Vietnam J ournal of Mechanics, VAST, Vol. 26, 2004, No.

1 (39 - 54)

BEAM ELEMENT FOR LARGE DISPLACEMENT

ANALYSIS OF ELASTO-PLASTIC FRAMES
NGUYEN DINH KIEN 1 AND D o Quoc QuANG 2
1
Institute of Mechanics
Vietnamese Academy of Science and Technology
2
Research Institute of Technology for Machin ery, Ministry of Industry

ABSTRA CT. The present paper develops a non-linear beam element for analysis of elasto-
plastic frames under large displacements. The finite element formulations are derived by
using the co-rotational approach and expression of the virtual work. The Gauss quadrature
is employed for numerically computing the element tangent stiffness matrix and internal force
vector. A bilinear stress-strain relationship with isotropic hardening is adopted to update
the stress. The arc-length technique based on the Newton-Raphson iterative method is
employed to compute the equilibrium paths. A number of numerical examples is employed
to assess the performance of the developed element. The effects of plastic action on t he
large displacement behavior of the structures as well as the expansion of plastic zones in the
loading process are discussed.

1 Introduction
In the previous work [1, 2], the authors investigated some beam elements for assessing the
behavior of elastic frames under large displacements. The finite element formulations in
the work have been developed by using the co-rotational approach, in which an element
attached coordinate system which continuously rotates and moves during the element
deformation process was employed. The approach allowed to derive the finite element
formulations in a local system, and then transfer them to a global one with the aid of
the transformation matrices . In add ition, the stress-strain relationship was assumed to be
linear, and this assumption enabled to derive the element formulations from the expression
of strain energy. As consequences , t he explicit forms of the element tangent stiffness matrix
and internal force vector have been obtained.
The present work aims to develop a beam element for large displacement analysis of
elasto-plastic frames by lifting the restriction of linearly elastic assumption in [1, 2], and t he
behavior of the structural material is now supposed to be elasto-plastic. T he co-rotational
approach is still adopted herewith, so that the work will mainly focuss on t he derivation
of formulations in the local system. The extension of the finite element formulat ions from
the elastic case to the elasto-plastic one is, however, not straightforward and not simple
either. With the involvement of plasticity, the structure became a dissipate system [3],
and one could not able to derive the expression of strain energy. As a result, the principle
of stationary potential energy, a powerful tool for elastic systems, is not able to apply for
constructing the element formulations . Furthermore, the numerical integration is often

39
needed to compute the tangent stiffness matrix and internal force vector, since these
quantities are hardly derived in explicit forms for the plastic case. As a consequence, the
computation of any real structure involving plasticity is very computer time consuming.
Taking elasto-plastic law of the material behavior into account, the derivation of finite
element formulations for large displacement analysis of frames is rather a complicated
task, since both the geometrically and materially nonlinearities should be considered si-
multaneously. The number of related elements found by the authors in the literature is
quite limited. An early element, a hybrid type, was formulated by Backlund [4], in which
the element stiffness matrix was formulated from the inversion of a flexibility matrix. In
[5], Cichon constructed a beam element in the context of total Lagrangian formulation,
which requires higher-order terms for assessing the new element length. In [6], Hsiao and
co-workers also employed a 'body-attached' coordinate system for developing element , but
the method is far from the co-rotational approach initiated by Belytschko et al. [7, 8],
due to their way of choosing the local coordinates. As a result , the method by Hsiao and
co-workers lent the local tangent stiffness matrix and internal force vector with the same
sizes as the global ones .

z current configuration

2
initial configuration
x

Fig . 1. Element configurations and kinematics

In [8], Belytschko et al. presented a co-rotational rod element for dynamic analysis of
frame structures. In the same context of co-rotational approach in [8], Crisfield formulated
a beam element based on linear definition of the local strain for instability analysis [9].
Both the work discussed in [8 , 9] employed a local coordinate system, which continuously
moves and rotates with the element during the deformation process as above mentioned.
As a result , the local formulations are simple, and the size of these formulations is much
smaller than that of the global one. The material behavior in both the work in [8, 9] is,
however, assumed linearly elastic.
The present paper adopted the co-rotational approach discussed in [8 , 9] to develop a
nonlinear beam element for large displacement analysis of elasto-plastic frames . Thus, the
main difference with the work in [8 , 9] lies on derivation of the local formulations , which is
presented below in Section 4. These local formulations, formulating from a bilinear elasto-
plastic behavior, cannot be derived explicitly, and thus require the numerical integrations.

40
The involvement of the plastic deformation makes the element formulations to be complex,
and the analysis is much more expensive. Besides that , a direct approach for constructing
the relationship between the local and global nodal displacements, presented in Section 2,
is also different from the approach discussed in [9], where a geometric method has been
adopted.

