Applied Elements Method

Table of Contents ...........................................................................................................................

1 Introduction............................................................................................................................

2 Basic Theory behind AEM ..................................................................................................

2.1 Connectivity Springs ...............................................................................................................

2.1.1 Matrix Springs .....................................................................................................................................
2.1.2 Reinforcement Springs ........................................................................................................................

2.2 Constitutive Models and Failure Criterion ...........................................................................

2.3 Elements Collision ...................................................................................................................

2.4 Degrees of Freedom .................................................................................................................

2.5 Global Stiffness Matrix ...........................................................................................................

2.6 Modeling of 3-D Structures at AEM vs. FEM ......................................................................

2.6.1 Connectivity between Different Components ......................................................................................
2.6.2 Modeling of Steel Sections ..................................................................................................................
2.6.3 Modeling of Reinforcement Bars.........................................................................................................

2.7 AEM Validation .......................................................................................................................

2.8 References ................................................................................................................................

4
1 Introduction
A structure may be subjected during its lifetime to extreme loading conditions that exceed its
design loads. Amongst these loading conditions are major earthquakes, explosions, unexpected
impact forces, and fire. Typically, structures are not being designed to resist such extreme loads
due to economic reasons. Life safety considerations necessitate that in the event of an extreme
loading condition on a building, people can be evacuated safely before the building collapses.
This requires a forecast about whether a building would eventually collapse in such an event or
not, and such a forecast is applicable for new structures as well as existing structures. By
reviewing the casualties caused by previous major earthquakes around the world, it was found
that more than 90% of death toll was due to structural collapse of buildings and bridges. A
forecast that the towers of the World Trade Center would collapse from the extreme impact load
and fire resulting from the plane crashes on Sept. 11th, 2001 may have saved thousands of
human lives.

Design of structures to resist progressive collapse has been the focus of building code changes in
the past few years. To be able to prevent progressive collapse, the design engineer should be able
to accurately predict how the structure would respond to the sudden loss of one or more of the
supporting columns.

Computer simulation is an important key to determine performance of structures in extreme

loading conditions. The Finite Element Method (FEM) was proven to be an accurate method for
structural analysis under different kinds of loads. However, for analysis of structures under blast
loads or for progressive collapse, FEM cannot be easily used. FEM is based on continuum
mechanics where elements are connected throughout the analysis. Separating these elements
from each other is possible, but needs a qualified researcher to be able to perform this task. This
is why FEM applications for progressive collapse were kept only for researchers rather than
engineers. On the other hand, analysis methods based on rules of discrete material cannot be
used to predict behavior of continuum elements. Structures during a collapse situation pass
through a continuum stage first, then through a discrete stage. The computer simulation needs to
follows the two behavior stages in order to answer the following questions:

 Will the structure collapse in an extreme loading event?

 What is the mode of collapse of the structure?

1
  How long would it take for the structure to get to a complete collapse?

 What happens to adjacent structures when falling debris collide with them?

 What is the footprint of debris due to the collapse of a structure?

These questions are samples of questions that cannot be answered without having a forecast of
the structural performance when subjected to an extreme loading event.

Applied Element Method (AEM) was developed at the Disaster Mitigation Engineering Lab at
Tokyo University in 1995 to create a link between the advantages of FEM for continuum
mechanics and DEM for discrete mechanics. It was proven that the method can follow the
behavior of the structure through the application of loads, crack initiation and propagation,
element separation, formation of debris and collision between falling debris and other structural
components in reasonable time, reliable accuracy and using relatively simple material models.

The purpose of this report is to validate the AEM. Since AEM was developed more than 30
years ago, it was important to present the validity of AEM through research and application that
AEM could perform besides to the tasks conducted at this proposal with blind numerical tests.

This report consists of 8 chapters including the introduction (1) above.

Chapter 2 covers the basic theory behind the AEM. It discusses the element and spring
formulation, degrees of freedom, stiffness matrix formation, element collision, etc. It also shows
the ways to model reinforcement bars and steel sections. It also discusses advantages of the AEM
compared to the FEM when creating complicated structural models.

Chapter 3 discusses the effects of Poisson’s ratio, which assumed equals to zero in the AEM
formulation. It shows the conclusions of some researchers regarding this material property and
when it should be considered or when it can be ignored. It also shows various progressive
collapse and blast studies where Poisson’s ratio was not considered while the results obtained
were accurate and reliable.

