See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/335104899

Improved Displacement-Based Timoshenko Beam Element with Enhanced-

Strains

Article  in  Journal of Structural Engineering · March 2020

DOI: 10.1061/(ASCE)ST.1943-541X.0002549

CITATIONS READS
0 441

2 authors:

De-Cheng Feng Jian-Ying Wu

Southeast University (China) South China University of Technology
47 PUBLICATIONS   299 CITATIONS    54 PUBLICATIONS   981 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Special Issue on 'Data Mining in Civil Engineering', Advances in Civil Engineering View project

Strain localization in softening inelastic solids View project

All content following this page was uploaded by De-Cheng Feng on 24 December 2019.

The user has requested enhancement of the downloaded file.

1 An Improved Displacement-Based Timoshenko Beam Element with
2 Enhanced-Strains

3 De-Cheng Feng1 and Jian-Ying Wu2

4
1 Key Laboratory of Concrete and Prestressed Concrete Structures of the Ministry of Education,
5 Southeast University, Nanjing 210096, China, email: [email protected]
6
2 State Key Laboratory of Subtropical Building Science, South China University of Technology,

7 Guangzhou 510641, China, with corresponding author email: [email protected]

8 ABSTRACT
9 This paper presents an enhanced-strain formulation for conventional displacement-based (DB)
10 Timoshenko beam elements accounting for shear deformations. The proposed beam element is
11 an extension of the original enhanced-strain formulation developed for DB Euler-Bernoulli beams
12 and the inaccuracy due to weak satisfaction of equilibrium is remedied. A strain enhancement
13 is introduced to the axial elongation of the section in such a way that the axial equilibrium is
14 strictly fulfilled. The element formulation is established based on a locking-free finite element
15 approximation, and a modified fiber model is adopted for the section behavior. The numerical
16 implementation of the element needs only minor changes to the state determination of the original
17 DB element. Two typical reinforced concrete (RC) columns are analyzed to demonstrate the
18 capacity of the novel Timoshenko beam element with enhanced-strain.
19 Keywords: Timoshenko elements, Displacement-based beam elements, Enhanced strain, Equili-
20 brium condition, Flexure-shear interaction

21 INTRODUCTION
22 Beam or frame elements may be the most widely used structural elements in practical structural
23 engineering owing to its numerical accuracy and computational efficiency. In accordance with the
24 interpolation formulations, beam elements can be categorized into three groups, i.e., displacement-
25 based (DB) (Hellesland and Scordelis, 1981; Mari, 1984), force-based (FB) (Taucer et al., 1991;
26 Spacone et al., 1996; Scott and Fenves, 2006; Feng et al., 2016) and mixed ones (Limkatanyu and
27 Spacone, 2002; Taylor et al., 2003; Alemdar and White, 2005). In the DB element the displacement
28 field of the element is interpolated, and the the principle of virtual displacements is adopted to
29 derive the element formulation. Despite its simplicity, the weak satisfaction of equilibrium does
30 not guarantee accurate solution for the local responses (e.g., section axial force), particularly when
31 coarse meshes are adopted in simulations. Consequently, refined meshes are usually required for
32 DB elements (Calabrese et al., 2010). Comparatively, the FB element is formulated based on the
33 force distribution of the element and the principle of virtual forces. As the equilibrium condition
34 is strictly satisfied, coarse meshes can be adopted. Being a combination of the above two groups
35 of elements, the mixed element is established based on the Hellinger-Reissner or the Hu-Washizu
36 principle. As the force field is assumed, the mixed element is also of high coarse mesh resolution
37 as the FB one.

1 D.C. Feng, August 12, 2019

38 Among the aforesaid three types of beams elements, the DB element is the easiest for numerical
39 implementation, and has been provided in most commercial finite element analysis packages.
40 However, though the axial force of the element is correct, the axial forces at the integration points
41 are not since there is no equilibrium condition at the section level (Koutromanos and Bowers, 2016).
42 To resolve this issue, Koutromanos and Bowers (2016) and Tarquini et al. (2017) proposed enhanced-
43 strain formulations in which the equilibrium along the axial direction is fulfilled. Unfortunately,
44 both formulations apply only to Euler-Bernoulli beams. That is, they can only be used for analysis
45 of slender beam/column members.
46 This paper aims at extending the enhanced-strain formulation for Euler-Bernoulli beams to
47 Timoshenko ones such that members with significant shear deformations, e.g., short columns,
48 coupling beams, etc., can be dealt with. The objectives of the present work are four-fold: (1) to
49 introduce enhanced strains at the section level to ensure the axial equilibrium (note here the
50 distributed axial load is neglected); (2) to overcome shear-locking typically for DB Timoshenko
51 elements using a locking-free finite element approximation scheme; (3) to represent the axial-
52 flexure-shear interaction at the material level by a modified fiber section model; (4) to avoid mesh
53 size dependence by regularizing the softening responses of the concrete behavior based on fracture
54 energy. Two representative examples are presented to demonstrate the performance of the proposed
55 Timoshenko element.

56 BASIC FORMULATION
57 In this section, the basic formulation of the proposed Timoshenko beam element with axial
58 strain enhancement is introduced.
59 Beam kinematics
60 The two-dimensional (2D) Timoshenko beam theory with linear geometric transformation is
61 considered, though the resulting beam element can be easily extended to general three-dimensional
62 (3D) cases. As shown in Fig. 1, the element has two nodes and each node has three degrees-of-
63 freedoms (DOFs), resulting in the following element displacement vector u (x) at position x
n oT
64 u (x) = u (x) , w (x) , θ (x) (1)

65 where x is the longitudinal axis of the beam; u (x) and w (x) are the axial and lateral translations,
66 respectively, with θ (x) being the rotation.
67 The corresponding section deformation and force vectors  (x) and σ (x) are then expressed as
n oT
 (x) = ε (x) , γ (x) , κ (x)
68 n oT (2)
σ (x) = N (x) , V (x) , M (x)

69 where ε (x) and N (x) are the axial elongation and force; γ (x) and V (x) are the shear deformation
70 and force; κ (x) and M (x) are the bending curvature and moment.