2 Local and global relationships

The formulations needed for transferring the local formulations to the global ones in the
elastic case can be derived with the aid of strain energy expression [1 , 2] , but this approach,
as discussed above, is not proper for a dissipative system of present consideration of the
elasto-plastic structure. The formulations, therefore will be derived in this section by using
expression of the virtual work, instead of the strain energy expression. Fig. 1 shows the
configurations of a beam element in the initial and current deformed states. The element
consists of two nodes , and in respective to the global system (x, z), there are three nodal
degrees of freedom (d.o.f) at each node: axial and transversal displacements and a rotation.
The element attached coordinates (x1, z1 ) (also called the local system below) are chosen
as well as its original is always placed at node 1, and the x1 axis is directed to node 2.
With this definition, the axial displacement at node 1 and both transversal displacements
in the local system are vanished. Accordingly, vectors of the element nodal displacements
in the systems are respectively given by

d1 = {u en B12}r, (2.1)

d = {u1 w1 B1 u2 w2 B2}T , (2.2)

where and later, the subscript l stands for 'local '; ( ...)T denotes the transpose of a vector
or a matrix; u is the local axial displacement of node 2. The local d.o.fs are related to the
global ones by (see Fig. 1)

(2 .3)
where l 0 , ln are the initial and current lengths of the element

v(x2 - x 1) 2 + (z2 - z1) 2 l (2.4)

V(X2 + U2 - X1 - U1) 2 + (z2 + W2 - Zl - wi) 2 , (2.5)

and er is the rigid rotation, which can be computed from geometry consideration as

sin er = SCo - CSo ==} Br = arcsin( SCo - CSo) l (2.6)

with
X2 + U2 -
c = cos e = - - -- - - -
X1 - U1
(2.7)
Zn '

(2 .8)

41
where Ba, B are the initial and current inclined angles of the x 1 axis. Eq. (2.6) is prevailed
with Br ::; f7r / 2 f, but by computing cos Br, the extension to the range of 21T is straightfor-
ward.
From Eqs . (2.4)-(2.8), the local virtual d .o.fs can be computed as

c(6u2 - 6u1) + s(6w2 - 6wi) = [-c - s 0 c s 0] 6d , (2.9)

be1 - ber = be 1 - be, (2 .10)
be2 - ber = be 2 - be, (2 .11)

where bd = { bu1 bw1 be1 bu2 bw2 be2 V is the vector of virtual global d .o.fs . The far
righthand sides of Eqs . (2.10) and (2 .11) have been written with notice that be0 = 0. By
differentiating sine, as defined in (2 .8), one can obtain the expression for be as

be = (bw2 -:-- bw1) - sc(bu2 - bu1) - s 2 (bw 2 - bw 1) .

(2.12)
cln
With some calculations, Eq. (2 .12) can be written in the form

1
i5e = Zn [s - c 0 -s c OJ. (2.13)

From (2.9)-(2.11) and (2 .13) we get

bd1 = Bbd , (2.14)

with the transformation matrix B given by

l[- cln - sZn 0

B = - -s c Zn (2.15)
Zn - s c 0

where c and s are given by Eqs. (2.7) and (2 .8), respectively.

Having the transformation matrix B derived, we can now relate the local and global
formulations by equating the virtual work in both the coordinate systems as

(2.16)

where f1 = {N1 Mn M12V and f = {N1 Q1 M1 N2 Q2 M2V are the vectors of local
and global internal forces, respectively. From (2.16), the global internal force vector is
given by

(2.17)

The element tangent stiffness in the global system kt can be obtained from differentiation
of the global internal force vector f , Eq.(2.17) , as

(2.18)

42
where Bi, (i = 1.. .3) is the ith column of the matrix BT. From (2.15) one can get

6B1 = zz:T 6d, 6B2 = 6B3 = (rzT; zrT) 6d, (2.19)

n
where r and z denote the vectors with following components

r = [-c - s 0 c s OJT, (2.20)

Z = [s - C 0 - s c OJT . (2.21)
From Eqs. (2.18)-(2.21) , the expression for the tangent stiffness matrix kt can be written
as
T zzT (rzT + zrT)
kt = B
(6 x 6)
kt1 B + -z-Nl
(3x3) n
+ z2
n
(Mn+ M12) , (2.22)

with kt1 = 8fl/ 8d1 is the local element tangent stiffness matrix.
A combination of Eqs. (2.17), (2.22) with Eqs. (2.15), (2 .20) and (2.21) completely
defines the global internal force vector and tangent stiffness matrix through their coun-
terparts, and the remaining work for obtaining the element formulations is to formulate
the local internal force vector f1 and the tangent stiffness matrix k 1, and this work will be
presented in Section 4 below.