Chapter 4 shows results of the first blind numerical test of 9 walls with different supporting
conditions subjected to various blast loads. The results are compared to the FEM results output
from LS-Dyna®. The effects of element size and time step are also studied using one of the wall
models.

2
Chapter 5 shows the comparison of the AEM to a reinforced concrete (RC) column subjected to
blast loads. The column had severe deformations but did not collapse. The analysis results of
AEM are compared to the physical test results. Different modeling options of reinforcement
rebar are also presented.

Chapter 6 presents the blind numerical test comparison of the AEM to an RC column subjected
to vertical loads plus high pressure loads from a blast. The AEM results are compared with the
physical test results.

Chapter 7 presents an extensive study of a 5 story reinforced concrete column- flat slab building
subjected to different columns removal. The building behavior is highly complicated and
dependent on the reinforcement curtailments, cracking and crushing of concrete, yield of
reinforcement rebar and element separation. AEM is compared to a proprietary FEM software
program and the results are discussed in detail.

Chapter 8 presents the general conclusions of Phase I

Finally the Appendix A discusses samples of research using the AEM in the fields of progressive
collapse, earthquake engineering, blast analysis, etc. The research was conducted at well-known
respected universities all over the world. It also includes various validation examples that were
performed in-house by ASI scientists and engineers. This appendix also summarizes some of
previous demolition analysis and collapse analysis that were conducted by ASI.

3
2 Basic Theory behind AEM
The Applied Element Method (Meguro and Tagel-Din 2000, 2001, 2002) is an innovative
modeling method adopting the concept of discrete cracking. In the Applied Element Method
(AEM), the structures are modeled as an assembly of relatively small elements, made by dividing
of the structure virtually, as shown in Figure 2-1b. The elements are connected together along
their surfaces through a set of normal and shear springs. The two elements shown in Figure 2-1c
are connected by normal and shear springs located at contact points, which are distributed on the
element faces. Normal and shear springs are responsible for transfer of normal and shear stresses,
respectively, from one element to the other. Springs represent stresses and deformations of a
certain volume as shown in Figure 2-1c.

Concrete spring

b Volume represented by
a normal spring and 2
shear springs

a
Reinforcing bar spring

a. Structure b. Element generation for c. Spring distribution and area of

AEM influence of each pair of springs
Figure 2-1 Modeling of Structure to AEM

2.1 Connectivity Springs

AEM elements are connected together through a series of springs that connect adjacent element
faces. The generation of these springs is automatically performed in the software. These springs
represent continuity between elements. The springs reflect the different material properties.
Strains, stresses, and failure criteria are all calculated and estimated using these springs. This
section describes the different types of springs used by AEM

2.1.1 Matrix Springs

Referring to Figure 2-2, “matrix springs” are those springs that connect two adjacent elements
and are representing the main structural material. For example, for a reinforced concrete
structure, these springs represent the concrete material; for a steel structure, these springs
represent the steel material. These springs adopt all material type and properties defined in the
user interface. In order to automatically generate the springs between faces, the faces must be
4
exactly in the same plane. Otherwise, springs will not be generated at connecting faces, as
demonstrated in Figure 2-3, and the two elements behave as if they belong to two separate
objects separated by a gap.

Z Y Element 1 Element 2

Element 1 Element 1 Element 1

Normal Springs Shear Springs x-z Shear Springs y-z

Figure 2-2 Connectivity Matrix Springs

Connectivity
springs generated

No connectivity
springs generated

>>
Figure 2-3 Condition for Creation of Matrix Springs
When the average strain between these two adjacent faces reaches a specific limit called the
separation strain which is specified in the material property section in the user interface, springs
between these two faces are removed and it is assumed that these elements behave as two
separate rigid bodies for the remainder of the analysis. Even if the same two elements meet
again, they meet as a contact between two separate rigid bodies.

5
At every calculation point, three springs are set for matrix springs. One for normal stresses and
the other two springs are for the shear stresses. Average normal strain is calculated by having the
average of absolute values of strains on each face. When the average strain value at the element
face reaches the separation strain, which is usually a large strain that confirms full element
separation, all springs at this face are removed and elements are not connected any more until
they collide. If they collide together, they collide as rigid bodies as will be discussed at Sec. 2.3.

