- 2.

1 -

COMPUTER MODELLING PRINCIPLES AND FEM

MODELLING PRINCIPLES
Why need for modelling?

 Analysis is an essential part of the structural design process. The aim of

structural analysis is to determine the behavioural responses of a
structure with which to justify the adequacy of the structure in terms of
structural design criteria such as strength, stability and stiffness.

STRUCTURE

EXCITATION RESPONSES
Loads Displacements
Vibrations Accelerations
Thermal changes Stresses and strains
Settlements Internal forces and moments

ANALYSIS To determine the Responses

of the structure to excitations

To ensure that the structure can

DESIGN sustain the excitation within an
acceptable level of response

 Structural engineering
is the art of using materials
that have properties which can only be estimated,
to build real structures
that can only be approximately analysed
to withstand forces
that are not accurately known
so that our responsibility with respect to public safety is satisfied.
 - 2.2 -

 Before the construction of the structure we cannot analyse ‘the structure’.

During the design stage of a structure, we can only analyse a
‘mathematical model’ of the structure whose properties and behaviour
we aim to select so that it adequately represents the real structure.
Therefore, we need tools to model the structure and to analyse the
model.

 The approximations and assumptions which must be made to create this

mathematical model mean that there can be no such thing as an exact
model. It is commonly believed that present day computer aided
analysis is necessarily more ‘accurate’ than earlier analytical hand
methods. However, while results of computer analysis may be ‘precise’,
unless the analytical model is properly established, otherwise the results
may be misleading and could be dangerously wrong.

Basic Principles in Modelling

Before embarking on an analysis it is necessary to consider what

analysis is needed:

- What needs to be demonstrated, what will dictate the design?

Before any analysis is done, the engineer must be able to visualise

how he or she expects the structure to work, and must have
preliminary ideas of the magnitude and form of answers anticipated.

- What is the simplest model which can demonstrate this?

Never use a model that is more complex than necessary. The

simplest model will be the easiest to understand, the quickest to set
up and use and so probably the most useful. Different models may be
used for different needs.

- What effects will be ignored or misrepresented by this model? Are

they significant?

e.g. the analysis will normally assume linear elastic materials and
small deflections.

- What results will be produced? Can they be used to demonstrate the

desired effect?
 - 2.3 -

Never accept the results of a computer analysis without passing

through rigorous scrutiny, especially if you cannot provide an
engineering explanation for the behaviour predicted by the computer.

- What will happen to the reactions from this model?

e.g. can the foundations resist the forces generated?

FUNDAMENTALS OF FINITE ELEMENT METHOD

What is Finite Element Method (FEM)?

 It is a numerical analysis technique for obtaining approximate solutions

to a wide variety of engineering problems. In a finite element analysis
procedure, a complex continuum body is discretized into many smaller
parts of simple shape, called finite elements. The material properties
and the governing relationships are considered over these elements
and expressed in terms of unknown values at element corners, called
nodes. An assembly process, duly considering the loading and
constraints, results in a set of linear algebraic equations. Solution of
these equations gives us approximate behavior of the original
continuum, under the prescribed loading.

 
 

OR

SHEAR WALL CONTINUUM FINITE ELEMENT MODEL BEAM COLUMN MODEL

 - 2.4 -
 The purpose of modelling is to try to represent the behaviour of a real
structure. There are three basic steps in the modern approach to creating
an analytical finite element model of the structure

1. Assumptions are made about the behaviour of a small piece of the

material – the so-called material model or constitutive relationship of
stress and strain of the material.

2. An integration process is performed on the infinitesimal pieces of

material to define the behaviour of an element which represents a
finite part of the structure – hence the term ‘finite element’.

3. The final step is an assembly of finite elements to form the finite

element model which represents the behaviour of the complete
structure. At this stage the boundary restraints are imposed and
loading conditions defined.

 The fundamental equations of structural mechanics can be placed in

three categories.

- The stress-strain relationship contains the material property

information which must be evaluated by laboratory or field
experiments

- The total structure, each element, and each infinitesimal particle with
each element must be in force equilibrium in their deformed position.

- Displacement compatibility conditions must be satisfied.

Minimum energy principles or methods of virtual-work are often used to

derive the general equations of elasticity for linear structural analysis.
 - 2.5 -

Continuum
Body Stress-Strain Equation
Structural Model
Equilibrium Equation
 K U   F 
Compatibility Equation
“Algebraic Equations”

“Differential (Principle of Virtual Work)

Equations”

 Advantages:

The main advantage of FEM, compared to conventional analytical

methods, is its versatility. The method can be applied to a variety of
physical problems. The structure analysed can be of arbitrary shape
with arbitrary applied loading and restraints. The structure mesh can
mix elements of different types, shapes and physical properties.

 Disadvantages:

A powerful numerical analysis tool requires the use of a computer and

a higher level of competency of the engineer. Experience and good
engineering judgement are required by a user in order to define a
representative model.

Material Properties

 The initial step in creating a finite element model is to define the

behaviour of the material or the constitutive relationship which relates the
stress and strain of the material.

 Linear elasticity is the most common material model in structural

engineering. Assumptions adopted in linear elasticity are:
 - 2.6 -
1. As stress increases the resulting strain increases in linear proportion.

2. As stress decreases the resulting strain decreases in the same linear

proportion.

3. Strain induced at right angles to an applied strain is linearly proportion

to the applied strain (Poisson’s ratio effect).

4. When the stress is removed, the strain is recovered to its initial value.

 The linear stress-strain relationships contain material property constants.

The mechanical material properties for most common material are
defined in terms of three numbers – modulus of elasticity E, Poisson’s
ratio  and coefficient of thermal expansion . In addition, the unit weight
 or the unit of mass  are also considered to be fundamental properties.
 Prior to the development of FEM, most analytical solutions in solid
mechanics were restricted to materials which are isotropic (equal
properties in all directions) and homogenous (same properties at all
points in the solid). With the introduction of FEM, we can consider a more
general definition of anisotropic materials having different properties in
different directions.

Sign Convention for

an Infinitesimal Element

 The most general form of the three dimensional stress-strain

relationship, for linear elastic structural materials subjected to both
stresses and temperature change, can be written in the following matrix
form
 - 2.7 -

E5 E6

E5 E6

E5 E6

E5 E6

E6

E5

In symbolic matrix form

d  Cf  T

where f

C = the compliance matrix in which each column of the matrix

represents the strains due to the application of a unit
stress
 = the strain matrix due to a unit change in temperature

Note that from equilibrium  12   21 ,  23   32 and  13   31 .

 Orthotropic Materials

An orthotropic material is a type of anisotropic material in which shear

stresses, acting in the three reference planes, cause no normal strains.
The constitutive relationship for orthotropic material can be written as
 - 2.8 -

For orthotropic material the C matrix has nine independent material

constants and there are three independent coefficients of thermal
expansion. Material such as wood, fiber composite laminate exhibits
orthotropic behaviour.