3 Elasto-plastic law
A bilinear stress-strain model with isotropic hardening rule for the material as shown
in Fig. 2.a is adopted in the present work. The model is represented by the so-called
'elasto-plastic' modulus E, which defined as

jj; = 80' = { E if IO' I < O'y or unloading

(3.1)
8E Et if l(J'I 2:: O'y,
where E , Et are the elastic and tangent modulus , respectively; O'y is the current yield
stress with initial value O'Q, :which can be determined from a simple tensile test. For the
reverse loading path in Fig. 2.a, the yielding occurs when the stress value reaches O's-the
stress at unloading point B. The kinematic hardening rule, which takes Bauschinger effect
into consideration [10], is also widely employed in the field of structural analysis, but the
present work does not investigate the effects of different hardening rules.
Suppose yielding has already occurred, then an strain increment de , from point B to
point C in Fig. 2.b can be regarded as composed of elastic and plastic parts: de= dep+dee·
The corresponding stress increment dO' can be computed in various ways as (see Fig. 2.b)

(3.2)
where H is the strain hardening parameter, which related to E and Et by

H= Et
1 - (Et/E)
or Et = E (1 - E+H
E ) · (3.3)

43
If E is finite , and Et = 0, then H = 0, and the material is 'elastic-perfectly' plastic [10].
If the material makes a transition from elastic to plastic, such as from point A to point
B in Fig. 2. b, the stress increment is computed by calculating the 'elastic' fraction m as

m= CTo-ICTAI
ICT Bel - ICT Ai
sothat d [ E
CT= m + (1 - m
)E]d
t c:,
(3.4)

where CTA is the stress at point A; CT B e is the 'elastic' stress at point B , computed by
assuming the total stress increment is elastic; de: =EB - EA is the total strain increment.
The absolute values of the stress are needed for the case of compressive stress .
cr
loading path
CYB 1--------=-r-"~

cr
cr o
crlle / dep dee
unloading path I -.- - -1>r-+<-
- I I I
aB t dcr; C
cr o /

reverse loading

(a) (b)

Fig. 2. Bilinear elasto-plastic model (a) , and a stress increment (b)

4 Local formulations
This section formulates the local internal force vector and tangent stiffness matrix. To this
end, a definition for local strain should be adopted, and in accordance with the previous
work in [2], the shallow arch theory is employed herewith

Ex l = Ee f + ZP!l =
1 [
loJo
10
[ou1
OXl
(8w1)
1
+ 2 ox1
2
]
dx1 - Zl OXf ,
8w1 2
(4.1)

where u1 is the local curvature; Eef is the effective strain, as explained in [2], employed
in order to avoid the locking problems [9, 11]. To express the strain in Eq. (4.1) in terms
of the nodal displacements defined by Eq. (2 .1) , the usual interpolation schemes, a linear
function for u1 and the Hermitian polynomials for w1 are adopted as

u1 = ~
-u
-
wi-
_{ ((ee -- 1)1)(~+1)
(~ - 1)} T { Bn }
012 ,
(4.2)
lo
where~= (2xlfl 0 - 1) , -1:::;: ~:::;: 1, is the natural coordinate, which introduced in regard
of the later development on numerical integration. Substitute (4.2) into (4.1) and notice
that
0 1
~ = l~ : ( one can easily get the expression for the local strain, which is now, for
the sake of simplicity, denoted by c:

c: = lo1 u- + 1 ( Bn - 2 2
1 BnB12 + B12 2) + zzlo [(1 - 3~)Bn - (1 + 3~)012] . (4.3)
15

44
The element virtual work in the local system can be written as

(4.4)

where Vis the element volume; dV = dAdxz = (la/2)dAd~, with A is the cross-sectional
area of the beam ; u is written in place of O"x l· From (4.3), the virtual strain bE can be
easily computed as

(4.5)

Substitute (4.5) into (4.4), the local internal forces are obtained as

Nz = ~
2
f
- 1
1 {
}A
udAd~, (4.6)

Mn = 2
la ( 2
15
en - 1_ Jr 1
30
e12 )
1

1
A udAd~ + 21
_ Jr
1
1

1
A zz(l - 300"dAd~, (4.7)

(~ez2 - ~en) 1 ~1
1 1
M12 = la { O"dAde - { zz(l + 3e)O"dAde. (4.8)
2 15 -1lA 30 -llA
2
Having the local internal forces computed, the coefficients for the local tangent stiffness
matrix ku are obtained from differentiation of fz , with respective to the local d.o.fs. Notic-
ing bO" = E&, with 8E given by Eq. (4.5), we can get

ku(l , 1) =
2~a 1:LEdAd~, (4.9)

ku(l, 2) = ~2 (~en
15
- ~ e12 )
30
1 1

- ljA
{EdAd~ + ~
2la
11 { 1
- ljA
z (1 - 3~)EdAd~, (4.10)

1
~ ( 125 e12 - EdAd~ - 2~a j_ 1 Lzz(l+3~)EdAde,
1

ku(l , 3) =
1
30
en) j_1L (4 .11)

ku(2 , 2) = -la
15
11 .
- 1
1

A
O"dAd~ + -la ( -2 en - -1
2 15 30
1 1-
e12)
2 1

- 1 A
EdAd~ (4 .12)

+ ( -2 en - -1 e12 )
15 30
11- 1
1

A
zz( l - 3~)EdAd~ - + 11 - 1l
2a
1

- 1 A
z 12 (1 - 3~) 2 EdAd~,
-

45
 kt1(3 , 3) = l
0
fl j O'dAd~ + -l
0
( 2 2
-81 - - 1 On )
2fl j _ EdAd~ (4.14)

I: i
15 - 1 A 2 15 30 -1 A
2
- ( 15 812 -
1
30
Bn) z1 (1 + 3~)EdAd~ + 2~0 [
1
1
iz[ (1 + 3() EdAd~,
2

and ku ( 2, 1) = ktt ( 1, 2) ; ku (3, 1) = ku ( 1, 3) ; ku ( 3, 2) = ku ( 2, 3) .

