Uiuc Extensions To The Library of Elements For Drain-2Dx: Vancouver, B.C., Canada August 1-6, 2004 Paper No. 2530

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

13th World Conference on Earthquake Engineering

Vancouver, B.C., Canada


August 1-6, 2004
Paper No. 2530

UIUC EXTENSIONS TO THE LIBRARY OF


ELEMENTS FOR DRAIN-2DX

Omer O. ERBAY1, Seung-Yul YUN2, Douglas A. FOUTCH3, and Mark ASCHHEIM4

SUMMARY

Extensions to the library of elements in the nonlinear structural analysis computer program, DRAIN-2DX,
were developed as part of research activities at the University of Illinois at Urbana-Champaign. Two new
elements were added: a connection element capable of modeling connection fracture and degradation for
steel moment frames and a pinching and stiffness degrading element capable of modeling symmetric and
non-symmetric reinforced concrete members. The features and capabilities of each element are discussed
and illustrated with examples.

INTRODUCTION

DRAIN-2DX is a general purpose computer program for static and dynamic analysis of plane structures.
The program was originally developed by Kanaan and Powell [4] as DRAIN-2D. The code was
restructured to its current form, known as DRAIN-2DX, by Allahabadi and Powell [1] and Prakash et. al.
[8, 9]. Both linear and nonlinear static and dynamic analyses can be performed with DRAIN-2DX.
Dynamic loads can be imposed as ground accelerations, ground displacements, and initial velocities. In
addition to static and dynamic analyses, DRAIN-2DX can be used to compute mode shapes and periods of
vibration, and can be used to and carry out linear response spectrum analyses.

Nonlinear response is captured through a step-by-step solution scheme in which imposed loads (static or
dynamic) are incrementally applied to the model of the structure. The load increments can be either
variable (events are considered within steps) or fixed (events are not considered within steps). An event is
defined as any kind of state change that causes a change in the stiffness of the structure. In the variable
load increment case, the program initially utilizes the load increment that is defined by the user and
reduces it until the error in calculated response at the end of the analysis step is less than a defined

1
Senior Engineer, Ph.D., Simpson Gumpertz & Heger, Waltham, MA, USA. Email: OOErbay@sgh.com.
2
Research Engineer, Caterpillar, South Research Park, University of Illinois at Urbana-Champaign,
Urbana, IL, USA. Email: yunseungyul@yahoo.com.
3
Professor of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana,
IL, USA. Email: d-foutch@uiuc.edu.
4
Associate Professor of Civil Engineering, Santa Clara University, Santa Clara, CA, USA. Email:
maschheim@scu.edu.
tolerance level. In the fixed load increment case, user specified load increment is used and kept constant
throughout the analysis.

The DRAIN-2DX program comes with a built-in element library. The element library contains an
inelastic truss bar (Element 01), a simple inelastic beam column element (Element 02), a simple inelastic
translational and rotational connection spring (Element 04), an elastic panel element (Element 06), an
inelastic link element that can act in compression/tension with initial gap or axial force (Element 09), and
a fiber beam-column element for steel and reinforced concrete members (Element 15). In addition to
these elements, the main program is structured to allow addition of new elements. Detailed discussion on
the incorporation of new elements to DRAIN-2DX can be found in Prakash [8].

Following the guidelines provided in Prakash [8], two new elements were developed as part of the
research efforts at the University of Illinois at Urbana-Champaign. The new elements include a
connection element capable of modeling connection fracture and degradation for steel moment frames and
a pinching and stiffness degrading element capable of modeling symmetric and non-symmetric reinforced
concrete members such as Tee beams and walls. In the following sections, the features and the
capabilities of each element are discussed and illustrated with examples. The extended version of the
element library, the associated FORTRAN source codes, and the executable DRAIN-2DX file can be
downloaded from the National Information Service for Earthquake Engineering (NISEE) web site, NISEE
[5].

CONNECTION ELEMENT FOR STEEL MOMENT FRAMES

Introduction
The steel frame connection element, Element 10, is a simple inelastic element for modeling structural
connections with rotational and translational flexibility. The element was developed as part of the SAC
steel project. The connection element that comes with DRAIN-2DX, Element 04 (Powell [7] and Prakash
et. al. [9]), was taken as the basis and extended to include four additional “elasticity codes.” The
additional features can represent ductile flexural behavior, connection fracture, and degradation of steel
moment frame connections. Element parameters allow some or all of these behaviors to be represented.