 Isotropic Materials

An isotropic material has equal properties in all directions and is the most
commonly used approximation to predict the behavior of linear elastic
materials. The constitutive relationship for isotropic material can be
written as
 - 2.9 -

It appears that the compliance matrix has three independent material

constants. It can be shown that the G=E/2(1+). Therefore, for isotropic
materials only Young’s modulus E, Poisson’s ratio  and the coefficient
of thermal expansion  need be defined.

 Plane Strain Isotropic Materials

If 3, 13 and 23 are zero the infinitesimal element is in a state of plain
strain. If the material is isotropic, then 13 and 23 are also zero and the
compliance matrix is reduced to a 3 x 3 array. For plane strain and
isotropic materials the constitutive relationship is

where

The cross sections of dams and tunnels with a near infinite dimension
along the longitudinal 3-axis can be considered in a state of plane strain
for loading in the 1-2 plane. For the case of plane strain the strain in the
longitudinal 3-direction is zero. However, the normal stress in the 3-
direction can be derived as

 Plane Stress Isotropic Materials

If 3, 13 and 23 are zero the infinitesimal element is in a state of plane
stress. For this case the compliance matrix is reduced to a 3 x 3 array.
For plane stress and isotropic materials the constitutive relationship is
 - 2.10 -

E
where E 
(1  v 2 )

The membrane behavior of thin plates and shear wall structures can be
considered in a state of plane stress for loading in the 1-2 plane.

 Axisymmetric Materials

A large number of common types of structures are made of circular cross

sections. For examples, pipes, pressure vessels, silos, fluid storage
tanks and rockets, fall into the category of axisymmetric structures. Many
axisymmetric structures have anisotropic materials. For the case of
axisymmetric solids subjected to non-axisymmetic loads, the compliance
matrix can be written in terms of the r, z and  cylindrical reference
system and the constitutive relationship is given as

6
 - 2.11 -
 Use of material properties within computer program

The stress-strain equations are the fundamental constitutive laws for

linear materials. Most of the modern computer programs for finite
element analysis require that these equations be expressed in terms of
forces and displacements, i.e.

f  Kd  fo

in which K = C-1 is the stiffness matrix and fo initial zero-strain thermal

stress. Note that the terms K, C-1 are not a function of the load and are
only the material properties of the elements. The numeric inversion of
these matrices can be easily performed within the computer program.

 Properties of Standard Materials

Basic properties for commonly used structural materials taken from

relevant design standards are given in the table below.

Material Young’s Poisson’s Coefficient Density

modulus Ratio of Thermal (kg/m3)
(kN/mm2) Expansion
(strain/C)
Steel 205 0.3 12e-6 7850
Aluminium 70 0.3 23e-6 2710
Timber 4-20 // grain N.A. N.A. 290-1080
0.2-1.3 X grain
Concrete 23.7 (C30) 0.2 10e-6 2400
27.7 (C45)
31.1 (C60)
35.1 (C80)

For reinforced concrete tall buildings, which are made up of two very
different materials, the gross concrete properties are normally used
during the preliminary design process. If the concrete materials are
subjected to loading without causing net tension, then an equivalent
Young’s modulus taken into the consideration of the stiffening
contribution of the steel reinforcement in concrete, can be given as
follows.
 - 2.12 -
Ac A
Eeq  Ec  s Es
A A

where Eeq = equivalent modulus of elasticity

Es = modulus of elasticity of steel
Ec = modulus of elasticity of concrete
Ac = net concrete area; As = steel reinforcement area
A = As + Ac

For example, consider grade 60 concrete with 2% steel reinforcement,

the equivalent short-term modulus of the reinforced concrete is

Eeq  200  0.02  31.1 0.98  34.5 kN / m 2

which is approximately 11% higher than that of the gross concrete.

Element Behavior

 The element addressed here are finite parts of the structure whose
properties are described in relation to forces and displacements at nodes
on the element. ‘Forces’ and ‘Displacements’ are generic terms. While
‘forces’ denote the internal forces and moments at a node,
‘displacements’ represent both translational and rotational movements at
the node. A ‘degree of freedom’ denotes a displacement freedom
associated with a nodal point.

 In FEM, there are two classes of elements:

Frame elements – which do not require mesh refinement. Only prismatic

bar and beam elements are discussed since these elements are the most
widely used in structural modelling.

Finite elements – which do need mesh refinement to improve the

accuracy of the solution. As the mesh is refined, the result should
‘converge’ to the ‘exact’ value.

There are now a large number of elements in use. Only the basic
assumptions for the most commonly used elements are discussed.
 - 2.13 -
 To establish an element stiffness matrix, four basic assumptions are
needed:

1. The geometry needs to be defined, e.g. the element can be defined

as a straight line, a flat straight-sided quadrilateral, a tetrahedron with
parabolic shaped sides, etc.

2. The degrees of freedom are chosen. This choice depends on the

accuracy required from the element and other factors.

3. The constitutive relationship is selected. For example, isotropic or

anisotropic material.

4. Assumptions are made regarding the state of the displacements or

shape functions where the independent variables are the nodal
displacements.

 When the element chosen involves the use of displacement functions,

the displacements within the element may or may not be compatible.
When there is boundary compatibility, then the element is said to be
‘conforming’ or ‘compatible’. Where boundary compatibility is not
satisfied the element is ‘non-conforming’.

 A mesh of conforming elements in general gives stiffer predictions than

would result from an exact solution. An ‘influence coefficient’ (i.e. the
calculated deflection vs the exact solution) will thus give a lower bound
value.

 A lack of full boundary compatibility causes a relaxation in the stiffness

of the element model and therefore may result in an upper bound to an
influence coefficient. The solution in general is more flexible than that
resulting from an exact solution.
 - 2.14 -
COMMON ELEMENTS

 JOINT ELEMENTS

A joint element is normally comprised of a set springs which connects

two nodes or two freedoms. The springs are characterized by
relationships of the form:

p  k s

where p is a direct force or moment, ks is a spring stiffness,  is the

movement in the direction of p.

In-plane Joint Element

The two nodes can normally be in the same positions or at two separate
locations. The figure in the above shows an in-plane joint element. For a
3D joint element, up to three translational springs and three rotational
springs can be included with various stiffness properties.

Various Joint Stiffness

 - 2.15 -
Three common types of joint stiffness are shown in the figure. (a) shows
a linear elastic spring stiffness; (b) shows elasto-plastic behaviour for a
rotational spring which is normally used to model a plastic hinge in a
frame; (c) shows a contact spring relationship which has a low (or often
zero) stiffness in tension direction and a gap.

 BAR ELEMENTS

The assumptions adopted for a prismatic bar elements are:

1. Geometry – the bar is straight and lies on the local X-axis. It has
uniform cross-section throughout its length.

2. Degrees of freedom – there is an axial translational freedom at each

end of the bar.