2.1.2 Reinforcement Springs

In reinforced concrete structures, the “Matrix Springs” are those springs representing the
concrete material as shown in Figure 2-2. The reinforcement springs are those springs
representing the existence of steel bars as shown in Figure 2-4.

Z Element 1 Element 2
Y Reinforcing
bar
x

Element 1 Element 1 Element 1

Normal Springs Shear Springs x-z Shear Springs y-z

Figure 2-4 Reinforcement Springs
The reinforcement springs represent the material properties, exact location and area of the
reinforcement bars. Similar to matrix springs, three springs are used to represent all the possible
force components. These springs are set at the intersection of the reinforcement bar and the
element boundary, as shown in Figure 2-4, and they are generated automatically by the ELS®
software. The normal spring takes the direction of reinforcement bar irrespective of the element
face direction. The other two springs represent reinforcement bar behavior in shear. The
reinforcement bar springs are cut when the reinforcement bar stresses satisfy the Von Misses
failure criteria illustrated in Figure 2-7

6
2.2 Constitutive Models and Failure Criterion
Fully nonlinear path-dependent constitutive models for reinforced concrete are adopted in the
AEM as shown in Figure 2-5. For concrete in compression, elasto-plastic and fracture model is
adopted (Maekawa and Okamura 1983). When concrete is subjected to tension, linear stress-
strain relationship is adopted till cracking of concrete springs, where the stresses drop to zero.
For steel bars, Ristic et al (1986) model is used for the envelope and interior loops.

The concrete is cracked when the principal tensile stresses reaches the cracking strength of
concrete as shown in Figure 2-6.

Referring to Figure 2-5, the concrete behavior in shear is linear till it reaches the cracking strain,
which is calculated based on principle stress criteria. Once the springs reach the cracking
criterion, the whole shear strength value at the face of the crack is redistributed (RV=1.0). Once
the crack is closed, the shear behavior is back in effect in linear way and it is dominated by the
friction coefficient as far as the crack is closed. This indicates that all cracks at the model are
considered as physical cracks that open and close in a way close to what happens in reality.

Steel
Concrete

Figure 2-5 Constitutive Models for Concrete and Steel

7
 z

xz
x
yz
xy
y y
yx zy
zx z
x

xz x
yz
y xy
y
yx zy
zx z
x Plane of major
principal stress
Figure 2-6 Cracking Criterion in AEM
Implicit steel structure is modeled as shown in Figure 2-8. The cross section is divided into cells
where steel springs are put at the location of steel section elements. There are three main
assumptions for the modeling of steel sections:

1- The interaction between the normal stresses in the two perpendicular directions of the
plates constituting the steel section is not taken into consideration. This is in fact acceptable
when dealing with frame structures where the longitudinal direction represents the main
direction of stresses in the steel section.

2- If two steel plates segments meet at the same element, then these two segments are always
connected and never separate. Since the element itself is not allowed to separate, the failure
will always be allowed at the element boundary instead of the intersection between the
steel plate segments. Hence, elements of small size should be used to allow for failure
inside the steel connections. When elements of large size are used at connection, the failure
will occur at the connection interface with members rather than inside the connection itself.

3- The steel spring failure criterion, either RFT spring or Implicit Steel or Explicit Element, is
following Von Misses as shown in Figure 2-7

8
 Failure envelope

N
V

2 2 2
 N  V   M 
         1
 N P   VP   M P 

Figure 2-7 Failure criteria of steel springs

l
na
di
itu
ng
lo
io in
ct s
re g
n
di prin
S

Springs in section

Vacuum cells Steel Springs

Figure 2-8 Modeling of Steel Members

As for bare steel members, the same concept is followed where only elements containing parts of
the steel section are used while other elements are considered as vacuum elements with no
material inside as shown in Figure 2-8. It should be emphasized that the failure criteria for steel
bars or implicit steel springs is the same, following 1-D models and Von-Misses criterion for
cutting the reinforcement bars or implicit steel.

9
2.3 Elements Collision
One of the main break-through features in AEM is automatic element contact detection. The user
does not have to predict where or when contact/separation will occur. Elements may contact and
separate, re-contact again or contact other elements without any kind of user intervention. This
reduces the dependency of the obtained results upon the qualifications of the user.