Element model
The Element 10 formulation follows the Element 04 formulation. The nodes that are connected by the
element must have identical coordinates (i.e. this is a zero-length element). The element can connect
either the rotational displacements of the nodes or the translational displacements. Positive actions
(moments or forces) and deformations as well as available connection types are shown in Figure 1. For a
translational connection, the connected displacement should be aligned with the global coordinate axes
(e.g. horizontal or vertical displacements), displacements aligned in non-coordinate directions cannot be
connected with the element.
Figure 1 Connection types and sign convention for positive deformations (Figure adapted from Powell
[7]).

The Park and Ang [6] damage index model was used to model the effect of low cycle fatigue on fracture
of a connection. The damage index expressed below, Equation 1, is a linear combination of the damage
caused by excessive deformation and the dissipated hysteretic energy. A positive and a negative fracture
deformation under monotonic loading ( δ u+ and δ u− ) are used as input. The input parameters β and α
are the linear and the power coefficient of dissipated hysteretic energy, respectively.

δm β
δ u M yδ u ∫
Damage Index, DI = + dE (1)

where, δ m = maximum response deformation under earthquake


δ u = ultimate deformation capacity under monotonic loading
M y = calculated yield moment
dE = incremental dissipated hysteretic energy
β = non-negative parameter
α = non-negative parameter

The details of the element formulation are beyond the scope of this paper. Interested readers may refer to
Shi [10].

Element features
As mentioned earlier, Element 10 is an extended version of Element 04. For this reason, Element 10
includes all the force-deformation relationship models that Element 04 has. These models are
summarized in Figure 2 and basically include bilinear inelastic, bilinear elastic, and bilinear inelastic with
gap force-deformation relationships.
(a) Bilinear inelastic (b) Bilinear elastic (c) Bilinear inelastic with gap
(Elasticity code = 0) (Elasticity code = 1) (Elasticity code = 2)
Figure 2 Force-deformation relationships in Element 04 (Figures taken from Powell [7]).

In addition to these models, Element 10 also includes four new force-deformation relationships. These
relationships include two models to simulate the failure behavior of steel connections, another model to
represent the behavior of reinforced concrete connections, and a final model to represent the behavior of
panel zones at steel connections. Figure 3 summarizes the features and the hysteresis curves associated
with each model.

The main difference between the steel connection models in Figure 3a and 3b is that pinching behavior
can be represented in the first model. Furthermore the first model can only be used in dynamic analysis
whereas the second model can be used in both dynamic and static analyses.

For model simulating fracture in the steel connections, a modification was required in the solution
algorithm in order to account for the force-unbalance associated with brittle connection fracture. The
modified code is compatible with existing DRAIN-2DX elements.

(a) Steel moment connection hysteresis model (b) Steel moment connection hysteresis model
(Elasticity code = 3) (Elasticity code = -3)
Figure 3 Force-deformation relationships implemented in Element 10.
(c) Concrete connection hysteresis model (d) Steel panel zone tri-linear hysteresis model
(Elasticity code = 4) (Elasticity code = 5)
Figure 3 Force-deformation relationships implemented in Element 10 (continued).

BEAM ELEMENT FOR SYMMETRIC AND NON-SYMMETRIC REINFORCED CONCRETE


MEMBERS

Introduction
The reinforced concrete plastic hinge element, Element 07, uses discrete plastic hinges at the ends of the
element to represent the behavior of ductile reinforced concrete beams and walls. The element is capable
of representing non-symmetric cross sections (e.g. Tee beams and flanged walls), stiffness degradation
(based on Takeda model [11), and pinching. The lack of symmetry of the cross section leads to differences
in the elastic cracked section stiffness and strength for positive and negative moment. A special solution
scheme is introduced to reformulate the stiffness matrix based on movement of the inflection point for this
element.

This element is partially based on the work of Tang and Goel [12], who implemented a stiffness-degrading
element in the program Drain-2DM and subsequent modifications made by Hueste and Wight [3].

