the in-plane displacements vanish, i.e. \(u(x,y,z=0)=v(x,y,z=0)=0\) that implies that points located in the middle-plane only moves vertically; (iii) the plate thickness remains unchanged; (iv) the plane stress assumption remains valid, i.e. \(\sigma _{zz}=0\) and (v) a straight line normal to the undeformed middle plane remains straight but not necessarily orthogonal to the middle plane after deformation.

*w*is the vertical displacement (deflection) of the points on the middle plane and the rotations \(\theta _x\) and \(\theta _y\) coincide with the angles followed by the normal vectors contained in the planes

*xz*and

*yz*respectively in their motions.

In many cases, the complexity of the solution makes impossible the introduction of pertinent hypotheses for reducing the dimensionality of the model from 3D to 2D. In that case a fully 3D descriptions seems compulsory, and the in-plane-out-of-plane separated representations become particularly suitable.

## In-plane-out-of-plane separated representation

*N*terms, each one of them consisting in the product of two unknown functions, one depending on the in-plane coordinates (

*x*,

*y*) and one on the out-of-plane coordinate

*z*, i.e.:

However the proposed in-plane-out-of-plane separated representation remains too intrusive to be implemented in structural mechanics commercial softwares that instead, propose a variety of plate and shell elements.

For this reason in the next section we propose a method which allows enriching standard plate and shell discretizations while limiting intrusivity.

## Enriched formulations

### PGD-based enriched elements

We assume the 2D mesh defined on the middle plane of a plate geometry \(\Omega \), depicted in Fig. 2. Our goal here is to address one element of the mesh, for example the one whose boundary is highlighted in red and that is noted by \(\Omega _e\), by using a fully 3D description in order to extract its homogenized \(9\times 9\) element stiffness matrix corresponding to \(\Omega _e\), and therefore fully compatible with the plate kinematics enforced on the element boundary \(\partial \Omega _e\).

Thus, the components of the localization tensor result from the elastic problem solution in \(\Omega _e\) by prescribing the canonical boundary displacements, i.e. \({\mathbf {U}}^e_1 = (1,0,0,0,0,0,0,0,0)\), \({\mathbf {U}}^e_2 = (0,1,0,0,0,0,0,0,0)\), and so on, with nine associated 3D elastic problems solved by using the PGD-based in-plane-out-of-plane separated representation.

Of course, when considering more modes in the separated representation of the 3D displacement field \({\mathbf {u}}^e\), the resulting stiffness matrix \({\mathbf {K}}^e\) differs from the one associated to standard plate theory by including eventual 3D effects ignored in standard plate kinematics. However, in that case the expression of the effective enriched stiffness matrix \({\mathbf {K}}^e\) depends on the element geometry. To avoid the necessity of calculating online stiffness matrices for each triangle, the PGD rationale allows including different parameters into the separated representation as proved in our former works. The interested reader can refer to [10, 11] for an intrusive and a non-intrusive approach, respectively. In the case here addressed, the natural choice consists of considering the triangle geometry as parameters, and then as extra-coordinates within the PGD rationale.

### Static condensation based enrichment

In this section, we consider an alternative procedure for computing the homogenized element stiffness matrix using static condensation [12].

We consider again the problem defined in the domain \(\Omega _e\) depicted in Fig. 3. The nodal degrees of freedom used for discretizing the displacement field \({\mathbf {u}}^e\), \({\mathbb {U}}^e\), can be decomposed in the ones related to internal nodes \(\mathbb U^e_i\) and the ones related to nodes located on \(\partial \Omega _e\), \({\mathbb {U}}^e_b\).

An additional advantage of this second route is the fact of deriving an expression for the effective plate nodal forces that will be advantageously considered when addressing inelastic and dynamic behaviors.

## Numerical validation

We solve the elastic problem in the plate domain depicted in Fig. 5 and compare the solutions obtained using 3D FEM, the standard plate theory and the enriched formulations considered in the previous section. The domain geometry \(\Omega = [0,H_x] \times [0,H_y] \times [0,H_z]\) is defined in Table 1 whereas Table 2 specifies the considered mesh.