There are several types of contacts; element corner-to-element face contact, element edge-to-
element edge contact and element corner-to-ground contact. Figure 2-9, Figure 2-10 and Figure
2-11 illustrate the three types of contact, respectively. Whenever elements are in contact, normal
and shear springs are added at the contact location and they are removed when elements separate.

Shear spring in y Shear spring in X Normal Spring

Figure 2-9 Corner- to-Corner Contact
r
hea
spr tact s

ar
he
ing

s
Con

t
ac
ont g
C ri n
sp
Contact
normal spring

Figure 2-10 Edge-to–edge contact

Ground Ground
Ground

Shear spring in y Shear spring in X Normal Spring

Figure 2-11 Corner- to-Ground Contact

10
2.4 Degrees of Freedom
Each single element has 6 degrees of freedom; 3 for translations and 3 for rotations. Relative
translational or rotational motion between two neighboring elements cause normal or shear
stresses in the springs located at their common face as shown in Figure 2-12. These connecting
springs represent stresses, strains and connectivity between elements. Two neighboring elements
can be separated once all the springs connecting them are ruptured.

Figure 2-12 Stresses in springs due to relative displacements

2.5 Global Stiffness Matrix

Referring to Figure 2-13, for each spring connecting two elements, a 12x12 stiffness matrix is
generated (since every element has 6 DOF). This matrix depends at the spring material and
status. For example, it takes into account whether the spring is for steel or concrete. In case of
concrete, it considers whether the spring is already cracked or reached compressive failure
criteria. In case of steel, it also follows the constitutive stress-strain relations. These matrices are
assembled at the structural overall stiffness matrix. The dynamic equation and the solution
procedure is the same as the conventional implicit FEM as shown in Figure 2-14

11
 Figure 2-13 Assembly of Global Stiffness Matrix

M yi   Ci y i   K i yi   Fi  Incremental Equation of Motion

requested
Step-by-step integration (Newmark-beta) method

y i   yi t  yi t

Acceleration

yoo(ti+t)
y i   y i t  1 yi t 2  yi t 2
2

t

Time
ti ti +t
Figure 2-14 Solution of Dynamic Equations at AEM Model

2.6 Modeling of 3-D Structures with AEM vs. FEM

It should be emphasized that the AEM has unique advantages in modeling that make it much
easier that the FEM. These advantages are the reason behind the fast modeling performed for all
case studies presented in this report. The simplicity in modeling described below makes it easy
even for non-engineers to build a very complicated model. Since the engineer or scientist time is
always limited and expensive, having the model totally generated by non-engineers saves a lot of
time and makes it handy tool for many projects. The scientist or engineer role starts after
building the models to generate proper loads or boundary conditions which are usually fast
compared to building the model itself. The main reasons behind the fast modeling are
summarized in this section:

12
2.6.1 Connectivity between Different Components
It is well known that building a reliable FEM model takes, in complicated cases, more time than
the time needed to solve the problem itself. The reason behind this is that elements should all be
connected through nodes. This is one of reasons why the Mesh-Free methods are developed to
make it easy to build complicated 3-D meshes without needs to connect nodes through elements.

Referring to Figure 2-15, the two FEM elements are not connected due to partial overlap
between elements. The only way to connect these two elements is to make a special constraint to
link the adjacent nodes to each other. This process is of course time consuming and needs a
qualified user to make it. The use of transition elements to connect adjacent components is also a
complicated process especially in 3-D and results in many added elements. Contrary to FEM, the
same two elements are connected automatically with AEM. This feature is one of reasons why
AEM is much faster in modeling.

FEM AEM

No Connectivity Connectivity included

Figure 2-15 Connectivity between Elements at FEM vs AEM
For example, consider a single bay frame with two columns and one girder modeled using 3-D
brick elements. Both FEM and AEM can generate the mesh easily for each of the columns and
the girder. While this is what is needed for AEM, the FEM user still needs to adjust the
connectivity between the columns and the girder.