Element model
The conceptual reinforced concrete member that the element is intended to represent is shown in Figure 4.
The geometric features of the element are also shown in the same figure. In simple terms, the element
formulation consists of an elastic beam connected in series to rigid-plastic hinges located at either end of
the beam. Rigid end zones may be specified, and elastic shear deformations can be included by specifying
an effective shear area. Although the element computations are based on rigid-plastic hinges, the input
data are based on the moment-curvature response determined for cross sections located at either end of the
beam. The flexural stiffnesses at each end of the beam, which may differ for positive and negative
moment, are assumed to extend to midspan, exclusive of rigid end offsets, if present.
Figure 4 Conceptual reinforced concrete member and the geometric properties of the element
(Figure adopted from Hueste and Wight [3]).

Yielding takes place only at the plastic hinges. Each plastic hinge may have a different elastic stiffness
and yield strength for positive and negative bending, in order to represent the cracked section stiffness and
flexural strength of non-symmertric reinforced concrete members under positive and negative moment.
Uncracked behavior is not modeled; rather, the member ends are assumed to be pre-cracked. The plastic
hinges are assumed to yield only in bending; inelastic axial deformation (e.g. gap opening) is not modeled.
This corresponds to plastic flow along the M direction only, not along the normal to the yield surface of a
P-M interaction diagram. The hysteretic behavior of the hinge is based on the Takeda model. The user
may specify the exponent of the decay function to mimic the degradation in the unloading stiffness.
However, the Takeda rules were not implemented for the small cycles that occur when unloading
commences prior to reaching a previous peak. For these small unloading cycles, the response follows the
unloading slope calculated for the most recent peak in the response quadrant. Additional changes were
made to introduce pinching to the hysteretic behavior. Further information on the details of the element
formulation can be found in Erbay and Aschheim [2]

Status checks incorporated in the element were primarily developed for an event-to-event solution
strategy. Two overshoot tolerance values are defined to capture two different types of events; each event
forces the structure stiffness matrix to be re-formed. Small or zero overshoot tolerances are used for
moment overshoot and small overshoot tolerances are used for movement of the inflection point. The
response of each element can be tracked accurately using an event-to-event solution strategy; other
solution strategies that do not track events may be used but can be expected to be less accurate.

The first type of event concerns the acceptable error in moment associated with overshoot in the non-
linear hysteretic response of the plastic hinge. As is in Element 02 (Powell [7]), a change in stiffness can
introduce error if the stiffness matrix is not updated. An event is called to prevent the error in the
calculated moment from exceeding the specified overshoot tolerance. The change in stiffness associated
with each event may be due to yielding, inelastic unloading, etc.

The second type of event concerns movement of the inflection point (IP). Because the flexural stiffnesses
for positive and negative moment may differ (Figure 5), even for elastic response, any movement of the
inflection point is associated with a change in the stiffness of the element. The element stiffness is
reformulated each time the inflection point moves beyond the specified tolerance. The reformulation uses
the new location of the IP to identify the segments of the member where the positive or negative member
properties are active.
L/2 L/2

EI +j
EI i+

EI i− EI −j
inflection
Mi point, IP
+
- Mj
EI +j
EI i+
EI −j

Li Lij Lj

Figure 5 Possible flexural stiffness variation over the length of the element due to unequal
positive and negative stiffness values along the element length.

The way the IP tolerance is utilized can be explained as follows: Consider the analysis state where the
moment distribution over the member length is as shown in Figure 6a. At this state the location of IP is at
point "a". Based on the current stiffness matrix, subsequent steps in the analysis may cause the IP to move
to location "b" (Figure 6b). Movement of the IP location changes the lengths of the segments over which
the positive or negative member properties are active. This change causes a change in the element
stiffness, which therefore changes the element response. If the element stiffness is reformulated based on
the IP being at “b”, the calculated moment increments at the member end will differ, in general, as
indicated by the dashed line in Figure 6b. The updated moment increments will result in a different
moment distribution as well as a different IP location ("*b" in Figure 6c). At the end of each analysis time
step, the element subroutines calculate the updated moment increments and compare the normalized error
with the calculations based on the initial element stiffness by using Equation 2.