2 D.C. Feng, August 12, 2019

 w2

u2 ww
Q C F
T2 wx

*B
m J T
q
w1 n w( x )
Y
u1 y Middle axis
T1 u( x )

X T ( x)
x
z

Fig. 1. Beam kinematics

71 In accordance with the Timoshenko beam theory, the following compatibility relation holds


 ∂u  

∂ x
 
ε
 
(x)
 
 ∂w

 
 
 

 (x) = γ (x) = − θ = L u (x)

  
 

72 (3)
 κ (x) 
  ∂x
  

  ∂θ
  


 

 ∂x 

 


73 for the Laplacian operator L (·).

74 Strain enhancement
75 Similarly to Koutromanos and Bowers (2016) and Tarquini et al. (2017), an enhancement
76 is introduced to the axial elongation of the section to improve the inaccuracy due to the weak
77 satisfaction of equilibrium. That is, the enhanced section strain vector ¯ (x) is expressed as

78 ¯ (x) =  (x) + ˆ (x) (4)

79 where  (x) = L u (x) is the standard part given in Eq. (3); the enhanced strain ˆ (x) is given by
n oT
80 ˆ(x) = e (1 − 2x/L) , 0, 0 (5)

81 for the element length L and a constant parameter e that needs to be determined at each step. Note
82 that the enhanced strain should satisfy two following requirements
∫ ∫
83 ε̄ (x) dx = u x, ˆT σ dx = 0 (6)
L L

3 D.C. Feng, August 12, 2019

84 where ε̄ (x) is the enhanced section axial elongation while u x is the axial elongation of the entire
85 element.
86 Eq. (6)1 states that the axial compatibility still holds in an integral form, and Eq. (6)2 indicates
87 that the enhancement is orthogonal to the stress vector, such that the standard principle of virtual
88 work is expressed as
∫ ∫ ∫  
89 δ¯ σ dx =
T
δ σ dx =
T
δuT b dx + δuT t |ΓB (7)
L L L

90 where δ (·) denotes the admissible variations of related variables; b = {n, q, m}T is the distributed
91 force vector; t = {F, Q, C}T is the concentrated nodal force vector; ΓB is the boundary of the
92 applied concentrated nodal force, as shown in Fig. 1.
93 In fact, Eq. (6)1 is satisfied automatically since the enhancement term, i.e., e (1 − 2x/L), is
94 an odd function and thus the corresponding integral vanishes. On the other hand, for the second
95 requirement, substituting Eqs. (2) - (4) into Eq. (6)2 yields
∫
96 (1 − 2x/L) N (x) dx = 0 (8)
L

97 which should be enforced during the state determination of the whole element.
98 Moreover, the axial force is approximated by the following interpolation scheme

99 N (x) = (1 − x/L) · N | x=0 + x/L · N | x=L (9)

100 Upon substitution of Eq. (9) into Eq. (8), the second constraint (6)2 becomes
∫ ∫
ˆ σ dx =
T
(1 − 2x/L) N (x) dx
101
L L (10)
1 
= N | x=0 − N | x=L = 0
6
102 Eq. (10) indicates that the strain enhancement leads to identical axial forces for the two end
103 sections. At the same time, recalling the force interpolation function Eq. (9), axial force of an
104 arbitrary section at position x becomes
N (x) = (1 − x/L) · N | x=0 + x/L · N | x=L
105 (11)
= N | x=0 = N | x=L
106 That is, the axial equilibrium is not only satisfied in an average sense but holds for anywhere of the
107 whole element. Meanwhile, Eq. (11) also allows us to calculate the constant e in the enhancement,
108 which will be addressed in the next section.
109 It should be noted that in the above derivation, the distributed axial force is neglected, i.e.,
110 n (x) = 0. This assumption may lead to some errors since the exact solution ∫ of N | x=0 − N | x=L
111 is non longer zero but equal to the integral of the axial distributed load L n (x) dx. However,
112 this inaccuracy is not a major problem since the axial distributed load is generally small for most
113 situations in structural engineering (Koutromanos and Bowers, 2016). For instance, the distributed
114 axial load only contributes to less than 1% of the total axial force for a RC column with an axial ratio
115 of 0.38. Accordingly, the axial distributed load is not accounted for as in other studies (Tarquini
116 et al., 2017).

4 D.C. Feng, August 12, 2019

117 Finite element approximation
118 Based on the classical DB approach, the displacement field of the element is approximated by
119 certain shape functions, leading to the following standard displacement and deformation vectors
120 u (x) = N (x) d,  (x) = B (x) d (12)
121 where d is the nodal displacement vector; N (x) is the adopted shape function matrix and B (x) =
122 L N (x) is the compatibility matrix.
123 Generally, the standard piece-wise interpolation functions can be used. Nevertheless, such an
124 approximation leads to different orders of terms ∂w/∂ x and θ presented in the shear deformation
125 γ = ∂w/∂ x − θ. This incompatibility usually causes the well-known shear-locking issue for
126 Timoshenko elements in the thin beam limit (Crisfield et al., 2012; Ehrlich and Armero, 2005;
127 Wu, 2013). An efficient strategy to overcome the shear-locking is to introduce a bubble degree-of-
128 freedom θ 3 at the center of the element and adopt the mixed interpolation between the transverse
129 displacement and rotation fields. In this case, the shape function and compatibility matrices
130 associated with the nodal displacement vector d = {u1, w1, θ 1, u2, w2, θ 2, θ 3 }T are given by (Wu,
131 2013; Li et al., 2016; Feng et al., 2017)
1 − ξ 0 0 ξ 0 0 0 
N (x) =  0 1−ξ (1 − ξ) 0 ξ − 2x (1 − ξ) (2ξ − 1) (1 − ξ)
 x 2x

132
2 3 (13)
 0
 0 1−ξ 0 0 ξ 4ξ (1 − ξ) 