Eqs. (4.6)-( 4.14) combining with Eqs . (2.17) and (2.22) give the global internal force vector
and tangent stiffness matrix, and completely define the element formulations. If the elastic
behavior is assumed, O' = Er=:, the integrations in (4.6)-(4 .14) can be computed explicitly,
and Eqs.(4.6)-(4.14) are reduced to the following explicit forms

N = EA
I
(ii 0[1
l 0 + 15
_ BnB12 0[2)
30 + 15 ' (4.15)

u + -0[1 - -BnB12- + -Bf2 ) + -2 EI(Wa + 012 ),

Mn = lo EA(4Ba - 812) ( -l (4.16)
30 0 15 30 15 l0

M12 = lo EA ( -Bn + 4B12 ) ( -lii + -Bia - -Ba 812

- + -Bf2 ) + -2 EI(Bn + 2B1 2) (4.17)
30 0 15 30 15 l0

and
1
ku(l, 1) = lo EA , (4 .18)
1
kt1(l , 2) = EA(4Bn - 812) , (4.19)
30
1
ku(l, 3) = EA(-Bn + 4B12 ) , (4.20)
30

kt1(2 , 2)
lo
= 30 EA
[1
30
2
(4Bn - 812) + 4
(ii
lo+
0[1
15 On 812
- 3Q 0[2)]
+ 15
4
+~EI , (4.21)

1
ku(33) = -lo EA [ -(-Ba+4B12) 2 +4 ( -
ii
+0[-
1 On-
-- 812+ -0[2 ) ] +-EI
4 (4.22)
'· 30 30 · l0 15 30 15 l0 '

k (2 3)
tl '
=~EA
30
[(4011 - 812)(4012 -Ba) -
30
(ii Br1 - 811812
l 0 + 15
Bl2)]
30 + 15
'}_EI
+ l0 '
(4.23)

kt1(2 , 1) = kt1(l , 2) ; kt1(3 , 1) = kt1(l, 3) ; ku(3, 2) = kt1(2, 3).

It is noted herewith Eqs. (4.15)-( 4.23) are previously derived in [2] for the case of shallow
arch element .

5 Computational procedures
With involvement of the plastic deformation, the integrations in Eqs. (4.6)-(4.14) cannot
be computed analytically, and the numerical integration should be employed. The Gauss
.quadrature with four points along the length and seven points through the depth of the

46
element is adopted in the present work. Two Gauss-point quadrature along x 1 axis gives
exact integration for the elastic case (since the strain energy U = ~ fv Ec: 2dV
is a second
2
order polynomial in e), but four-point Gauss quadrature is employed in order to investi-
gate the plastic expansion along the beam length. The positions and weights for Gauss
quadrature are given in [12] .
The obtained element formulations are assembled into the internal force vector and
tangent stiffness matrix at the structural level in order to construct the equilibrium equa-
tions, which can be written in the forms [8, 9]

R(D , >.)=Fin - Fex =Fin - Afex = 0, (5.1)

where R is the out of balance force vector; Fin and F ex are the vectors of internal and
external forces, respectively; >. is the load parameter, and fex is the fixed normalized
external load vector. Eq. (5.1) can be solved by the incremental-iterative strategy based
on Newton-Raphson method. In order to obtain the complete equilibrium paths, Eq. (5 .1)
is supplemented by the constrain equation to form the arc-length method as discussed in
the authors' previous work [2] .
To compute integrals in Eqs. (4.6)-(4.14) , the values of stress and 'elasto-plastic' mod-
ulus at the Gauss point should be evaluated, and comparing to the elastic analysis, this
procedure is an additional part. Because the incremental nature of plastic equations,
Eqs. (3 .2) and (3.4), in order to compute the new stress from a strain increment, values
of the strain and stress at the Gauss points should be stored for each equilibrium point.
Furthermore, since the hardening effect is taken into consideration, values of the yield
stress are also required to store and update. To illustrate this computational process, a
computer subroutine written in MATLAB [13] , a Matlab function called eplaSE, is pro-
vided in the Appendix. The function shows how a strain increment is computed from
the current global nodal d.o.f, and then the stress is updated by using Eqs . (3 .1) , (3 .2) or
(3.4). The strain hardening is taken into consideration in the function by updating the
yield stress at Gauss points for each increment. The function is developed in close with
the discussed theory, but it may not be optimized in the computational context.