Problem parameters

\(H_x\): Length in the | 250 |

\(H_y\): Length in the | 50 |

\(H_z\): Length in the | 2 |

| \(2 \cdot 10^{11}\) |

\(\nu \): Poisson coefficient | 0.25 |

Parameters values used in the simulation

\(N_x\) (number of elements in the |

\(N_y\) (number of elements in the |

\(N_z\) (number of elements in the |

The number of elements with enriched behavior (according to the procedures described in the previous section) can be arbitrarily increased, and it was noted, that as expected, by increasing it the solution approaches the reference one.

Parameters values used in the simulation for the reference solution

\(N_x\) (number of elements in the |

\(N_y\) (number of elements in the |

\(N_z\) (number of elements in the |

Thus one could conclude that the proposed technique could become advantageous when used for enriching the mechanical description inside an element or patch (as discussed later) but whose effects remain confined in the interior of that element (or patch) and when approaching to the element border the plate kinematic is accurate enough.

We consider the same problem as in the previous example but now we suppose that in the 3D element there is an hole as depicted in Fig. 7 (in the other elements we consider the enrichment presented before in absence of holes).

Here, there is clearly a double advantage in using the enrichment methodology. First, if the effect of the hole remains confined inside the element, vanishing when approaching the element boundary, the element enriched suffices and the plate meshing can be alleviated because the hole only exist for the element that will be accordingly enriched, but remains invisible at the plate level. On the other hand, its local 3D effects will be described very accurately.

Once the holed element stiffness matrix \({\mathbf {K}}^e\) is computed using the procedure presented in the previous section, the problem is solved and the generalized displacement field \({\mathbf {U}}^e\) can be extracted at the vertices of the holed triangle. Then the 3D solution inside the element can be easily computed by using (17). Figure 8 depicts different solutions. Even if \(\Omega _e\) accounts the hole presence (the PGD and static condensation based enrichment proceed on a mesh that describe the hole presence), the plate mesh becomes unaltered.

As discussed, after solving the problem at the plate level, using the enriched stiffness matrices for the different elements, 3D fields can be reconstructed. Thus, Figs. 9, 10 and 11 represent the out-of-plane stress tensor components, including the component \(\sigma _{zz}\) that is neglected when using standard plate theory. Moreover, a parabolic evolution of the components \(\sigma _{xz}\) and \(\sigma _{yz}\) along the domain thickness can be noticed in Fig. 11.

## Extension of the method to patches

In this section we extend the method previously presented to the construction of an enriched patch (super-element). In fact a single triangle element is too small for capturing rich events which can occur on a large area of the domain. Our goal is to select the patch delimited by the red line in Fig. 12, \(\Omega _p\), treat it using a 3D description by employing one of the two procedures previously presented (the PGD in-plane-out-of-plane separated representation or the static condensation procedure) and then construct the \((3 {\mathcal {N}}_b)\times (3 {\mathcal {N}}_b)\) patch stiffness matrix (being \({\mathcal {N}}_b\) the number of nodes in the border of the enriched patch). In the case here addressed numerically, the stiffness matrix size will be \(84\times 84\). In the construction procedure it is assumed that on the patch border the kinematic is the standard plate kinematic applied in the surrounding elements.

In the numerical validation we consider again the problem depicted in Fig. 5, where the geometrical and mechanical properties are defined in Table 4. On the right face (the blue area in Fig. 5) a vertical traction is applied, \(T = (0,0,1100000) \mathrm {N/m^2}\) whereas on the opposite face the displacement is prevented.

Figure 15 depicts the different solutions, where 3D FEM coarse and refined (reference) solutions are related to the meshes specified respectively in Tables 5 and 6. The solution referred as “with all patch elements” refers to the solution computed replacing all the plate elements with the 9 patches (super-elements) depicted in Fig. 16. The 3D reference solution has been computed using a mesh as fine as the one considered to describe the enriched solution at the patch level.

The solution involving the nine patches is the closest to the reference solution. However, because at the border of the patches the usual plate kinematics is enforced, a gap between the nine-patches solution and the refined 3D FEM solution (reference solution) persists. In any case, as the patches solution involved less constraints (plate kinematics) than the one related to enriched elements where the kinematic constrains are applied to all the elements boundaries, its accuracy is superior.

Since it is not pretended that plates should be discretized using enriched super-elements, our main interest is adopting accurate descriptions of rich behaviors that could be assumed confined inside a region (our patch). This is the route retained for replacing 3D FEM discretization in regions exhibiting rich behaviors by enriched super-elements keeping its 2D computational complexity.