−1 0 0 1 0 0 0 
1
B (x) =
 
133
 0 −1 −L/2 0 1 L/2 −2L/3  (14)
L 
0 0
 −1 0 0 1 4 (1 − 2ξ)
134 for the normalized coordinate ξ = x/L.
135 Afterwards, substitution of (12) into the principle of virtual work (7) yields the element internal
136 force F int , external force F ext and stiffness K ele
∂F int
∫ ∫
137 F =
int
B σ dx, F =
T ext
N T b dx + t |ΓB , K ele = (15)
L L ∂d
138 By recalling the enhanced-strain (4), the element stiffness can be further derived as
T ∂σ ∂ ¯ ∂ ¯
∫ ∫
K =
ele
B dx = BT DT dx
∂ ¯ ∂ d ∂d
∫L L
∂ ∂ ˆ

139 = BT DT + dx (16)
∂d ∂d
∫L
= BT DT (B + aBed ) dx
L
n oT
140 where DT = ∂σ/∂ ¯ is the tangent stiffness of the beam section; a = ∂ /∂e
ˆ = 1 − 2x/L, 0 , 0 is
141 the enhanced strain vector. The constant parameter e and the additional matrix Bed are determined
142 from the axial force equilibrium Eq. (11), say,

 N (x) = DT,1 ¯ (x) = DT,1 (Bd + )



 ˆ
143 (17)
 N | x=0 = N | x=L



5 D.C. Feng, August 12, 2019

144 where DT,1 is the first row of the section stiffness DT . The expansion of the above relations gives
DT,1 B| x=L − DT,1 B| x=0 ∂e DT,1 B| x=L − DT,1 B| x=0
145 e= d, Bed = = (18)
DT,11 | x=L + DT,11 | x=0 ∂d DT,11 | x=L + DT,11 | x=0
146 with DT,11 being the first component of DT,1 .
147 Modified integration rule
148 The solution to the integrals in Eqs. (15)-(16) can be evaluated by some numerical quadrature
149 schemes, e.g., Gauss-Legendre quadrature or Gauss-Lobatto quadrature. Meanwhile, one needs to
150 check the axial equilibrium condition (17) between the generalized section stresses at the two end
151 points of the element. However, the end points are not necessarily coincident with the integration
152 points. On the on hand, if the Gauss-Legendre quadrature with no integration points at the element
153 ends is adopted, we need call for the sectional constitutive relations for the two end sections and
154 for the integration points to determine the constant parameter e using Eq. (18) and to calculate the
155 integrals in Eqs. (15)-(16), respectively. On the other hand, if the Gauss-Lobatto quadrature with
156 the integration points placed at the element ends is used, one still needs to calculate the integrals in
157 Eqs. (15)-(16)from the sectional constitutive relations for the interior sections. In order to improve
158 the numerical efficiency, a modified integration rule (Koutromanos and Bowers, 2016) is used
159 herein. That is, the integrals in Eqs. (15)-(16) are approximated by a linear interpolation of the
160 generalized stresses at the two end points, i.e.,
∫ h x   x  i
F =
int
BT 1 − σ x=0 + σ  x=L dx
L L
161 ∫L (19)
n x  h i xh i o
K =
ele
B T
1− DT (B + aBed )  + DT (B + aBed ) 
 
dx
L L x=0 L x=L

162 With the above approximation, only the constitutive relations for the end sections are needed.
163 In this case, Koutromanos and Bowers (2016) showed that the Gauss-Legendre quadrature and the
164 Gauss-Lobatto quadrature yield almost identical results. Therefore, the two-point Gauss-Legendre
165 quadrature is sufficient to evaluate the integrals (Tarquini et al., 2017). In fact, for most geometrically
166 linear FB elements with no distributed loads, the above force interpolation is exact.

167 FIBER SECTION AND MATERIAL MODELS

168 In order to account for axial-flexure-shear interactions, the beam section behavior is described by
169 a modified fiber model. To this end, the distribution of shear deformations has to be known. Strictly
170 speaking, an iterative procedure is required (Kagermanov and Ceresa, 2017) as the distribution of
171 shear strain across the section is non-uniform and sometimes rather complex. In practice, a
172 predefined distribution of the shear strain is usually assumed in order to enhance the overall
173 computational performance of structural analyses. Either uniform distribution and parabolic one
174 has been considered in the literature (Ceresa et al., 2009; Mullapudi and Ayoub, 2013; Li et al.,
175 2016). It is shown that both distributions yield rather close global load-displacement responses
176 (Vecchio and Collins, 1988; Petrangeli et al., 1999). Therefore, for the sake of simplicity, in
177 this work shear deformations are assumed to be uniformly distributed along the section depth.
178 Moreover, shear forces are assumed to be resisted only by concrete. Note again that the above
179 simplification should not be regarded as the limitation of the proposed method, but rather, it aims
180 only at improving the computational efficiency.

6 D.C. Feng, August 12, 2019

 Steel fiber yy
, yy

xy
, xy

xx
, xx
Concrete fiber
Stress state

Fig. 2. Modified fiber section model

181 Upon the above assumption, the concrete fibers are subjected to multi-axial stresses while the
182 steel fibers to uniaxial one; see Fig. 2. In this work, the multi-dimensional softened damage-
183 plasticity model (Feng et al., 2018a,b) is used for concrete fibers and the uniaxial Menegotto-Pinto
184 (M-P) model (Filippou et al., 1983) is used for steel fibers. With the above considerations, the
185 flexure-shear interaction can be described at the material level.
186 Moreover, a regularization strategy is needed to suppress the localization issue arisen from
187 strain-softening problems. Either the nonlocal approaches (Feng et al., 2016), the strong discon-
188 tinuity approaches (Wu, 2011) or the phase-field models (Wu, 2018) can be considered. In this
189 work, the crack band theory is adopted to determine the parameters involved in the softening regime
190 of the concrete model based on a constant fracture energy (Wu et al., 2006; Feng and Ren, 2017).
191 As only two strain components of concrete fibers, i.e., the normal one  x x and shear one  xy ,
192 are available at the section level, an iterative process is required to compute the lateral component
193  yy by imposing the equilibrium σyy = 0. Moreover, the static condensation is also needed as in
194 Feng et al. (2017) and Feng and Xu (2018).