3. Constitutive relationship written in the form of force and displacement

as

N  EA du / dx

where N is the axial force in the bar, EA = axial stiffness, u is the

displacement in the X-direction.

4. Internal displacement function – or shape function

u  1   2 x , where 1 ,  2 are coefficient terms.

Differentiating the displacement function gives du / dx   2 , i.e. constant

strain along the length of the bar element. Since the axial force N is
constant along the length of the element, the constitutive relationship
also infers constant strain and therefore the assumed displacement
function gives a correction solution for the constitutive equation, requiring
no mesh refinement for accuracy.
 - 2.16 -
The uniform stress assumption is good for bars in tension where their
ends are pinned. However, the assumption is only acceptable for pinned
ended ‘short’ columns with negligible eccentricity of the applied load.
With ‘long’ column where the effects of eccentricity and the second order
P-delta effect are significant, the first order linear elastic assumption for
bar elements may be inadequate.

 BEAM ELEMENTS

The term ‘beam’ tends to denote bending but in the context of finite
element analysis it normally refers to a general line element which has
bending, shear, axial and torsional deformation.

Beam element under pure bending

The basic assumptions for a beam element under pure bending are:

1. Geometry – the planar line element has uniform cross-section

throughout its length. It lies on the local X-axis of the X-Y plane.

2. Degrees of freedom – there is a translational and a rotational freedom

at each end of the bar.

3. Constitutive relationship for beam bending in the form of moment and

curvature as

M  EI d 2v / dx

where M is the bending moment, EI = flexural stiffness, v is the

displacement in the Y-direction.

4. Internal displacement function – or shape function

v  1   2 x   3 x 2   4 x3 , where 1 , 2 , 3 , 4 are coefficient terms.

 - 2.17 -

Differentiating the displacement twice gives d v / dx  2 3  6 4 x

2 2

which is linear. With only end moments applied, the bending moment
profile in the element linear and therefore the curvature derived from
the displacement function is linear. As a result, the assumption
displacement function gives a correct solution for the governing
constitutive equation, requiring for no mesh refinement for accuracy.

Using the conditions of equilibrium and compatibility, the element force-

displacement relationship of the planar flexural beam can be written as

Beam element under bending and shear

As compared to the nonlinear deformed shape of a beam under bending,

the deformed shape (i.e. dv / dx ) of an infinitesimal element under a
shearing action is linear.

Bending Mode Shear Mode

 - 2.18 -
The constitutive relationship for shear deformation under the shear
action is given as

dv
S  AG
dx
dv
where S = shear force, A = shear area and = the rotation due to
dx
shear.

Shear deformation causes the end rotations to increase. Consider the

beam with a moment m A1 applied at end 1, the shear is constant over
m A1
the length with a value of .
L

Note that the end rotation due to shear can be given as

 s1  mA1 /( L AG ) . Using the principle of virtual work, based on the
conditions of equilibrium and compatibility, the element moment-rotation
relationship can be modified by the shear deformation as

Notice that the total end rotation at the end 1,  A1 , is now an average
rotation at the end of the member, including the rotation due to bending
and the rotation caused by the shear force.
 - 2.19 -

Beam element under torsion

As beam element is under torsional loading, the element will twist and
exhibit both St. Venant and warping torsional behavior.

While St. Venant torsion can give good results for closed sections, the
warping torsion is more applicable to open sections made of thin plates.
The constitutive relationship for St. Venant torsion is

dT / dx  GJ d 2 / dx 2

where the dT / dx = torque per unit length, GJ =torsional stiffness of the

cross section,  = angle of twist of the cross section.

The constitutive relationship for warping torsion is

dT / dx  EI w d 4 / dx 4

where EI w = warping stiffness of the cross section, I w = the second

moment of the sectorial area of the cross section.

Warping torsion is not well understood by most structural engineers

mainly because structural members are often not required to resist
torsional loading. Indeed most FEM software in structural engineering
has neglected the effect of warping torsion and only St. Venant torsion is
considered.
 - 2.20 -
Typical Beam Elements in Frame Buildings

Plane frame elements – By combining a bar element and a bending

planar element, one gets a plane frame beam element. This represents
the most commonly used elements in building frames. For linear
elements the P-delta effect of the combined axial force and the bending
deformation is not considered.

Grillage frame elements – By combining a torsion element and a bending

element one obtains a grillage element.

Space frame elements – This represents the most general types of beam
elements which include a bar element, bending elements for bending
about both the X- and Y-axes and a torsion element. Each space frame
element has six degrees of freedom (d.o.f.) per node. If the warping
torsion terms are also included there are seven d.o.f. per node.
 - 2.21 -

 MEMBRANE ELEMENTS

Membrane elements are 2D planar elements taken into the consideration

of only in-plane stress and strain. Their accuracy depends on how such
a structure is meshed. In general, the finer the mesh, the more accurate
the analysis results. The thickness of a membrane element is usually
constant. The basic assumptions for some typical membrane elements
are presented in the following to help develop an insight into their
behavior.

Triangular Constant Strain Element

The four basic assumptions for this type of membrane elements are:

1. Geometry – as its name implies, it is a flat straight-sided triangle.

2. Degrees of freedom – there are two translational freedoms at each

node giving a total of six freedoms per element.

3. Constitutive relationships – It follows basically the plane stress or

plane strain constitutive relationship.

4. Displacement functions are

u  1  2 x   3 y
v   4  5 x   6 y
where u and v are the displacements in the X- and Y-directions
respectively and i are constant coefficients.
 - 2.22 -
Note that the direct strains are
u v
  2,  6
x y
and the shear strain is
u v
  3  5 .
y x

Therefore, all strains (and hence stress) components over the area of an
element are constant. In this element, displacements are linear and
strains are constant. This then is the main assumption made as to the
behavior of the element.

This triangular constant strain element is one of the simplest possible,

but less accurate 2D membrane element. If a beam in bending is
modelled using triangular constant strain elements the resulting in-plane
stress distribution at a section will be shown in a step manner. If
sufficiently large number of elements is used, the stepped distribution will
approach to a correct linear distribution.

Four-Noded Linear Strain Quadrilateral Element

 - 2.23 -

As an attempt to improve the accuracy of a membrane element, a four-

noded linear strain quadrilateral element was developed.

The four basic assumptions for this type of membrane elements are:

1. Geometry – quadrilateral shape with four straight sides.

2. Degrees of freedom – two translational freedoms at each corner,

giving a total of eight freedoms per element.

3. Constitutive relationships – same as the plane stress or plane strain

constitutive relationship.