6 Numerical examples
6.1 Cantilever beam under tip load
The cantilever beam with geometry and material data shown in Fig. 3 was proposed by
Kondoh and Atluri in [14] . The original data were given in English units of measurement,
but for easier imagination, they have been convected to SI units herewith. The beam is
analyzed by eight equal elements, and the dependence of displacements at the free end on
the external load P is investigated.
Fig. 4 shows the computed equilibrium paths of the beam, which are in excellent with
those obtained in [14]. The expansion of plasticity in the beam is investigated by examining
the values of stress at the integration points for each increment , and the zones with plastic
points in the beam at various load levels are shown in Fig. 5. From Fig. 5, one can see
that the plastic zone in the tensile region is slightly larger than the corresponding zone in

47
t he compressive region. This result is physically reasonable, since t he neut ral axis of t he
beam is in tension , and the strain, defined by Eq. (4.3) , is a combination of t he posit ive
membrane st rain and t he bending strain. T he computed values of stress and strain at t he
t hree sampling points (Gauss points) in F ig. 5, given in Table. I , furt her confirm t his fact .

200

Q l 50 L +u
O!J
c
Q..

b = 0.254 cm cross-section -0
h = 1.270 cm Gh .£ 100
'O
E = 2. 1 x 10 6 kgf/cm 2 Lil 0)

:a
z,w b 0..
E t = 7 x 10 4 kgf/cm 2 <t: 50
cr0 = 2.l x J0 3 kgf/cm 2
L x,u

3 6 9 12
L = 12.7 cm Displacements at free end (cm)

Fig . 3. Cantilever beam under t ip load: geom- Fig. 4. Cantilever beam: applied load versus
etry and material data (ref. [14]) displacements at free end

I 2 I 3 B 4 I ~ I 6 _ 7 I ~ .§.
14.6522 cm 1 c --.. I I I ~~ I
- - 1 - - 1I- -
A - - - ! · sampling lpoints I element !number

..
1

,
•• •• •••• •1• •• •• ·1
• •1• • •• •!• · • b .
-
••• ---~1 ~
co press1on I I
~ • •1• • · ~1~ =::1 p= 24.3020 kgf
.-.-- • ·1· • •• .-.----- 1- I I
-!tension I • plasticl point
• •
• • •••• • ~ i · • ••1. • • ~i ·
• el• • • • •
•
• -;- ~ - -- I I
•••
1...--- I
1 I I p = 38.3439 kgf
.!:-----.-- I I

p = 43.3312 kgf

•• •• •• •••1• •• •• ••1•• •• •• •• • • •• • • •!• • ~1

•I• • • •I• • • •I• • I
L •- -•- •1•
•• •• •• · 1·
• • • •• ~i ·
•
••
• ..
~1 · · ~1 · • ·- ··· 11··---·-
•• ·~ 11 ·· • • ·~1, ·· .---.
• • • • • •• • •• • • •• • ~, ·
.- ·c:::::
·- .1..--
•• :1•••• .-·-J
• •
I I
!
p= 55.8928 kgf

Fig. 5. Cantilever beam: posit ion of sampling points and plastic zones

48
 - - --

Table 1. Cantilever beam: values of stress and st~ain at

three sampling points in Fig. 5

p OA EA crs EB ere EC
(kgf) (10 3 kgf/cm2) 10- 3 (10 3 kgf/cm2) 10- 3 (10 3 kgf/cm 2 ) 10- 3
24.3020 2.1425 1.4737 -2 .1416 -1.4612 0.0132 0.0063
32.1975 2.7480 10.0860 -2.7431 -10.0157 0.0741 0.0352
38.3439 3.3114 18.0993 -3.3032 -17.9819 0.1239 0.0587
43.3212 3.7144 23 .8312 -3.7031 -23.6709 0.1690 0.0801
55.8928 4.5429 35.6157 -4.5232 -35 .3345 0.2965 0.1406

6. 2 Asymmetric frame
The asymmetric frame in Fig. 6 was firstly studied by Cichon in [5], and then by Hsiao
and co-workers in [6]. The analysis is performed herewith by using ten equal elements,
five for each beam. The applied load versus displacements at the loading point is shown
in Fig. 7, where for the purpose of comparison, the curves computed by Cichon are also
displayed.

p u

24cm 96 cm z,w -g 0 '

~
L= 120cm .£ / I

L A = 6 crn2 ] -o. 5
0.
o o o Cichon [5] :
I ,,
,, '\
I
"
I
! =2 crn4 ~ - - - - present elastic' , __ , , ' - w \ ,,: u
E = 7.2 x 10 5 kgflcm2 -1
- present elasto-plastic
Et= 0.1 E
· cro = !Ox 10 3 kgflcm2 -1. 50 20 40 60 80 100 120
Dispcacements at loaded point (cm)