195 IMPLEMENTATION
196 The proposed beam element can be numerically implemented by making minor modifications
197 to the standard DB elements. The only difference is that we need to solve iteratively the constant
198 parameter e according to the axial force equilibrium Eq. (17). Let us consider a typical time step
199 tn , with the given element nodal displacement vector denoted by dn . The iteration starts at j = 0
200 and is summarized as follows:

201 1. Assuming an initial guess to e j , which can be the vaule of the previous step, the enhanced
202 section strain is calculated as
j j T
¯n (x) = B (x) dn + ˆn (x) ˆn (x) = e j (1 − 2x/L) , 0, 0

203 with (20)

204 2. Calling for the fiber model, the section stress vector σn | x=0, L and stiffness vector DT n | x=0, L
205 at the two end points are computed from the enhanced strain vector ¯n | x=0, L .
3. Check whether  Nn | x=0 −Nn | x=L  < T ol, where T ol is a convergence tolerance. The influence
 

of it specific value on accuracy and efficiency will be discussed in the validation part. If
the equilibrium is satisfied, go to next step 4; otherwise, go back to step 2 with the updated

7 D.C. Feng, August 12, 2019

 section enhanced strain
 
Nn  − Nn 
 
x=0 x=L
e j+1 =e −
j   (21)
DT n,11  + DT n,11 
 
x=0 x=L

4. Update the element internal force Fnint and stiffness Knele according to Eq. (19) using σn  x=0, L

206

207 and DT n  x=0, L .

208 VALIDATION
209 The proposed element is implemented into the finite element analysis package ABAQUS through
210 the user subroutine UEL (Hibbitt et al., 2001). Three numerical examples are presented herein to
211 validate the proposed element. First an elastic cantilever beam is analyzed to show the locking-free
212 performance and the convergence rate of the element. Afterwards, two typical RC columns, i.e., a
213 flexure-dominated one and a shear-dominated one, are used as realistic examples to demonstrate the
214 superiority of the element over standard DB elements. In the analysis, 24 and 12 fibers are used to
215 discretize the core concrete and the cover concrete, respectively. Reinforcement fibers are assigned
216 according to the details of steel rebars. The confinement effect due to stirrups is considered through
217 the Mander model (Mander et al., 1988).
218 Elastic cantilever beam
219 The first example is an slender elastic cantilever beam under monotonically increasing tip
220 displacement, just as shown in Fig. 3. The beam length is 2000 mm, and section dimensions are
221 400 mm × 400 mm. The elastic modulus E=200000 MPa and Poisson’s ratio v=0.2. Evidently,
222 the beam is a slender type beam since its aspect ratio equals 5 so it can be used to demonstrate the
223 locking-free performance of the proposed element.
224 Fig. 4 (’ele’ in the figure represents ’element’) shows the theoretical and numerical results of the
225 tip displacement at the right end versus the reaction force at the left end, where Fig. 4(a) is obtained
226 by the proposed element and Fig. 4(b) is obtained by classical DB element. From the figures we can
227 found that the both the proposed element and classical DB element have no locking behavior since
228 both the results approaches the exact theoretical results very well. This is just because we adopted
229 a mixed interpolation strategy in Eq. (13) to avoid shear-locking. Furthermore, for the proposed
230 element, only one element can capture the accuracy result since the enhanced strain method can
231 overcome the weak satisfactory of the axial equilibrium. For the classical DB element, one need
232 to refine the meshes to approximate the accurate result. As seen in Fig. 4(b), refining the element
233 mesh will get converged results. Using one or two DB elements will significantly over-estimate the
234 structural response while using up to more than four DB elements can get nearly the same results
235 as the theoretical one.
236 Flexure-dominated RC column
237 The second example is a flexure-dominated RC column under cyclic loading, which was tested
238 by Saatcioglu and Grira (1999). The column height is 1645 mm, the section dimensions are 350 mm
239 × 350 mm, with a concrete cover of thickness 29 mm. Twelve deformed (ribbed) bars with diameter
240 19.5 mm are used as the longitudinal bars, and the diameter of the transverse bars is 6.6 mm with a
241 spacing 76 mm, respectively. The material parameters for concrete are: elastic modulus Ec =32500
242 MPa; compressive strength fc =34 MPa; compressive peak strain c =0.002; tensile strength ft =3.4

8 D.C. Feng, August 12, 2019

 V

400 mm
E, ν

L=2000 mm 400 mm

Fig. 3. Elastic cantilever beam

2 0 2 0
T h e o re tic a l s o lu tio n T h e o r e tic a l s o lu tio n
1 6 E n h a n c e d 1 e le 1 6 D B 1 e le
E n h a n c e d 2 e le D B 2 e le
D B 4 e le
D B 8 e le
1 2 1 2
k N )

k N )
3

3
L a t e r a l l o a d ( ×1 0

L a t e r a l l o a d ( ×1 0
8 8

4 4

0 0
0 2 0 4 0 6 0 8 0 1 0 0 1 2 0 0 2 0 4 0 6 0 8 0 1 0 0 1 2 0
T ip D is p la c e m e n t (m m ) T ip D is p la c e m e n t (m m )

(a) Enhanced element (b) Classical DB element

Fig. 4. Lateral load versus top displacement curves of the elastic beam

243 MPa ; tensile peak strain t =0.0001. The material properties for rebars are: elastic modulus
244 Es =200000 MPa; yielding strength fy =455.6 MPa; hardening ratio is α=0.01; default values are
245 adopted for the empirical parameters of the M-P model, i.e., R0 = 20, cR1 = 0.925, cR2 = 0.15,
246 a1 = a3 = 0.0, a2 = a4 = 1.0.
247 The column is firstly subjected to an axial load of 961 kN through force control, and then to a
248 cyclic lateral load through the displacement control. The comparison of numerical and experimental
249 results between shown in Fig. 5. As can be seen, for both the enhanced DB element and the standard
250 one, the results converges as the element number increases. Remarkably, only one enhanced DB
251 element gives satisfactory response which match the experimental result very well. Comparatively,
252 for the classical DB element the response will be significantly over-estimated, unless at least four
253 classical DB elements are used. It should also be noted that the two kinds of elements (i.e., the
254 enhanced one and the classical one) converges to the final result in different manners. That is, as
255 the number of elements increases, the results of the standard DB element become flexible while
256 the results become stiffer for the enhanced formulation. The latter is rational since the proposed
257 DB element with strain enhancement can be interpreted as a special mixed displacement-force
258 approach (Koutromanos and Bowers, 2016) which usually predicts stiffer result.
259 To give a more comprehensive description of the numerical results by different models, the
260 quantities of the predicted cyclic behaviors are given in Table 1. The yield load and displacement