4. The displacement functions are:

u  1  2 x   3 y   4 xy
v   5  6 x   7 y  8 xy
Unlike the triangular constant strain element, this element gives a linear
distribution of both direct strain and shear strain as follows.

u v u v
  2   4 y,   7   8 x,   3   4 x   6  8 y
x y y x

Quadrilateral Element with mid-side nodes

 

 
 - 2.24 -
As a further attempt to improve the accuracy of a quadrilateral
membrane element, a eight-noded quadrilateral element was developed.
This element is capable of producing a parabolic deformed shape of
each side of the element.

The four basic assumptions for this type of membrane elements are:

1. Geometry – the element has four sides, each side being capable of
having a parabolic deformed shape.

2. Degrees of freedom – two translational freedoms at each node, giving

a total of sixteen freedoms per element.

3. Constitutive relationships – same as the plane stress or plane strain

constitutive relationship.

4. The displacement functions used are:

u  1  2 x   3 y   4 xy   5 x 2 y   6 xy 2   7 x3  8 x 4
v   9 10 x  11 y  12 xy  13 x 2 y  14 xy 2  15 y 3  16 y 4

This eight-noded element is considered to be more accurate and is

available in most of today’s FEM software packages. However, they are
more cumbersome to use and require for higher computational efforts.

 PLATE BENDING ELEMENTS

Plate bending elements are used to model thin 2D planar elements

subjected to transverse loadings.

A
B
A
B
 - 2.25 -

The simplest kind of plate bending element is the rectangular thin plate
bending element or the so-called ACM rectangular element, which was
first developed independently by Adini and Clough (1961) and Melosh
(1963), hence the ACM name.

The four basic assumptions for this type of plate bending elements are:

1. Geometry – the element is a rectangular flat element.

2. Degrees of freedom – at each corner three degrees of freedom are

chosen corresponding to translation at the transverse direction (z-
axis) of the plane of the plate and rotations about the x- and the y-
axes of the plate.

3. Constitutive relationship – the thin plate bending constitutive

relationship can be given by the following equation.

  2 
    x 2 
 Mx  1  0  2 
  Eh 3
    
 y
M  2 
 1 0    y 2 
M  12(1   )
 xy 
 1    2 
 0 0    
 2  2 
 x y 

The constitutive relationship is based on the following assumptions

- Normals to the plane of the thin plate remain straight after

deformation.
- The normal stress in the thickness direction is assumed to be zero
(plane stress).
- Shear deformation in the thickness direction of the plate is neglected
(Kirchhoff thin plate theory). This is valid for thin plates normally with
span-to-depth ratios greater than 10:1.
- The deflections are small in comparison with the plate dimensions.
- There is no strain in the centroidal plane of the plate subject to pure
bending.
 - 2.26 -

Since the thin plate is under bending, the normal to the plane is
assumed to be straight but to rotate through an w / x so that

u w v w
  and  
z x z y

Therefore the normal strains can be rewritten as

u 2w v 2w
x    z 2 and  y   z 2
x x y y

The bending moment per unit width on the x-plane is

M x  hh/ 2/2  x z dz

For plane stress thin plate with isometric material,

E
x  ( x  v y )
1  v2
Therefore,
E  2w 2w  Eh3 2w 2w
Mx  h/2
 h / 2  z  vz 2  z dz  ( v 2 )
1  v 2  x 2 y  12(1  v 2 ) x 2 y
Similarly,
E  2w 2w  Eh3 2w 2w
My  h/2
 h / 2  vz 2  z 2  z dz  (v 2  2 )
1  v2  x y  12(1  v 2 ) x y
 - 2.27 -

Note that the thin bending plate may experience twisting under lateral
loading. The shear strain of a differential element of the plate is

u v 2w
 xy    2 z
y x xy

Therefore the twisting moment per unit width on the x-plane is

2w
M xy  
 hh/ 2/ 2 xy z dz  
 hh/ 2/ 2 G xy z dz  hh/ 2/ 2 G (2 z ) z dz
xy

Substituting G  E / 2(1  v) and performing integration gives

h3  2 w E h3  2 w 2w
M xy  2G ( ) 2  D(1  v)
12 xy 2(1  v) 12 xy xy

As a result, the constitutive relationship of the isotropic thin plate

bending element is

  2 
 2 
x
 Mx  1  0  2 
       Eh3
 M y   D  1 0  2  where D 
 y
M    12(1  v 2 )
 xy   0 0 1   
  2 
 
 x y 
 - 2.28 -

4. Displacement function used for the ACM rectangular plate bending

element is

w  1   2 x   3 y   4 xy   5 x 2   6 y 2   7 x 2 y   8 xy 2 +
 9 x3  10 y 3  11 x3 y  12 xy 3

which is the so-called Lagrange interpolation function in two

dimensions. In terms of boundary compatibility, this displacement
function is conforming in the displacement in the thickness direction.
However, the element involves a discontinuity of cross slope at inter-
element boundaries and therefore is non-conforming in terms of
rotations between elements.

To examine the inter-element compatibility, consider a x=constant

boundary. The displacement w is reduced to the following form

w  c1  c2 y  c3 y 2  c4 y 3

in which the four constant coefficients ci can be uniquely defined by

the four nodal displacements (i.e. 2 translations and 2 rotations) at the
two ends of the x=constant boundary.

Now let’s define the rotation by differentiating the transverse

displacement at any x=constant boundary as

w
y   c2  2c3 y  3c4 y 2
y

It is obvious that there are only two nodal rotations at the ends of
boundary to define the  y function where three are required. The
displacement function assumed is non-conforming in terms of inter-
element rotations.
 - 2.29 -

 OTHER PLATE BENDING ELEMENTS

In the past 30-40 years, several hundred plate-bending finite elements

have been proposed. Much effort has been made, but there is not yet a
clear guide to the best element for use. The accuracy of these finite
element types depends on how the structure is to be meshed. In general,
plate bending elements using quadrilateral shapes is more preferable
than triangular plate bending elements.

For thicker plates (normally with span-to-depth less than 10), the shear
deformation in the thickness direction is no longer negligible. It is then
advised to use the Reissner-Mindlin constitutive relationship (Mindlin
1951) to consider plate bending with the consideration of shear
deformation.

Reissner E. (1945) “The effect of transverse shear deformation on the bending of elastic plates”, J. of Applied
Mechanics, ASME, 12, 69-76.

Mindlin, R.D. (1951) Influence of rotary inertia and shear on flexural motions of isotropic elastic plates, J. of
Applied Mechanics, ASME, 18, 31-38.

 SHELL ELEMENTS

By combining a membrane and a plate bending element one obtains a

flat ‘shell’ element.
 - 2.30 -

Shell element is a useful form of element for slab analysis where

membrane effects are not negligible. It is also a common type of element
for concrete shear walls of arbitrary geometry. Since this shell element is
developed based on the assumption of a flat plate of uniform thickness,
it is not suited for curved shell structures and cannot be fitted to doubly
curved shells (with double curvature).