Fig. 6. Asymmetric frame under concentrate Fig. 7. Asymmetric frame: applied load ver-
load: geometry and material data (ref. [5]) sus displacements at loaded point
As seen from Fig. 7, the load-displacement curves obtained in the present study are
quite agrement with those reported by Cichon using the total Lagrangian approach, up
to the lateral displacement of above 60cm, at least . The work in [5], and in [6] either,
did not show the displacement beyond 60 cm. The large displacement behavior of the
frame obtained by the elasto-plastic analysis is clearly different from that obtained by
the elastic analysis . As seen from Fig. 7, by assuming the linearly elastic behavior , we
have considerably overestimated the limit load of the structure. From the numerical
investigation in this example, we can conclude that the effects of plasticity in some practical
cases is important and it should be taken into account in the large displacement analysis
of frame structures.
The authors have also employed some higher-order Gauss quadrature in analyzing this
example, but the obtained results show very little difference. It is noted that both Cichon

49
and Hsiao and co-workers employed Lobatto rule [9], four points along the element length
and five points through the depth, in computing their internal forces and tangent stiffness
matrix. The Lobatto rule has points on the beam surface which may sooner predict
the plasticity, but by increasing the number of integration through the beam depth, the
authors see almost no difference in the numerical results.

6.3 Elasto-plastic buckling of Euler beam

The elasto-plastic buckling of the simply supported beam subjected to axial forces , which
often called Euler beam in the literature, is investigated in this example. The beam with
length of 100 cm, and a quare cross-section of (5 cm x 5 cm) , is made of the same material
as the asymmetric frame in Fig. 6. To compute the secondary buckling paths, a small
transversal load, 1P with 'Y = 0.001 , is introduced as a perturbation load. A mesh of eight
equal developed elements is employed in the analysis.
Fig. 8 shows the equilibrium paths obtained by elastic and elasto-plastic analyses, where
7r2EJ
Pe = - ----V is the Euler load. As noted from the figure that the behavior of the beam
in the post-buckling region obtained by the elastic analysis is very different from that
obtained by the elasto-plastic analysis. For the case of elastic analysis, the computed path
shows the stable state in the post-buckling region, which is in agrement with the theory
of the 'elastica ' beam [15 , 16] . The path obtained by the elasto-plastic analysis, however
becomes unstable beyond point Po (see Fig. 8), where the material at the outer layers of
the beam makes a transition from elastic to plastic.
By computing the equilibrium paths with a range of tangent modulus Et of 0 to 0. 7E
as illustrated in Fig. 9, we can conclude that the position of the 'transition point ' Po in
the present example is not depended on the value of Et, but on cro - the initial yield stress .
The paths displayed in Fig. 9 also show that t he tangent modulus, that is the hardening
parameter H defined in Eq. (3.3) , affects the post-buckling strength, measured in term of
P / Pe . A higher hardening parameter is , the more post-buckling strength observers, and
this result is physically reasonable.

1.5 1.5
elastic Et=O.lE Et = 0.7 E
elasto- plastic Et = 0.5 E

I
Po

0.5

ii ti
L=lOO cm ~ Et = O
x,u

0 0.1 0.2 0.3 0 0.1 0.2 0.3

-w!L - w!L

Fig. 8. Euler beam: equilibrium paths ob- Fig . 9. Euler beam: equilibrium paths with
tained by elastic and elasto-plastic analyses various values of tangent modulus Et

50
 Fig. 10 shows t he plastic points and their covered zones in the beam corresponding to
the points P1, P2, and P3 in Fig. 8. In moving from point Pi to point P3 , the plastic zones
gradually expand towards t he beam neutral axis from the outer layers, and to the t wo
beam ends from the middle. In contrast to t he cantilever beam in Sub-section 6.1, the
plastic zone in compression in the present example is slightly larger than t he corresponding
zone in tension, and this fact is resulted from t he negative membrane strain defined by t he
first two parts of Eq. (4.3) . The computed values of stress and strain at the three sampling
points in Fig. 10 also further confirm this result.

element number

2 1 II 4 •1 B ~ 5 I :§ 7
..
--~
4Q 1?koI
~- - - 1
cm •I
- - _ .,.11
C ---- " I I
A ________::::!"" samplin points

---- . . .. . . .•!•. .• .• .. . . .------

•
,!,
com~ress10n -
· ~---..1 •
I I • • I •l.L--
I ~ II II p = 33.0598

t nsion
-
I
~·-
~I
.. . . .. . . .. -·--
- .- .1.-.-----L I
• plastic point
1
(10 3 kgf)

p = 20.9420
(103 kgf)

p = 16.6144
( J0 3 kgf)

Fig. 10. Euler beam: position of sampling points and plastic zones
It is necessary to note that , in moving from point Po to point P3 , we are not in the
unloading process as mentioned in Sec.3 and illustrated in Fig. 2.a, even the applied load
P (in absolute value) is in descent. The absolute values of both stress and strain, as given
in Table 2, are increased, and t he material is fl.owed plastically. That means we are still
in the loading process. The decrease in the external load may be explained by the lower
effective area of t he cross-section, which is resulted from expansion of the plastic zones as
suggested by Fung and Tong in [3] , in which t he plastic behavior of a bar under tensile
test has been investigated.