9 D.C. Feng, August 12, 2019

 3 0 0 3 0 0

2 0 0 2 0 0

L a te ra l lo a d (k N ) 1 0 0 1 0 0

L a te ra l lo a d (k N )
0 0

-1 0 0 -1 0 0 T e s t d a ta
T e s t d a ta D B 1 e le
E n h a n c e d 1 e le D B 2 e le
-2 0 0 E n h a n c e d 2 e le -2 0 0 D B 4 e le
E n h a n c e d 8 e le
D B 8 e le
-3 0 0 -3 0 0
-1 5 0 -1 0 0 -5 0 0 5 0 1 0 0 1 5 0 -1 5 0 -1 0 0 -5 0 0 5 0 1 0 0 1 5 0
T o p d is p la c e m e n t (m m ) T o p d is p la c e m e n t (m m )

(a) Enhanced element (b) Classical DB element

Fig. 5. Lateral load versus top displacement curves of the flexure-dominated column

TABLE 1. Quantities of the cyclic behaviors by different models for flexure-dominated column

Enhanced Classical DB
Quantity Test
1 ele 2 ele 1 ele 4 ele
Yield load (kN) 159.86 176.88 183.24 242.35 181.90
Yield displacement (mm) 21.03 17.89 18.60 21.02 18.57
Peak load (kN) 192.56 199.56 207.08 292.63 216.41
Peak displacement (mm) 34.00 31.82 31.82 49.72 31.88
Dissipated energy (J) 115643.95 89343.02 109794.24 103463.47 107520.77

261 are determined through the equivalent energy method (Park, 1988). It can be found that the errors
262 for the results by one and two enhanced elements at yield point are around 10-15%, while they are
263 around 5-8% at peak point. However, the errors of the results by one classical DB element are up to
264 50% at the yield and peak point. It will decrease to similar level of enhanced element when using
265 four classical DB element. Meanwhile, for the dissipated hysteretic energy, two enhanced elements
266 and four classical DB elements have more accurate results compared with the experiment, whose
267 errors are 5% and 7%, respectively.
268 The above phenomena can be explained by Fig. 6, which shows evolution of the section axial
269 force given by both the proposed beam element and the classical DB one. It can be seen from
270 Fig. 6(a) that the axial forces of the base section exhibit great fluctuations for the DB element,
271 even refined meshes are used. Comparatively, the axial force given by the enhanced element is
272 exactly the specified value (961 kN). Fig. 6(b) gives the axial forces of both the upper section and
273 the bottom one of column. When using one DB element, though the average axis force of the two
274 sections is equal to the specific value, their values exhibit severe oscillations since the equilibrium
275 condition is satisfied only weakly in an integral sense.
276 Fig. 7 further displays the distributions of the axial force in the column given by one enhanced
277 element and four classical DB elements at different times, respectively. It can be observed that the
278 section axial forces of the enhanced element are correct as expected, but those given by the classical

10 D.C. Feng, August 12, 2019

 0 0
D B 1 e le U p p e r s e c tio n
D B 2 e le B o tto m s e c tio n
-4 0 0 D B 4 e le -4 0 0 A v e ra g e
D B 8 e le
A x ia l lo a d (k N ) E n h a n c e d

A x ia l lo a d (k N )
-8 0 0 -8 0 0

9 6 1 k N 9 6 1 k N
-1 2 0 0 -1 2 0 0

-1 6 0 0 -1 6 0 0
0 1 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0 0 1 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0
S te p S te p

(a) Axial forces of the base section (b) Axial forces of the DB element

Fig. 6. Evolution of axial forces in the flexure-dominated column

279 DB element vary significantly due to the axial equilibrium in an average sense. As the axial force
280 has a great impact on the moment-curvature (or moment-rotation), this inaccuracy inevitably leads
281 to inaccurate sectional responses.
282 Fig. 8 compares the experimental and numerical moment-rotation relations of the base section.
283 Evidently, the results from one or two enhanced elements both match the test data, while that
284 from one DB element significantly overestimates the test result due to the inaccurate axial force.
285 This fact justifies the results presented in Fig. 5, since the global force-displacement response is
286 determined from the local section force-deformation results. Note that maximum rotation is not
287 reached in the numerical simulation. The reason may be that in the actual experiment multiple
288 (three) cycles were performed at each level of the displacement while in the numerical simulation
289 only the first cycle was applied for simplicity. The degradations caused by multiple loading cycles
290 at each displacement level definitely affect the damage and plasticity evolutions, which cannot be
291 considered in the simulation, such that the numerical results under-estimated the maximum rotation
292 of the column.
293 The computational cost of the proposed element is also studied. Here one enhanced element and
294 four classical DB elements are compared, since they have similar numerical accuracy. In order to
295 complete the simulation, the Newton-Raphson iterations are 4864 and 4258 for enhanced element
296 and classical elements, respectively. On a laptop with a CPU of i7-2.40 GHz and RAM of 16 GB,
297 the simulation time is 117 seconds for enhanced element and 88 seconds for classical element.
298 The computational cost increases around 20% due to the internal iteration process presented in the
299 enhanced element, but the overall efficiency is acceptable.
300 Fig. 9 also gives the internal iterations of the enhanced element with respect to different axial
301 force tolerance. It is found that the iterative process in calculation of the enhancement parameter e
302 does not bring about too much computational burden. If the tolerance is set as 1 kN (around 0.1%
303 of the applied force), though the corresponding axial force of the based section exhibits fluctuations
304 within the range ±1 kN, generally 2-3 iterations per step is sufficient to find the converged results.
305 If a smaller tolerance, i.e., 10−3 kN, is used, the iteration number per step increases to 8-10, but the
306 computed axial force is exact as the specified value.