Each node of a conventional shell element has 5 degrees of freedom (3

translations and two out-of-plane rotations). Therefore, a four-node
quadrilateral shell element has 20 degrees of freedom. In today’s tall
concrete building structures, the use of coupled shear wall systems
requires the finite element model to make use of a combination of shell
elements with coupling beam elements. Since the conventional shell
element does not consist of the in-plane rotation, it cannot be used to
effectively model coupled shear wall structures.

Drilling
D.O.F

QUADRILATERAL SHELL ELEMENT

WITH DRILLING D.O.F.

Most modern shell elements have added on the in-plane rotation d.o.f.
or the so-called drilling d.o.f., resulting in 6 d.o.f. (3 translation and 3
rotations) per node. Such shell elements provide rotational continuity
between shear wall shell element with flexural lintel beams.

As in the case of plate bending, the shell element has the option to
include transverse shearing deformation. Shell elements defined as
Thick Plate include transverse shear deformation using the
Mindlin/Reissner theory and only affect the plate bending behavior, not
the membrane behavior of the element. The thin plate elements use the
Kirchhoff theory.
 - 2.31 -

 SOLID ELEMENTS

Solid elements represent the most comprehensive types of elements for

solid structures. The simplest types of solid elements are the four-noded
tetrahedron elements and eight-noded brick elements with each node
having three translational degrees of freedom (i.e. totally 12 d.o.f. for a
tetrahedron and 24 d.o.f. for a brick).

Higher order solid elements with larger number of d.o.f. are available; but
they require a significant increase in computational efforts and the results
are more difficult to interpret. Solid elements are used for solid structures
such as dams and soil continuum. However, they are seldom used in tall
building structures and thus are only briefly described herein.

SYSTEM BEHAVIOUR

This section is concerned with the techniques for setting up models of

structural systems using conventional FEM methods. The most popular
computer software packages for analysis of buildings in Hong Kong are the
ETABS and SAP2000 software. Both of these software packages are
available in our computer lab. While ETABS is a special-purpose analysis
software for buildings structures, SAP2000 software package is a general-
purpose finite element software for buildings as well as any other kinds of
structural systems. Some general considerations regarding modelling of
structural building systems are to be considered in the following.
 - 2.32 -

 FRAMES

A frame is basically an assemblage of line beam elements, in which these

members are capable of resisting axial, shear, bending and torsional
twisting actions. Often times these members are interconnected and are
assumed to have zero cross section dimension, but with sectional
properties (i.e. axial area, shear areas, bending and torsional stiffness).

Finite size of members:

Because real members are not of zero thickness, they never meet at a
point.

Center-to-Center Model
Steel Beam on Columns

Conventional frame models assuming center-to-center lengths is likely to

underestimate the flexural stiffness of the beam. For the case of the frame
under lateral loading, the beam will be bent in a double curvature manner
such that the relationship between end moment and rotation for the beam
is

6 EI
M 
L(1   )3

where  = C/L = column width / center-to-center length L.

 - 2.33 -

When considering a typical face-to-face length of a beam with a rigid

offset equal to 10% of the center-to-center length, the stiffness of the
beam can be significantly increased by slightly over 37%. Beam
elements framed to finite columns can be modelled using a single
element with a transformed stiffness, or element with rigid ends.

Beam Modeled with Rigid Offsets

Connection flexibility:

The connection at this idealised point in the model can be one of two
types:

- Rigid joint, in which there is no relative rotation between elements

meeting at a point.

- Pinned joint, in which no moment is transferred between elements

meeting at a joint.
 - 2.34 -

True pins are rare, unless special connection detail is provided. Truly rigid
joints never occur, although many joints are stiff enough for such an
assumption to be made. A partially fixed or semi-rigid joint, where a
rotational spring is used to transfer moment across a joint while allowing
some rotations to be included in a model. However, the spring stiffness
of a semi-rigid joint is generally not readily available.

Joints in steel structures even when apparently stiff are usually modelled
as pinned which simplifies joint design and ensures that members
designed to resist sway carry all horizontal loads. Where a nominally
pinned joint has to be able to rotate, the capacity of the actual connection
to accommodate those rotations must be checked.

For normal purposes full connection fixity is acceptable for:

- welded joints or bolted joints with web and flange cleats

- continuous concrete frame connection.

Shear Deformation:

If shear deformation of the members of a rigid frame is neglected (usually

by not providing the shear area of a frame member), the model will tend
to deform less than the actual frame. This is probably acceptable when
the span-to-depth ratio of the members is relatively large with a value of
10 or more. Deep beams with a span-to-depth ratio of 2.5 or less are
unlikely to behave as flexural beams. They should be designed as deep
beams and can be better modelled by a strut and tie model.
 - 2.35 -

In practical tall building frameworks, shear deformations tend to increase

with the number of bays or as the bay size reduces. For tall tubular steel
structures, shear deformation can account for as much as flexural
deformation of the total deformation.

In general even if shear deformation affects the overall displacements of

a frame significantly, it tends not to affect the member forces to the same
degree. With today’s modern FEM, it is advisable that the shear
deformation of members be always included, particularly for the case of
structural steel frameworks.

Members with Non-uniform Sections:

Haunched beams and perforated beams are often found in building

frames.

Haunched beams can be normally treated by choosing a mean stiffness

or by modelling different parts of the beams with different properties. It
should be borne in mind that minor differences in stiffness due to
haunching can have less effect than some of the other assumptions
made during modelling (e.g. using gross uncracked section properties
for concrete). Variations in depth of +/- 25% can generally be ignored.
With greater variation in depth a haunched beam can be modelled by
three elements, one representing each haunch, with the moment inertia
taken as the average of the support and span values and the middle one
with the constant mid-span section properties.
 - 2.36 -

Isolated holes or recesses which extend over less than 5% of the length
of a perforated beam do not significantly affect its stiffness and can be
ignored in analysis. Larger holes with long slots or a perforated beam
can be represented in the model by splitting the member into sections.

Note that the slotted section can be modelled by two elements and the
their end nodes are rigidly linked to the centroidal nodes of the normal
end sections. The slotted section should be modelled as a beam with
reduced shear area.

A castellated beam with a line of holes, or even a truss can be modelled

as single shear beam.

The sectional properties of a single castellated beam can be defined as

It
It
 - 2.37 -

The structural action of a continuous truss is similar to that of a beam

Chords of the truss => flanges of a beam

Diagonals of the truss => web of a beam

Ad Ad Av
Av

Infilled Frames:

Brickwork or blockwork which infills the area between beams and

columns in a frame can make a major contribution to the lateral stiffness
behaviour. Under lateral load the infilled material can be modelled as
diagonal compression struts (without the capability of carrying tension).
The axial area of such a strut can be approximately equal to one tenth of
the diagonal length of the infill panel times the wall thickness.