Table 2. Euler beam: values of stress and strain at

three sampling points in Fig. 10

-P O"A EA O"B EB O"C cc

(10 3 kgf) (103 kgf/ cm 2 ) 10 - 3 (10 3 kgf/ cm 2 ) 10- 3 (10
3
kgf/cm 2 ) 10- 3
33.0598 10.4561 20 .2246 -10 .8615 -25.8542 -2 .0267 -2 .8148
20 .9420 13.4701 62.0848 -14.0446 -70 .0633 -2. 8723 -3 .9893
16.6144 15.6498 92.3580 -16.0989 -98.5957 -2 .2456 -3.1189

51
7 Conclusions
A non-linear beam element for in-plane analysis of elasto-plastic frames under large dis-
placements has been formulated in this paper. The element , taking the axial displacements
into account , has been formulated by using the co-rotational approach and expression of
the virtual work. The Gauss quadrature has been employed in computing the local in-
ternal force vector and the tangent stiffness matrix. The structural material has been
assumed to be bilinear with isotropic hardening rule, and this law has been employed to
update the stress and yield stress at the Gauss points .
The investigation from the numerical examples has shown a good performance of the
developed element. It has also shown that, by lifting the linearly elastic assumption, the
behavior of the frames under large displacements has been clearly changed. In particular
cases, such as the structure in Sub-section 6.2 , the limit load might be considerably over-
estimated by ignoring the plastic deformation. In the instability analysis , as shown by the
example in Sub-section 6.3, with the involvement of plasticity, the nature of post-buckling
behavior of the structure might completely be different. The investigation of plastic ex-
pansion has contributed to the understanding the behavior of the elasto-plastic frames
under large displacements.
Acknowledgements . The authors would like to thank Prof. Ku Mo Hsiao of the
National Chiao Tung University, Taiwan for useful advice in the numerical algorithm.
The practical help by Mr. Duong Quang Viet, a Ph.D student at the Tokyo Institute of
Technology is gratefully acknowledged.
Appendix. This Appendix lists the MATLAB commands for computing values of the
new strain and stress, and updating the yield stress at Gauss points of the element. To
this end, a natural coordinate, ( = ~ zz, is introduced, so that with zz E [- ~, ~] ---+ ( E

[-1 , 1], where h is the beam thickness. Besides the geometry and material parameters ,
the following input data should be included: d-the vector of nodal displacements, defined
in Eq. (2.2); nGX, nGZ-the number of Gauss points along xz and zz axes, respectively;
epsO, sigO , sigYO-the vectors of strains, stresses and yield stresses at the Gauss points
of the element. The function calls other function, namely ptwt2 (nGX, nGZ) , in which the
positions and weights for Gauss quadrature are listed [12] . The explanations, including
the citations of developed formulations are provided for the sake of easily understanding.
[eps,sig,sigY] = eplaSE(xl,x2,zl,z2,b,h,E,Et,d,nGX,nGZ,epsO,sigO,sigYO);
u1=d(1,1); w1=d(2,1); r1=d(3,1);
u2=d(4,1) ; w2=d(5,1); r2=d(6,1);
Lx=x2-x1; Lz=z2-z1;
Lo=sqrt(Lx -2+Lz-2); % initial length
Ln=sqrt((Lx+u2-u1) -2+(Lz+w2-w1)-2); % current length
% below rigid rotation
s=((w2-w1)*(x2-x1)-(u2-u1)*(z2-z1))/(Lo*Ln); % sin(tr), eq.(2.6)
c=((x2-x1+u2-u1)*(x2-x1)+(z2-z1+w2-w1)*(z2-z1))/(Lo*Ln) ; % cos(tr)
if ((s>=O & c>=O) I (s<=O & c>=O))
tr=asin(s);
elseif (s>=O & c<=O)
tr=acos(c) ;