11 D.C. Feng, August 12, 2019

 1 .0 1 .0
a t 2 5 m m
a t 5 0 m m A v e ra g e
a t 1 0 0 m m
0 .8 0 .8 9 6 1 k N

N o rm a liz e d c o lu m n le n g th
0 .6 0 .6

0 .4 0 .4

0 .2 0 .2

0 .0 0 .0
9 3 0 9 4 0 9 5 0 9 6 0 9 7 0 8 0 0 1 0 0 0 1 2 0 0 1 4 0 0
A x ia l fo rc e (k N ) A x ia l fo rc e (k N )
b y e n h a n c e d 1 e le b y c la s s ic a l 4 e le

Fig. 7. Axial force distribution at different times

500 500
400 400
300 300
Bending moment (kN⋅m)

Bending moment (kN⋅m)

200 200
100 100
0 0
-100 -100
-200 -200
Test data Test data
-300 -300 DB 1 ele
Enhanced 1 ele
-400 Enhanced 2 ele -400 DB 4 ele

-500 -500
-0.10 -0.05 0.00 0.05 0.10 -0.10 -0.05 0.00 0.05 0.10
Total rotation (rad) Total rotation (rad)

(a) Enhanced element (b) Classical DB element

Fig. 8. Moment-rotation relation of the flexure-dominated column

307 Shear-dominated RC column

308 The final example is a RC column with shear failure tested by Arakawa et al. (1989). The
309 column height is 225 mm, while the section depth and width are both 180 mm, with a concrete
310 cover of thickness 10 mm. The steel rebars consist of eight longitudinal bars of diameter 12.7
311 mm, and the transverse bars of diameter 4 mm and spacing 64.3 mm. The concrete properties are:
312 elastic modulus Ec = 32500 MPa; compressive strength fc = 33 MPa; compressive peak strain
313 c = 0.002; tensile strength ft = 3.3 MPa; tensile peak strain t = 0.0001. The steel properties are:
314 elastic modulus Es = 200000 MPa; yielding strength fy = 340 MPa; hardening ratio is α = 0.01;
315 the default values as before are set to the empirical parameters in the M-P model. The applied axial
316 load is 476 kN.
317 The simulation results are shown in Fig 10. Similarly to the first example, the proposed enhanced
318 Timoshenko element can capture the test results fairly well with a small number of elements, as
319 shown in Fig. 10(a), while the results given from the classical DB element only converge to the test

12 D.C. Feng, August 12, 2019

 2 0 9 8 0
T o l. 1 0 -3
k N
T o l. 1 0 -3
k N
1 6 T o l. 1 0 0
k N
9 7 0 T o l. 1 0 0
k N

E n h a n c e m e n t ite ra tio n s
1 2 9 6 0

A x ia l fo rc e (k N )
8 9 5 0

4 9 4 0

0 9 3 0
0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0 2 5 0 0 3 0 0 0 3 5 0 0 4 0 0 0 0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0 2 5 0 0 3 0 0 0 3 5 0 0 4 0 0 0
S te p S te p

(a) Iteration numbers (b) Axial force for base section

Fig. 9. Computational cost of the enhanced element for flexure-dominated column

TABLE 2. Quantities of the cyclic behaviors by different models for shear-dominated column

Enhanced Classical DB Euler-Bernoulli

Quantity Test
1 ele 2 ele 1 ele 4 ele 2 ele
Yield load (kN) 111.28 112.25 115.48 133.07 120.65 161.00
Yield displacement (mm) 0.94 0.61 0.57 0.55 0.56 0.50
Peak load (kN) 111.28 123.39 123.77 139.89 128.08 161.00
Peak displacement (mm) 0.94 0.95 0.95 0.95 0.95 0.50
Dissipated energy (J) 610.53 710.56 694.57 521.01 760.65 2293.99

320 data with refined mesh as shown in Fig. 10(b). Comparatively, as the section shear deformations
321 are not accounted for, the results given from Euler-Bernoulli elements largely over-estimate the
322 load capacity, loading/unloading stiffnesses and the energy dissipation; see Fig. 10(c). In this
323 shear-dominant case (the span-to-depth ratio of the column is only 1.25), Euler-Bernoulli elements
324 should be avoided.
325 The predicted quantities during the test from different models are summarized in Table 2. Note
326 that the experimental yield point coincides with the peak point. As can be seen, the proposed
327 enhanced Timoshenko element is far more superior to both the standard DB element and the
328 Euler-Bernoulli one.
329 The local responses of the column are also studied. Fig. 11 presents the evolution of the axial
330 force at the base section. As can be seen, the axial force by the enhanced Timoshenko element
331 is accurate and the moment-curvature responses given by one or two elements are very close.
332 Comparatively, though the accuracy is improved by increasing the element numbers, the axial
333 equilibrium condition of classical DB elements is still not strictly satisfied. As shown in Fig.12,
334 this deficiency inevitably affects the numerical results of the section moment-curvature responses,
335 evidenced by the distinct results given by one or four classical DB elements.
336 The stress profiles of the base section at the top displacement 2.75 mm and -2.82 mm are plotted

13 D.C. Feng, August 12, 2019

 2 0 0 2 0 0

1 0 0 1 0 0
L a te ra l lo a d (k N )

L a te ra l lo a d (k N )
0 0
T e s t d a ta
T e s t d a ta
D B 1 e le
-1 0 0 E n h a n c e d 1 e le -1 0 0 D B 2 e le
E n h a n c e d 2 e le D B 4 e le
E n h a n c e d 8 e le D B 8 e le

-2 0 0 -2 0 0
-5 0 5 -5 0 5
T o p d is p la c e m e n t (m m ) T o p d is p la c e m e n t (m m )