## Extension of the method to plasticity

In this section we extend the method to problems in which plastic behavior can occur. In fact localized plasticity phenomena can occur in several situation, as in spot-welds during crash simulations.

Model parameters

\(H_x\): Length in the | 250 |

\(H_y\): Length in the | 250 |

\(H_z\): Length in the | 2 |

| \(2 \ 10^{11}\) |

\(\nu \): Poisson coefficient | 0.25 |

In order to solve the resulting nonlinear problem, various computational procedures have been proposed and extensively used, among them [15, 16, 17, 18, 19, 20, 21, 22].

*n*load step solves

Coarse 3D FEM

\(N_x\) (number of elements in the |

\(N_y\) (number of elements in the |

\(N_z\) (number of elements in the |

Refined 3D FEM considered for defining the reference solution

\(N_x\) (number of elements in the |

\(N_y\) (number of elements in the |

\(N_z\) (number of elements in the |

*j*the problem to be solved reads

## Extension to structural dynamics

The boundary \(\partial \Omega \) is decomposed according to \(\partial \Omega = \partial _D \Omega \cup \partial _N \Omega \) where displacement and tractions are prescribed.

Model parameters

\(H_x\): Length in the | 3 |

\(H_y\): Length in the | 3 |

\(H_z\): Length in the | 0.1 |

| \(2 \ 10^{11}\) |

\(\nu \): Poisson coefficient | 0.25 |

Coarse FEM mesh

\(N_x\) (number of elements in the |

\(N_y\) (number of elements in the |

\(N_z\) (number of elements in the |

Refined FEM mesh considered for computing the reference solution

\(N_x\) (number of elements in the |

\(N_y\) (number of elements in the |

\(N_z\) (number of elements in the |

An enriched stiffness can be derived at the element or patch levels, where the use of the the procedure based on the condensation allows properly addressing the forces \({\mathbf {p}}^{*}_{j+1}\), and therefor combining dynamics and plasticity.

For evaluating the performances of the proposed enrichment procedure, we consider again the problem defined in Fig. 5 and Table 7.

On the right face (the blue area in Fig. 1) a vertical traction is enforced, \({\mathbf {T}}(t) = (0,0,4.1 \cdot 10^6 \sin (2 \pi \omega t)) \mathrm {N/m^2}\), whereas on the opposite face displacement is prevented. Without loss of generality homogeneous initial conditions \({\mathbf {u}}({\mathbf {x}},t=0) = 0\) and \({\mathbf {v}}({\mathbf {x}},t=0) = 0\) are assumed. Table 8 reports the considered coarse mesh whereas Table 9 defined the refined one from which the reference solution is calculated.

Figures 18 and 19 depict respectively the solutions computed using the different techniques at four different times and the vertical displacement time evolution at the right border. Figures 20 and 21 present similar results but in a case where plasticity takes place. Again the best solutions are the ones related to a nine-patches discretization for the reasons widely exposed before.

## Conclusions

This paper proposed two different procedures to capture 3D behaviors by enriching elements (or super-elements) connected to plate discretization. This goal is performed by integrating, in a non-intrusive manner, 3D elements (or patches) in plate or shell based commercial codes. In order to compute this enriched 2D element, two different techniques were presented: one based on the PGD in-plane-out-of-plane separated representation and the other on the static condensation. The method was firstly developed in linear elastic settings and then successfully extended to structures exhibiting inelastic behaviors or dynamics.

## Notes

### Author's contributions

All the authors participated in the definition of techniques and algorithms. G. Quaranta, developed moreover the simulation software, assisted by M. Ziane. All authors read and approved the final manuscript.

### Acknowledgements

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 675919.

### Competing interests

The authors declare that they have no competing interests.