 SHEAR WALL SYSTEMS

Concrete shear walls are commonly modelled as in-plane membrane

elements. If coupled beams are used to connect shear wall elements to
provide additional stiffness to form a coupled shear wall system,
membrane elements with in-plane rotation drilling degree of freedom
must be used.
 - 2.38 -
When finite element meshing is required, it should be borne in mind that
the accuracy of the model is dependent on how the structure is meshed.
When choosing a mesh of membrane elements for shear walls, the
following general rules can be considered.

- Use a finer mesh where the stress gradients are high and a coarse
mesh in the areas of more uniform stress.
- Choose a mesh that gives a converging analysis results.
- Keep the elements as ‘square’ as possible, i.e. keep length to breadth
ratio as close to unity as possible. The length-to-breadth aspect ratio
of 10 or higher will introduce inaccurate numerical errors.
- At a re-entrant corner the theoretical elastic stresses are infinite and
therefore mesh refinement in such areas can show very high
stresses. In reality there is likely to be localised cracking or plastic flow
such that the peak predicted stress may not occur.
- Higher order quadrilateral elements (e.g. eight-noded membrane) or
elements with more than four sides can help to improve accuracy, but
they should be avoided whenever possible. Higher order elements
require more computational effort to set up and their behaviour is
more difficult to interpret.
- With linear analysis use quadrilateral elements whenever possible in
preference to triangular elements since triangular membrane
elements produce uniform cross sectional stresses whereas
quadrilateral elements results in linear cross sectional stresses which
are more accurate.

In-plane stress distribution

between triangular and Two triangular elements Quadrilateral element is
quadrilateral elements are less accurate more preferable
 - 2.39 -

Case study 1:

Consider a cantilever wall structure for which the stress in the area of the
applied load is to be investigated.

The following figure shows a common error made on an attempt of mesh

refinement. At point ‘X’ there is a node which relates to the smaller
elements but not to the larger ones unless they have mid-side nodes or
a constraint is used to ensure that the boundary of the larger element is
compatible with the node X.
 - 2.40 -

As an attempt to improve accuracy, an economical refinement in terms

of numbers of elements and degrees of freedom is shown as follows.
While this mesh refinement is acceptable for computation, it may not
necessarily result in higher accuracy since the less accurate triangular
elements are used near the refined location.

The following two mesh refinements use rectangular elements which is

helpful for generating the mesh. The second one provides a refinement
without the use of triangular elements. When compared to the above
economical refinement, these two refinements can improve the
accuracy; however in some cases they may result in long and narrow
element shapes that are not all satisfactory.

The following figure presents a spider web approach which gives good
proportioning of elements but may be difficult to generate.
 - 2.41 -

Case study 2:

Consider low-rise wall with height = 3 m, width = 3 m and thickness =

0.15 m. It is modelled as sixteen membrane elements with three load
cases as shown.

Note that relatively high lateral deformation of point A in case 2 which

has the entire single point load at A. An even spreading of the loading
in case 3 results in a even distribution of internal forces and therefore the
top nodes of the walls are found to displace in an even manner. It is
interesting to note that an equivalent single flexural beam including shear
stiffness gives a fairly good estimate of the deflection.
 - 2.42 -

For the case where the shear wall is used to support a floor slab at the
top, the floor slab can be modelled as a rigid plate with infinite axial
stiffness so as to mobilise the top nodes of the highest levels to move
together with the same amount of lateral displacements.

Planar Shear Walls with Openings

If a planar shear wall has a complex geometry, with openings in different

places on different floors, planar membrane elements should be used to
model the structure and to identify possible concentration around the
openings. Out of plane bending of shear walls is normally neglected.
 - 2.43 -

Where the wall is supported by transfer beams and columns at ground

floor level, the basic mode of stress transfer is altered. Vertical stress is
transmitted to the columns by an arch action and the transfer beam at
the first floor level acts mainly as a tie to the arch.

Note that when a transfer beam is used to support the shear walls,
special care is needed to ensure that the beam is well connected to the
intermediate shear wall membrane elements throughout the length of the
beam so as to allow the loading from the walls to be resisted by the
beams in bending and shearing. Attention is also needed to make sure
that the axial stiffness of the transfer beam is modelled and tension
cracking is checked on the beam due to tying action.
 - 2.44 -
For simplicity, simple coupled shear walls with regular openings can be
modelled as an equivalent frames with sufficient accuracy. Indeed,
whenever practicable and easy to do, one should treat a wall system as
equivalent frame. This simpler model provides the simplest way to
understand the behaviour of the wall system especially during the
preliminary design stage. Another advantage of the frame model of a
coupled shear wall system is that the forces and moments obtained on
the equivalent beam elements can be used directly for steel
reinforcement calculations.

In the simplified frame model, the finite sizes of the beams and the
columns can be modelled as rigid arms with sufficiently large values of
bending stiffness.

The connecting beams in the frame model consist of a lintel beam with
or without certain effective width of the floor slab. In the cases of coupled
shear wall buildings in Hong Kong, the depth of the lintel beams (in the
order of 450 – 600 mm or higher) is normally much larger than the
thickness of the slab (typically 100-150 mm), the contribution of the floor
slab to the flexural beam is usually neglected.

Furthermore, lintel beams are normally sized to have their width to be

made the same as the thickness of adjacently aligned wall panels for
ease of construction. However whenever it is necessary to increase the
coupling action so as to improve the lateral stiffness of the coupled shear
walls, the width of the lintel beams can sometimes be made wider than
the adjacently connected wall panel.
 - 2.45 -

Floor Plates in Bending and Grillage Systems

Floor structures form horizontal flat planes in buildings. They join with the
vertical elements of the lateral stability system, allowing the building to
respond to both lateral and vertical loadings as an integral total structure.

While the primary function of a floor framing system is to transmit gravity

forces to the vertical columns and/or walls, it can also serve as a
horizontal diaphragm to distribute lateral loads to the vertical elements.
In the lateral load resisting system of building models, floor slabs are
often modelled as rigid floor diaphragms or membrane panels with
infinitely stiff in-plane axial stiffness.

When subjected to gravity loading, floor slabs can be more accurately

modelled by plate bending or shell elements. Note that the accuracy of
modelling using these elements depends on the meshing of the floor
slab.

Case Study 3:

Consider a symmetrical quarter of an interior panel of a flat slab system

with a uniformly distributed gravity loading. Due to symmetry, only one
quarter of the interior plate is modelled and it is supported vertically at A
and rotational restraints are required are imposed on the four sides about
axes parallel to the sides.

3.0
 - 2.46 -

At the column support, the floor slab is fixed or restrained using springs
to model the column stiffness. Note that a point support gives a
singularity on the solution which for elastic conditions gives a theoretical
stress infinity. To improve accuracy, it is best to model the finite size of
the column where slave nodes around the perimeter of the column are
constrained to deform the same as the master center node.