(a) Enhanced Timoshenko element (b) Classical DB Timoshenko element

2 0 0

1 0 0
L a te ra l lo a d (k N )

-1 0 0
T e s t d a ta
E u le r-B e rn o u lli 2 e le
-2 0 0
-5 0 5
T o p d is p la c e m e n t (m m )

(c) Enhanced Euler-Bernoulli element

Fig. 10. Top displacement-lateral load curve of the shear-dominated column

-3 6 0
D B 1 e le
D B 2 e le
-3 9 0 D B 4 e le
D B 8 e le
E n h a n c e d
A x ia l lo a d (k N )

-4 2 0

-4 5 0

4 7 6 k N
-4 8 0
0 1 0 0 0 2 0 0 0 3 0 0 0
S te p

Fig. 11. Evolution of axial forces in the shear-dominated column

14 D.C. Feng, August 12, 2019

 4 0 40
3 0 30
2 0 20

Bending moment (kN⋅m)

B e n d i n g m o m e n t ( k N ⋅m )
1 0 10
0 0
-1 0 -10
-2 0 -20
E n h a n c e d 1 e le
-3 0
DB 1 ele
E n h a n c e d 2 e le -30 DB 4 ele

-4 0 -40
- 0 .2 - 0 .1 0 .0 0 .1 0 .2 -0.2 -0.1 0.0 0.1 0.2
C u rv a tu re (m −1
) Curvature (m−1)

(a) Enhanced Timoshenko element (b) Classical DB Timoshenko element

Fig. 12. Local section responses of the shear-dominated column

0.10 0.10
Axial stress σxx
Lateral stress σyy
0.05 0.05 Shear stress τxy
Section depth (m)

Section depth (m)

0.00 0.00

-0.05 Axial stress σxx -0.05

Lateral stress σyy
Shear stress τxy
-0.10 -0.10
-25 -20 -15 -10 -5 0 5 10 -20 -15 -10 -5 0 5 10
Stress (MPa) Stress (MPa)

(a) At top displacement 2.75 mm (b) At top displacement -2.82 mm

Fig. 13. Local fiber responses of the shear-dominated column

337 in Fig. 13. As can be seen, the lateral equilibrium of the fiber stress state is well fulfilled. Though
338 the shear strain is assumed to be uniform, the predicted shear stresses are non-uniformly distributed.
339 As expected, the maximum shear stress is presented around the center of the section.
340 Finally, regarding the computational performances, 3454 global Newton-Raphson iterations
341 are needed for one enhanced element and 3449 iterations for four classical DB elements. The
342 corresponding computation time is 77 seconds and 76 seconds, respectively. Fig. 14 also shows
343 the iterations for calculating the axial enhancement parameter on the element level. It can be seen
344 that in most steps 2-4 iterations are sufficient to get the converged results.

345 CONCLUSION
346 This paper presents an improved DB Timoshenko beam element with enhanced-strains. The
347 strain enhancement is introduced to the axial elongation of the section. As the axial equilibrium
348 condition is now strictly enforced, the inaccuracy of classical DB elements due to the weak
349 satisfaction of equilibrium are effectively avoided. The shear-locking is overcome at the element

15 D.C. Feng, August 12, 2019

 1 2

E n h a n c e m e n t ite ra tio n s
4

0
0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0 2 5 0 0 3 0 0 0
S te p

Fig. 14. Computational cost of the enhanced element for shear-dominated column

350 level with a locking-free finite element approximation, and at the material level a multi-dimensional
351 constitutive model is used for concrete fibers in order to account for axial-flexure-shear interactions.
352 Numerical results of two typical RC columns under cyclic loading are presented to validate the
353 proposed enhanced element. It has been found that the proposed Timoshenko beam element with
354 enhanced strains can well capture both flexure and shear responses of RC columns, with less
355 elements compared to the classical DB element. Moreover, the computational cost does not grow
356 too much even if the additional element-level iteration for the axial enhancement parameter is
357 needed.

358 DATA AVAILABILITY STATEMENT

359 Some or all data, models, or code generated or used during the study are available from the
360 corresponding author by request.

361 ACKNOWLEDGEMENTS
362 This work is supported by the National Key R&D Program of China (2017YFC0803300), the
363 Natural Science Foundation of Jiangsu Province (Grant No. BK20170680) and the National Natural
364 Science Foundation of China (Grant No. 51708106, 51878294).

365 REFERENCES

366 Alemdar, B. N. and White, D. W. (2005). “Displacement, flexibility, and mixed beam-column finite
367 element formulations for distributed plasticity analysis.” Journal of Structural Engineering,
368 131(12), 1811–1819.

369 Arakawa, T., Arai, Y., Mizoguchi, M., and Yoshida, M. (1989). “Shear resisting behavior of short
370 reinforced concrete columns under biaxial bending-shear.” Transactions of the Japan Concrete
371 Institute, 11, 317–324.

372 Calabrese, A., Almeida, J. P., and Pinho, R. (2010). “Numerical issues in distributed inelasticity
373 modeling of rc frame elements for seismic analysis.” Journal of Earthquake Engineering, 14(S1),
374 38–68.

16 D.C. Feng, August 12, 2019

375 Ceresa, P., Petrini, L., Pinho, R., and Sousa, R. (2009). “A fibre flexure-shear model for seismic
376 analysis of RC-framed structures.” Earthquake Engineering & Structural Dynamics, 38(5), 565–
377 586.

378 Crisfield, M. A., Remmers, J. J., Verhoosel, C. V., et al. (2012). Nonlinear finite element analysis
379 of solids and structures. John Wiley & Sons.

380 Ehrlich, D. and Armero, F. (2005). “Finite element methods for the analysis of softening plastic
381 hinges in beams and frames.” Computational Mechanics, 35(4), 237–264.

382 Feng, D.-C. and Ren, X.-D. (2017). “Enriched force-based frame element with evolutionary plastic
383 hinge.” Journal of Structural Engineering, 143(10), 06017005.

384 Feng, D. C., Ren, X. D., and Li, J. (2016). “Implicit gradient delocalization method for force-based
385 frame element.” Journal of Structural Engineering, 142(2), 04015122.