Referring to Figure 2-16, the bridge components are all meshed independently and this makes
the whole process of building the model fast as the user does not have to adjust the mesh at
interfaces. The same concept applied for the inclined shear wall shown in Figure 2-17. While it is
complicated to adjust the wall nodes to the slab nodes, no action is needed using AEM. Finally
the model shown in Figure 2-18 may takes days to adjust the mesh of each brick to the adjacent
brick and also to adjust the glass window mesh to the window frame, while no action is needed
13
using AEM. It is still recommended to have elements of comparable size adjacent to each other;
however, the elements do not have to match exactly the corners to each other.

Easy Mesh Generation

Figure 2-16 Simplicity of Modeling a Bridge Deck using AEM

FEM AEM

Auto elements
connectivity

Difficult meshing
and For
Compatibility
Merge Nodes of
Slab and Column

Figure 2-17 Simplicity of Modeling Shear Wall at AEM Compared to FEM

14
 Concrete Slab

Girder

Win
dow
Fra
me
Column

Brick
Wall

Glass Windo
w

Figure 2-18 Simplicity of Connecting Different Components at AEM

2.6.2 Modeling of Steel Sections

Considering the simplicity of connecting adjacent structural components to each other which was
described at section 2.6.1, it is now easy to build real 3-D steel structures using AEM. It is
already known that using 3-D brick elements with FEM to build a fully 3-D steel structure, as the
model shown in Figure 2-19 and Figure 2-20, is time consuming considering the connectivity
between different steel plates especially when the plates are of different width or thickness.
Referring to Figure 2-20, using AEM modeling, the steel sections are generated and connected
fully from styles shown in Figure 2-19. So, the user can be even non-engineer who can build the
whole steel structure model. No further actions are needed to connect these two adjacent sections
and this reduced the modeling time and modeler qualifications considerably.

15
 Figure 2-19 Modeling of Complicated Steel Structures using Styles

Figure 2-20 Automatic Connectivity at Interfaces Makes it Easy to Build Complicated Steel Structures.

Figure 2-21 Reinforcement Details are Generated from Cross Section Styles

16
2.6.3 Modeling of Reinforcement Bars
Since reinforcement bars are modeled using springs, it is sufficient for the software to know the
start, the end, area and material of each reinforcement bar segments in the model to generate all
steel springs. In brief, the user interface deals with real cross section reinforcement details,
covering, stirrups spacing, etc., which the user is familiar with. Referring to Figure 2-21, it is
only important to know the girder and pier cross section details, like concrete cover, number of
bars, bar diameter, etc., to generate the whole reinforcement detail for the whole bridge model.
The calculation core converts such reinforcement bars into normal and shear springs at the
appropriate locations. Using springs to model rebar means that they are perfectly bonded to the
concrete. However, it is still an option to the user to model steel bars using elements to control
the bond behavior between the rebar and the concrete.

2.7 AEM Validation

Since AEM was developed 1995, a lot of research has been conducted using AEM in 2-D and 3-
D. Appendix A of this report shows the current research conducted using 3-D AEM.

2.8 References
1. Maekawa, K. and Okamura, H. (1983). “The deformational behavior and constitutive equation of concrete
using the elasto-plastic and fracture model”, Journal of the Faculty of Engineering, The University of Tokyo
(B), 37(2), 253-328.

2. Meguro, K., and Tagel-Din, H. (2001) “Applied Element Simulation of RC Structures under Cyclic
Loading”, ASCE, 127(11), (pp. 1295-1305).

3. Meguro, K., and Tagel-Din, H. (2002) “AEM Used for Large Displacement Structure Analysis”, Journal of
Natural Disaster Science, 24(1), (pp. 25-34).

4. Meguro, K. and Tagel-Din, H., (2002) “AEM Used for Large Displacement Structure Analysis,” Journal of
Natural Disaster Science, Vol. 24, No. 2, pp. 65-82.

5. Ristic, D., Yamada, Y., and Iemura, H. (1986) Stress-strain based modeling of hysteretic structures under
earthquake induced bending and varying axial loads, Research report No. 86-ST-01, School of Civil
Engineering, Kyoto University, Kyoto, Japan.

6. Tagel-Din, H. and Meguro, K., (2000) “Applied Element Method for Simulation of Nonlinear Materials:
Theory and Application for RC Structures,” Structural Eng./Earthquake Eng., International Journal of the
Japan Society of Civil Engineers (JSCE) Vol. 17, No. 2, 137s-148s, July.

17