## References

- 1.Oñate E. Structural analysis with the finite element method. Linear statics. Volume 2: Beams, plates and shells. Berlin: Springer; 2010.zbMATHGoogle Scholar
- 2.Ahmad S, IB M, ZO C. Analysis of thick and thin shell structures by curved finite elements. Int J Numer Methods Eng. 1970;2(3):419–51.CrossRefGoogle Scholar
- 3.Bognet B, Bordeu F, Chinesta F, Leygue A, Poitou A. Advanced simulation of models defined in plate geometries: 3D solutions with 2D computational complexity. Comput Methods Appl Mech Eng. 2012;201–204(Supplement C):1–12.CrossRefGoogle Scholar
- 4.Chinesta F, Leygue A, Bordeu F, Aguado JV, Cueto E, Gonzalez D, Alfaro I, Ammar A, Huerta A. PGD-based computational vademecum for efficient design, optimization and control. Arch Comput Methods Eng. 2013;20:31–59.MathSciNetCrossRefGoogle Scholar
- 5.Kirchhoff G. Uber das Gleichqewicht und die Bewegung einer elastichen Scheibe. J Reine und Angewandte Mathematik. 1850;40:51–88.CrossRefGoogle Scholar
- 6.Reissner E. The effect of transverse shear deformation on the bending of elastic plates. J Appl Mech. 1945;12:69–76.MathSciNetzbMATHGoogle Scholar
- 7.Mindlin RD. Influence of rotatory inertia and shear in flexural motions of isotropic elastic plates. J Appl Mech. 1951;18(1):31–8.zbMATHGoogle Scholar
- 8.Bognet B, Leygue A, Chinesta F. Separated representations of 3D elastic solutions in shell geometries. Adv Model Simul Eng Sci. 2014;1:4.CrossRefGoogle Scholar
- 9.Quaranta G, Bognet B, Ibañez R, Tramecon A, Haug E, Chinesta F. A new hybrid explicit/implicit in-plane-out-of-plane separated representation for the solution of dynamic problems defined in plate-like domains. Comput Struct. 2018;210:135–44.CrossRefGoogle Scholar
- 10.Chinesta F, Keunings R, Leygue A. The proper generalized decomposition for advanced numerical simulations. A primer. Berlin: Springer; 2014.CrossRefGoogle Scholar
- 11.Borzacchiello D, Aguado JV, Chinesta F. Arch Computat Methods Eng (2017). https://doi.org/10.1007/s11831-017-9241-4
- 12.Wilson LE. The static condensation algorithm. Int J Numer Methods Eng. 1974;8(1):198–203.CrossRefGoogle Scholar
- 13.Owen DRJ, Hinton E. Finite elements in plasticity: theory and practice. Swansea: Pineridge Press; 1980.zbMATHGoogle Scholar
- 14.Rv Mises. Mechanik der festen Krper im plastisch- deformablen Zustand. Nachrichten von der Gesellschaft der Wissenschaften zu Gottingen, Mathematisch-Physikalische Klasse. 1913;1913:582–92.zbMATHGoogle Scholar
- 15.C Zienkiewicz O, Valliappan S, King I. Elasto-plastic solutions of engineering problems ‘initial stress’, finite element approach. J Numer Methods Eng. 1969;1:75–100.CrossRefGoogle Scholar
- 16.Gallagher RH. Stress analysis of heated complex shapes. ARS J. 1962;32(5):700–7.CrossRefGoogle Scholar
- 17.Argyris JH. Elasto-plastic matrix displacement analysis of three-dimensional continua. J R Aeronaut Soc. 1965;69(657):633–6.CrossRefGoogle Scholar
- 18.Swedlow JL, Williams ML, Yang WH. Elasto-plastic stresses and strains in cracked plates. California Institute of Technology: Firestone Flight Sciences Laboratory; 1965.Google Scholar
- 19.Pope GG. A discrete element method for the analysis of plane elasto-plastic stress problems. Aeronaut Q. 1966;17(1):83–104.MathSciNetCrossRefGoogle Scholar
- 20.Reyes SF, Deere DU. Elastic-plastic analysis of underground openings by the finite element method. Lisbon: Int Soc Rock Mech Rock Eng; 1966.Google Scholar
- 21.Marcal PV, King IP. Elastic-plastic analysis of two-dimensional stress systems by the finite element method. Int J Mech Sci. 1967;9(3):143–55.CrossRefGoogle Scholar
- 22.Popov EP, Khojasteh-Bakht M, Yaghmai S. Bending of circular plates of hardening material. Int J Solids Struct. 1967;3(6):975–88.CrossRefGoogle Scholar
- 23.Newmark NM. A method of computation for structural dynamics. J Eng Mech Div. 1959;85(3):67–94.Google Scholar

## Copyright information

**Open Access**This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.