 ∆*M in, +j 1 − ∆M in, +j 1 


NE = max  (2)
 M yi , j 
 

where NE = the maximum of the normalized error for end i and j, ∆M in, +j 1 = moment increment at end i
(or j) by using the stiffness corresponding to IP being at point a, ∆* M in, +j 1 = moment increment at end i (or
j) by using the stiffness corresponding to IP being at point b, and M yi , j = yield moment at the response
quadrant (positive or negative) for end i (or j).
*
Min ∆ *Min+1 Min+1
Moment increments Min +1 IP Shift
a ∆ Min+1
Mjn *
b
∆ *M in,+j 1 − ∆M in. +j 1 ∆ *Mjn+1 b
NE = ∆ Mjn+1
M yi, j Total moment variation
*
Mjn+1
Mjn +1

(a) (b) (c)


Figure 6 Normalized error calculations in moment increments for checking the event
corresponding to excessive movement of the inflection point.
If the error calculated by Equation 1 exceeds the user-specified IP overshoot tolerance, an event is called
and the time step is divided in proportion to estimate the deformation at which the error exceeds the
tolerance. The change in stiffness is a nonlinear function of the change in inflection point. Therefore, a
linear proportioning of the time step only gives a first order approximation to the deformation at which the
error exceeds the tolerance. Once the step at which the error exceeds the tolerance is determined, the
element stiffness is reformulated using the new IP location at that step. It should be noted that the
normalized error in moments is calculated on a step-by-step basis (i.e. using incremental moments
associated with the incremental rotations determined by for the entire structure in each analysis step).
Because the errors may accumulate over successive steps (particularly when small steps are used), the
error (Equation 2) is accumulated over successive steps until the next event is called. Each time the
structure stiffness matrix is updated with the current element stiffness, the accumulated error is reset to
zero.

The default values for the moment and inflection point tolerances are each 1.0e-5. Guidance on the of the
overshoot tolerance for inflection point movement is provided in later sections..

Element features
In Element 07, there are four constants that define the response characteristics of the hysteresis model.
These four constants can be adjusted to represent various response modes that can be useful to
characterize the response of some reinforced concrete members. The description of each constant is in
Table 1 and illustrated in Figure 7. Also provided in Table 1 are the parameter ranges that have been
found to provide stable and convergent solutions.

Table 1 Possible ranges of values for the hysteresis model parameters.

Parameter Description Recommended range


Post-yield stiffness ratio of the moment- 0.001 (almost perfectly plastic) – 0.8 (high
α + and α -
curvature response. post-yield stiffness ratio)
Unloading stiffness degradation
γ + and γ - 0.001 (no reduction) - 0.6 (high degradation)
parameter
β + and β - Stiffness deterioration due to pinching 0.1 (high pinching) – 1.0 (no pinching)
Coefficient to determine the recover
δ + and δ - 0.1 (high pinching) – 1.0 (no pinching)
curvature for pinching

M M
α + EI +
M +y (φ + +
m, Mm )
EI + M +y
φy−
φy+ φ
γ+
 φ y+ 
EI − k +4 k +4 = k1+  + 
k1+ φ 
 m

α − EI − M −y φ y+ φ

Figure 7 Hysteresis model parameters in Element 07.


M
(φ + +
m , Mm )
M +y
(if no prior yielding in
+
current direction) k no pinching
k5+ = β + k +no pinching
k +6 k 6+ +
φ pin = δ +φ o+

(φ + +
pin , M pin ) (
M +pin = k5+ φ pin
+
− φ i+ )
k 5+

(φ ,0)
i
+ φy+ φo+ φ

δ +φo+

Figure 7 Hysteresis model parameters in Element 07 (continued).

Figure 8 shows some of the possible hysteresis loops that can be obtained using different combinations of
model parameters. The data of this figure represent the moment-curvature response at the base of a single
degree of freedom oscillator exposed to static positive and negative horizontal load cycles.

600 600 600


400 400 400
200 200 200
0 0 0
-200 -200 -200
-400 -400 -400
-600 -600 -600
-0.5 0 0.5 1 -1 0 1 2 3 -1 0 1 2 3

equal stiffness and strength, no unequal stiffness, equal equal stiffness, unequal strength,
stiffness degrd., no pinching strength, no stiffness degrd., no no stiffness degrd., no pinching
(Stf. ratio = 1:4, Str. ratio = 1:1) pinching (Stf. ratio = 1:1, Str. ratio = 1:4)
(γ =0.001, β =1.0, δ =1.0) (Stf. ratio = 1:4, Str. ratio = 1:1) (γ =0.001, β =1.0, δ =1.0)
(γ =0.001, β =1.0, δ =1.0)