386 Feng, D.-C., Ren, X. D., and Li, J. (2018a). “Cyclic behavior modeling of reinforced concrete shear
387 walls based on softened damage-plasticity model.” Engineering Structures, 166, 363–375.

388 Feng, D. C., Ren, X. D., and Li, J. (2018b). “Softened damage-plasticity model for analysis of
389 cracked reinforced concrete structures.” Journal of Structural Engineering, 144(6), 04018044.

390 Feng, D.-C., Wu, G., Sun, Z.-Y., and Xu, J.-G. (2017). “A flexure-shear timoshenko fiber beam
391 element based on softened damage-plasticity model.” Engineering Structures, 140, 483–497.

392 Feng, D.-C. and Xu, J. (2018). “An efficient fiber beam-column element considering flexure–
393 shear interaction and anchorage bond-slip effect for cyclic analysis of rc structures.” Bulletin of
394 Earthquake Engineering, 16(11), 5425–5452.

395 Filippou, F. C., Bertero, V. V., and Popov, E. P. (1983). “Effects of bond deterioration on hysteretic
396 behavior of reinforced concrete joints.” Report no., Earthquake Engineering Research Center,
397 University of California, Berkeley, CA.

398 Hellesland, J. and Scordelis, A. (1981). “Analysis of RC bridge columns under imposed deforma-
399 tions.” IABSE colloquium, 545–559.

400 Hibbitt, Karlsson, and Sorensen (2001). ABAQUS/Standard user’s Manual, Vol. 1. Hibbitt, Karlsson
401 & Sorensen.

402 Kagermanov, A. and Ceresa, P. (2017). “Fiber-section model with an exact shear strain profile for
403 two-dimensional rc frame structures.” Journal of Structural Engineering, 143(10), 04017132.

404 Koutromanos, I. and Bowers, J. (2016). “Enhanced strain beam formulation resolving several issues
405 of displacement-based elements for nonlinear analysis.” Journal of Engineering Mechanics,
406 142(9), 04016059.

407 Li, Z. X., Gao, Y., and Zhao, Q. (2016). “A 3D flexure-shear fiber element for modeling the seismic
408 behavior of reinforced concrete columns.” Engineering Structures, 117, 372–383.

17 D.C. Feng, August 12, 2019

409 Limkatanyu, S. and Spacone, E. (2002). “Reinforced concrete frame element with bond interfaces.
410 I: Displacement-based, force-based, and mixed formulations.” Journal of Structural Engineering,
411 128(3), 346–355.

412 Mander, J. B., Priestley, M. J. N., and Park, R. (1988). “Theoretical stress-strain model for confined
413 concrete.” Journal of Structural Engineering, 114(8), 1804–1826.

414 Mari, A. R. (1984). Nonlinear geometric, material and time dependent analysis of three dimensional
415 reinforced and prestressed concrete frames. Department of Civil Engineering, University of
416 California.

417 Mullapudi, T. R. S. and Ayoub, A. S. (2013). “Analysis of reinforced concrete columns subjected to
418 combined axial, flexure, shear and torsional loads.” Journal of Structural Engineering, 139(4),
419 561–573.

420 Park, R. (1988). “Ductility evaluation from laboratory and analytical testing.” Proceedings of the
421 9th world conference on earthquake engineering, Tokyo-Kyoto, Japan, Vol. 8, 605–616.

422 Petrangeli, M., Pinto, P. E., and Ciampi, V. (1999). “Fiber element for cyclic bending and shear of
423 RC structures. I: Theory.” Journal of Engineering Mechanics, 125(9), 994–1001.

424 Saatcioglu, M. and Grira, M. (1999). “Confinement of reinforced concrete columns with welded
425 reinforcement grids.” ACI Structural Journal, 96(1), 29–39.

426 Scott, M. H. and Fenves, G. L. (2006). “Plastic hinge integration methods for force-based beam-
427 column elements.” Journal of Structural Engineering, 132(2), 244–252.

428 Spacone, E., Filippou, F. C., and Taucer, F. F. (1996). “Fibre beam-column model for non-linear
429 analysis of RC frames: Part I. Formulation.” Earthquake Engineering & Structural Dynamics,
430 25(7), 711–726.

431 Tarquini, D., Almeida, J. P., and Beyer, K. (2017). “Axially equilibrated displacement-based beam
432 element for simulating the cyclic inelastic behaviour of rc members.” Earthquake Engineering
433 & Structural Dynamics, 46(9), 1471–1492.

434 Taucer, F., Spacone, E., and Filippou, F. C. (1991). A fiber beam-column element for seismic
435 response analysis of reinforced concrete structures, Vol. 91. Earthquake Engineering Research
436 Center, College of Engineering, University of California Berkekey, California.

437 Taylor, R. L., Filippou, F. C., Saritas, A., and Auricchio, F. (2003). “A mixed finite element method
438 for beam and frame problems.” Computational Mechanics, 31(1-2), 192–203.

439 Vecchio, F. J. and Collins, M. P. (1988). “Predicting the response of reinforced concrete beams
440 subjected to shear using modified compression field theory.” ACI Structural Journal, 85(3),
441 258–268.

442 Wu, J. Y. (2011). “Unified analysis of enriched finite elements for modeling cohesive cracks.”
443 Computer Methods in Applied Mechanics and Engineering, 200(45), 3031–3050.

18 D.C. Feng, August 12, 2019

444 Wu, J. Y. (2013). “New enriched finite elements with softening plastic hinges for the modeling of
445 localized failure in beams.” Computers & Structures, 128, 203–218.

446 Wu, J.-Y. (2018). “A geometrically regularized gradient-damage model with energetic equivalence.”
447 Computer Methods in Applied Mechanics and Engineering, 328, 612–637.

448 Wu, J. Y., Li, J., and Faria, R. (2006). “An energy release rate-based plastic-damage model for
449 concrete.” International Journal of Solids and Structures, 43(3), 583–612.

19 D.C. Feng, August 12, 2019

View publication stats