52
else
tr=-acos(c);
end
u=Ln-Lo; rl1=r1-tr; rl2=r2-tr ; % local d . o.f, eq . (2.3)
[point2,weight2]=ptwt2(nGX,nGZ); %Gauss posi.and wei .
nG=nGX*nGZ; %total Gauss points
eps=zeros(nG,1); sig=zeros(nG,1); sigY=zeros(nG,1); % initial values
epsM = u/Lo + (r11-2-0.5*rl1*rl2+rl2-2)/15; % mem.strain, eq . (4 .3 )
% below loop over Gauss points
for inX=1:nGX
xi=point2(inX,1);
for inZ=1: nGZ
ze=point2(inZ,2);
ipG = nGZ*(inX-1)+inZ; % current Gauss point
epsG=epsO(ipG,1); % old strain
sigG=sigO(ipG,1); % old stress
sigYG=sigYO(ipG,1); % old yield stress
kappa= h*ze*((1-3*xi)*rl1-(1+3*xi)*rl2)/Lo/2; % curvature, eq. (4.3)
epsN epsM+kappa; % new strain
deps epsN-epsG; % strain increment
dsig E*deps; % elastic stress inc.
sigE sigG + dsig; % elastic predictor
% below yielding check and elasto-plastic stress increment
fY = abs(sigG) - sigYG ; % yield function
i f (fY==O) %yielded before inc .
dsig = Et*deps ; % stress inc ., eq. (3 . 2)
elseif (fY<O & abs(sigE) > sigYG) % transition to plastic
m = (sigYG-abs(sigG))/(abs(dsig));
dsig m*deps*E + (1.-m)*deps*Et; % stress inc . , eq . (3 . 4)
else
dsig E*deps ; % elastic or unloading
end
eps(ipG,1) = epsN;
sig(ipG,1) = sigG + dsig ;
% below account for hardening
if abs(sig(ipG,1)) > sigYG;
sigY(ipG,1) abs(sig(ipG,1 ) );
else
sigY(ipG,1) sigYO(ipG,1);
end
end
end

References
1. Nguyen D . K ., A non-linear element for analyzing elastic frame at large deflections , Vietnam
Journal of Mechanics 22 (2000) 19-28 . .
2. Nguyen D. K. and Do Q. Q., Large deflections analysis of frames by elements contaning
higer-order terms, Vietnam Journal of Mechanics 25 (2003) 243-254.

53
 3. Fung Y. C. and Tong P. , Classical and Computational Solid Mechanics, World scientific,
Singapore, 2001.
4. Backlund J., Large deflection analysis of elasto-plastic beams and frames , International Jour-
nal of Mechanics Science 18 (1976) 269-277.
5. Cichon C., Large displacement in-plane analysis of elastic-plastic frames, Computer f3 Struc-
tures 19 (1984) 737-745.
6. Hsiao K. M., Huo F. Y. and Spiliopoulos K. V. , Large displacement analysis of elasto-plastic
frames, Computer f3 Structures 28 (1988) 627-633.
7. Belytschko T . and Glaum L. W ., Application of higher order co-rotational stretch theories to
nonlinear finite element analysis, Computer f3 Structures 10 (1979) 175-182.
8. Belytschko T ., Liu W. K. and Moran B. , Nonlinear Finite Elements for Continua and Struc-
tures, John Wiley & Sons, Chichester, 2000.
9. Crisfield M. A., Non-Linear Finite Element Analysis of Solids and Structures, Volume 1:
Essentials, John Wiley & Sons, Chichester, 1991.
10. Mendenson A., Plasticity: Theory and Application, The Macmillan Co., New York , 1968.
11. Luo Y. H. , Explanation and elimination of shear locking and membrane locking with field
consistence approach, Computer Methods in Applied Mechanics and Engineering 162 (1998),
249-269.
12. Mathews J. H. and Fink K. D., Numerical Methods Using Matlab , Third, Prentice Hall Inter-
national, New Jersey, 1999.
13. Matlab, High-Performance Numeric Computation and Visualization Software, Reference
guide, The Math Works Inc. , New York, 1992.
14. Kondoh K. and Atluri N., Large-deformation, elasto-plastic analysis of frames under non-
conserveative loading, using explicitely derived tangent stiffness based on assumed stresses,
Computational Mechanics 2 (1987) 1-25.
15. Shames I. H. and Dym C.L. , Energy and Finite Element Methods in Structural Mechanics,
McGraw-Hill, New York, 1985.
16. Timoshenko S. P. and Gere J . M., Theory of Elastic Stability, Second, McGraw-Hill , New
York, 1961.
Received August 8, 2003

' ' ">., "" , ,,, "

PHAN TU DAM CHO PHAN TICH CHUYEN VI LON
CUA KHUNG DAN-DEO .
Bai bao xay di,mg phan tit dam phi tuyen cho phan tfch khung dan-deo ch!u chuyen v! l&n.
Cong thli'c phan tit dm;rc thiet l~p nha phmmg phap h~ t<;>a d9 dong hanh va bieu thli'c cong ao.
Ma tr~n d9 cli'ng va vec-ta n9i h.rc phan tit dm;rc tinh nha phep cau phtrang Gauss. Ung suat t~i
cac diem cau phmmg dm;rc c~p nh~t tren ca s& gia thiet quan h~ li'ng suat-bien d~ng tu.in theo
quy lu~t luang tuyen tinh v&i lu~t tai ben dling hu&ng. Duang can bang ci'i.a ket cau duqc xay
di,mg tren ca s& thu~t toan d9 dai cung va phuang phap l~p Newton-Raphson. Doi xit ci'i.a phan
tit dm;rc danh gia qua cac vf dl,l. Ket qua so cho phep dua ra m9t so nh~n xet ve anh hu&ng ci'i.a
chay deo den doi xit ciia ket cau CO chuyen Vl lan va S\f lan truyen vung deo trong qua trinh chat
tai .
e-mail: ndkien@im01 .ac.vn

54