Four noded Kirchhoff thin bending plates are used to model the floor slab
and different mesh refinements have been used to for the model as
follows. Meshes BE1, BE2 and BE3 do not take account of the finite size
of the column whereas the other meshes do this by imposing a constraint
over the column area.
A

Y Y

[email protected]

A
 - 2.47 -

The results of different analyses are given in the following tables.

Bending

36 X 36

Meshes BE8 and BE9 are the most refined and it appears that the
predictions for moments are not significantly affected by mesh division
with this degree of refinement. Therefore the BE9 results are taken as
target values and are used for design purposes. Note that the empirical
method used by the BS8110 tend to produce conservative results.

Meshes BE4, 5, 6 and 7 all gives similar results. M YB is somewhat

underestimated by BE5 which is less accurate. Note that there is a
singularity at the support A for meshes BE1, 2 and 3 and at D for the
other meshes, and hence as the mesh is refined higher and higher
stresses are predicted at these points. The peak moment at these point
supports needs to be smoothed out to some degree.

As an alternative, the floor slab can be approximately modelled as

grillage system. Note that a grillage beam element requires both the
flexural and torsional stiffnesses to be properly modelled. As a common
practice, the second moment of area and shear area of a grillage element
is taken as that of the normal cross section but the torsion constant is
normally assumed to be half the value of the normal cross section.
 - 2.48 -

The results of the different grillage models are given in the following table
and are compared with that of the BE9 plate bending model. The G1 model
does not give useful results. The other models generally overestimate the
moments M XB , M YB and underestimate the moment M XC , M YC . G4, G5
and G6 give sensitive values for column moment M XA , M YA although the
column moment at the corner of the columns of BE9 cannot however be
treated as a target value since there is a singularity at the corner of the
column. In all cases, the mid-span vertical displacement at C is found to be
overestimated.
 - 2.49 -
MODEL EVALUALTION AND CHECKING

 The inexperienced engineer may be tempted to accept without question

the computer generated results of structural analysis packages.
Checking and verifying computer analysis results are important aspects
of design, which is not adequately covered in available literature. This
part of the lecture is concerned with the checking of analysis.

 When using computer models, the validity of results should be treated

with scepticism and every practicable measure taken to check them. A
careful check of the data input alone is not sufficient; the results should
also be compared with those obtained from a simplified hand calculation.
Order of magnitude errors will normally be revealed in this way.

 In treating a structure as an assemblage of elements, a mathematical

model is created in the computer analysis package. However, our
understanding of the behaviour of the structure as a whole is based on
models which rely on experience and intuition as well as on
mathematics; these models are most important on the checking process.
For clarity, the term ‘computer model’ is used for the computer-
processed mathematical model, and the term ‘checking model’ is used
for verification.

Identifying Data Errors

Errors can originate from three sources:

1. The main source of errors is in the data itself. Data must therefore be
checked very carefully. Syntax errors (data using the wrong format) can
be easily detected by most analysis software packages. Whenever this
happens the software program is terminated without producing any
results.
2. The second sources of errors are logical errors in the data. The software
package may execute the analysis and produce results. However the
results are manifestly wrong. This represents the most common and
difficult type of data to check.
3. Less commonly, program errors sometimes appear in commercial
software. There is no bug-free program in reality. When a computer
program grows big, it is unlikely not to have errors in it, but such errors
are more likely with new software than with well-used programs.
 - 2.50 -

Checking Data

Data items must be carefully checked in relation to:

- the format of the input

- the magnitude of the parameters

This is normally done against output or on screen. The following

procedures are good practice:

 An independent check by person not involved in setting up the data is

worthwhile.

 Check that fixed values of deformation (e.g. restraints) have been

properly specified. A misplaced restraint can cause significant
differences in the results.

 Check the boundary conditions. It is good practice to check that the sum
of the support reactions is equal to the applied load.

 Check for errors in the order of magnitude in material and section

properties.

 A screen check or printer plot of the geometry of the model provides a

visual check on a good proportion of the data.

 Check situations that may lead to ill-conditioning problems. Ill-

conditioning occurs particularly when there are large differences
between the stiffnesses of parts of a structure or when dealing with small
differences between large numbers. Typical ill-conditioning situations are

- a rigid structure on very flexible supports

- a member divided into many very refined elements
- an element with disproportional shape
- a free node with degrees of freedom not properly being restrained or
constrained to some other master nodes.
 - 2.51 -

Model Verification using approximate checking models

To check the validity of the output results of complicated structures, the best
approach is to develop simple approximate checking models and make use
of these checking models to check against the behaviour of the computer
models.

When comparing a checking model with the results from the computer
model, two possibilities arise:

1. The results from the two models are significantly different

In this case at least one of the models is wrong or both may be wrong.
Further checking data may reveal computer model errors. Examination
of the computer results may show trends which were not previously
recognised, which can in turn lead to a radical reinterpretation of the
behaviour of the structure. The knowledge thus gained makes computer
analysis a valuable aid to the understanding of structural behaviour.
Note that the checking model may also be incorrect. However the
chance of generating error in simpler model is likely reduced as
compared to the rigorous but complex computer model. Therefore, it is
important that the checking model should stay simple and often can be
processed by hand calculation.

2. The results compare favourably.

In this case an increasing confidence in the results of the computer

model is achieved. The validity of the computer model is likely confirmed.
Note that the possibility of their both being in error by the same amount
cannot be wholly discounted.

The following are the usual stages in model checking procedures.

Stage 1

The most basic level of checking is to assess the general form and
magnitude of the results; thus requires a qualitative understanding of
structural behaviour.

- Check the pattern and magnitude of displacement; is the general

deformed shape as expected?
 - 2.52 -

- Check the pattern and magnitude of stresses/internal forces; are they

sensible?

- Check for discontinuities in the results. If they occur where there are no
discontinuities in the structure or loading, then look for data errors.

Note that an error in the elastic modulus may only affect the predicted
displacements but not necessarily the internal stresses or forces.

Stage 2

The next stage is a hand calculation of the checking model. To make this
worthwhile, the checking model must be capable of taking account of the
behaviour of the structure to a reasonable degree of approximation. In this
stage, approximations are taken to reduce the original or a subset of the
complex computer model to a simplified form. For example, 3D structures
could be reduced and analysed as two dimensional.

A good strategy is to use the checking models which are known to give
results on one side of the expected values. For example, a model which has
internal relaxations such as torsionless grillage representation of a plate will
tend to be more flexible than the original structure, and will probably
overestimate stresses. If a stiffer model can also be used, then upper and
lower bounds to the solution can be established.

Further details on developing various approximated checking models will

be discussed later.

Stage 3

The last stage of checking is to repeat the computer model solution using
different software. It is best to ask an independent checker to do this task.
This is sometimes mandatory for important projects in China. However, this
does not guarantee that the model is correct.
 - 2.53 -

Some Simple Checking Models

Approximate Truss Model

A parallel chord truss has a structural action analogous to that of a beam.

The top and bottom chords are equivalent to the flanges while the posts
and diagonals are equivalent to the web. This observation leads to the
concept of treating the truss as an equivalent beam

Consider the given truss bent as shown,

The equivalent bending stiffness

EI e  EAcb 2 / 2

The equivalent shear deformation due to an applied shear S is

SLd Sb Sa Sa
s    
EAd sin 2  EAp EA sin 2   a EA a
d p
Ld b

Treating the truss bent as an equivalent shear beam of the length a, the
shear deformation of equivalent beam is
 - 2.54 -

Sa
s 
(GAs )e

Then, the equivalent shear stiffness can derived from

1 1 1 1 1
   
(GAs )e EA sin 2   a EA a EAd sin 2  cos EA cos
sin 
d p p
Ld b

Early approaches to finite element modelling of shear walls used equivalent

truss systems. Consider the cantilever shear wall of thickness t.

t
F

h h

b  b
Shear Wall Equivalent Truss Bent

Lateral shear deformation of the shear wall

Fh Fh
s  
GAs G  tb

The approximate lateral deformation of the equivalent truss, assuming a

negligible horizontal chord,

d ( F / cos ) Ld 1 FLd Fh
s     
cos EAd cos EAd cos 2  EAd cos 2  sin 

Therefore, the equivalent area of the diagonal of the equivalent truss is

2
Gtb tb  Ld   Ld  tL3d
Ad      
E cos 2  sin  2(1  v)  b   h  2(1  v)bh
 - 2.55 -

Using the concept of approximate truss model, complicated shear wall

structures can be simplified to equivalent braced frames as

Conversely, the concept of approximate truss model can also be used to

simplify complicated braced frame into simple analogous beam or column
models.

The flexural inertia of the equivalent column is assigned to have the same
values of the inertia of the column areas about their common centroid on
the braced frame. Also the shear area of the analogous column of each floor
level is assigned to have the total shear stiffness of the diagonals on the
level.
 - 2.56 -

Approximate Frame Model

A multi-bay rigid frame under lateral loading can be modelled very closely
with regard to its lateral behaviour by a single-bay rigid frame. The criteria
for equivalence are that the racking shear rigidity (GAs) as defined by the
column and beam flexural stiffnesses, the sum of column inertia, Ii and the
overturning bending stiffness as defined by the column sectional areas, are
at each level the same in the equivalent single bay frame as they are in the
multi-bay frame. The detailed derivation of these equivalent properties will
be given in the later part of the course.

A multi-story rigid frame of repetitive story heights can be simplified by

vertically combining groups of several beams (typically 3 – 5 beams) into
single beams, at the middle beam location, and assigning to them the
lumped properties of inertia and shear area.
 - 2.57 -

Similarly, repetitive coupling beams of a coupled shear wall structure can

also be lumped over several stories to a simplified coupled structure.

If shear walls are interconnected by rigid frame in the same plane to form a
wall-frame structure, the wall can be represented by an analogous “wide
column”, which is placed at the wall’s centroidal axis and assigned to have
the wall’s inertia and axial area, and rigid arms that join the column to the
connecting beams at each floor level.
 - 2.58 -

Simplification of 3D Building Structures

Nonplanar assemblies of shear walls are commonly found in buildings to

form elevator cores, which can be simulated by a single column located at
the shear centre of the core and assigned to have the principal sectional
properties of the core wall. If the core is effectively closed, box-like section,
the single column should be assigned with both bending and torsional
stiffnesses of the core.

If the core is an open-section, which translates and twists with warping

deformation, then the equivalent single column should be modelled with
both bending, torsional and warping properties. For such an open core, a
two-column approximate model can be used to give an approximate
representation of all the bending, St. Venant and warping torsional
properties of the open core.
 - 2.59 -

Reduction by Symmetry and Antisymmetry

A building structure that is symmetric in plan about the axis of horizontal

loading can be simplified and analysed as a half-structure subject to half of
the loads. The ends of the members along the cut by the line of symmetry
must be restrained against rotation and horizontal displacement in the plane
perpendicular to the direction of loading, and against rotation about a
vertical axis, while simultaneously being free to displace vertically and to
translate in the direction of the loading.
 - 2.60 -

A building structure that is symmetric in plane about a horizontal axis

perpendicular to the axis of horizontal loading behaves antisymmetrically
about the axis of symmetry. In such case, only half of the structure, to one
side of the axis of symmetry, and subjected to loads of half value, needs to
be modelled and analysed. The ends of the cut members along the line of
symmetry must be restrained against vertical displacement, but are free to
rotate in the axis of the anti-symmetry.
 - 2.61 -
The full model of a building structure is commonly built three dimensionally
in today finite element software packages. Under lateral loading analysis,
floor slabs are numerically modelled as rigid diaphragms with infinitely stiff
in-plane stiffness that mobilise all the joints on the same floor to move rigidly
together as on the rigid floor plate.

While it is relatively easy to model 3D building structures using today’s

engineering software, it is still difficult to check 3D models. Whenever
feasible and practicable, 3D building models should be reduced to planar
frameworks for checking and better understanding of their lateral behaviour.

Consider a four-story building with a typical floor plan given in the above.
The building can be simplified into three separate frameworks being linked
by axially stiff bar elements. Note that the reduction of a 3D building to a
planar equivalent structure forces the building to move in the way such that
its torsional effect is to be neglected. Therefore, this planar approximation
may underestimate the actual movement of the building and is good for
symmetric buildings or buildings with little torsional effects.
 - 2.62 -

For a crude checking, the planar structure can be further reduced to a

simple wall-frame with a single cantilever and a single-bay frame. Such a
crude model can be set up quite quickly and can be used to give a
reasonable estimate of the relative importance of the walls and the frames
in lateral load transmission.
 - 2.63 -
A more refined approximation can be developed by considering the 3D
building as a number of ‘support frames’ which are interconnected to form
the building structure. If the lateral stiffness of each individual support frame
can be accurately modelled as a spring, then the entire building can then
represented by a rigid plate supported by springs which simulate the
stiffnesses of the support frames.

The spring constant of each individual frame can be obtained by applying a

lateral load F and finding the a top lateral deflection  and then defining

F
ki 

 - 2.64 -
The building can be modelled by a rigid plate on springs. The in-plane action
is modelled by master-and-slave constraints with the springs being treated
either as joint elements or as bar elements with axial stiffness ki.

Alternatively, the in-plane rigidity of the floor plate can be modelled by stiff
beams with sufficiently large bending stiffness to mobilise the springs to
simulate the behavior of the building.

Unlike the approximate planar framework model, the above two spring
models can be used to capture both the lateral displacement and torsional
twisting effects of the building.