600 600 600


400 400 400
200 200 200
0 0 0
-200 -200 -200
-400 -400 -400
-600 -600 -600
-0.5 0 0.5 -0.5 0 0.5 -0.5 0 0.5

unequal stiffness and strength, unequal stiffness and strength, unequal stiffness and strength,
no stiffness degrd., no pinching no stiffness degradation, moderate stiffness degradation,
(Stf. ratio = 4:1, Str. ratio = 1:2) moderate pinching no pinching
(γ =0.001, β =1.0, δ =1.0) (γ =0.001, β =0.8, δ =0.8) (γ =0.4, β =1.0, δ =1.0)
Figure 8. Effects of model parameters on the hysteresis loops.
600 600 600
400 400 400
200 200 200
0 0 0
-200 -200 -200
-400 -400 -400
-600 -600 -600
-0.5 0 0.5 -0.5 0 0.5 -0.5 0 0.5

unequal stiffness and strength, unequal stiffness and strength, all parameters are unequal
no stiffness degrd., high high stiffness degradation., no (α +=0.03, α -=0.01,
pinching pinching γ +=0.1, γ -=0.6,
(β +=0.1, β -=0.1 (γ +=0.6, γ -=0.6) β +=0.8, β -=0.5
δ +=0.1, δ -=0.1) δ +=0.8, δ -=0.5)
Figure 8. Effects of model parameters on the hysteresis loops (continued).

Guidance on defining tolerance value to for movement of inflection point


As discussed in the previous sections, the overshoot tolerance for inflection point movement is compared
to the normalized error in moment increments (calculated by Equation 2) at the ends of the element. As
explained in Figure 6, the yield moment in the direction of the response is used to normalize the moment
error at the end of the element. The specified values of the yield moments determine whether the
normalized error is relatively small or relatively large, and this will influence the effectiveness of the
selected overshoot tolerance value. For example, an overshoot tolerance value of 0.001 may be ineffective
for the analysis of an element whose yield moment values lie in the range of 100000 and the moments
calculated during the analysis are on the order of 10. In contrast, an overshoot tolerance of 0.001 can be
too small if the specified yield moments are on the order of 1.0 and the calculated response moments on
the order of 10. A very small tolerance will result in many events and a large number of structure stiffness
matrix determinations, and therefore will require more computation time.
Normalized time/error, %

100 100
Normalized time, %

80 80 Moment Disp.
60 60
40 Time 40
20 20
0 0
0.000001 0.0001 0.01 1 0.000001 0.0001 0.01 1
IP overshoot tolerance IP overshoot tolerance

Variation of computation time with changes in IP Variation of normalized computation time and
overshoot tolerance. The time values are relative normalized error in maximum displacement and
to the maximum time measured for the smallest moment as a function of IP overshoot tolerance
overshoot tolerance
Figure 9 The effect of inflection point overshoot tolerance level on computation time and estimates of
response variables.
Relative error, % 20 20

Relative error, %
15 15
10 10
5 5
0 0
0.000001 0.0001 0.01 1 0.000001 0.0001 0.01 1
IP overshoot tolerance IP overshoot tolerance

Variation of error in maximum positive and Variation of error in maximum moment estimates, in
negative displacements, in percent, as a function percent, on either side of the beam as a function of
of IP overshoot tolerance IP overshoot tolerance
Figure 9 The effect of inflection point overshoot tolerance level on computation time and estimates of
response variables (continued).

Figure 9 shows the variation of computation time (on the ordinate) and the convergence of the results with
respect to different overshoot tolerance values (on the abscissa) for a simple structure subjected to
dynamic analysis. The structure consists of a single-bay frame (period = 0.8s) with equal positive and
negative yield moments of 300kN-m, preloaded by gravity loads, so that movement of the inflection point
occurs under lateral excitation. The variation of the calculated moments is on the order of 100kN-m. In
each analysis the same ground motion is used to calculate the response. The ground motion is scaled to
result in elastic response. For comparison purposes the analysis results for the lowest overshoot tolerance
are assumed to be "exact" in generating the values provided in the plots. The overshoot tolerance for
moment is taken as 1.0e-5 and kept constant for all analyses. As can be seen, the results converge to the
"exact" value for tolerance values smaller than 1.0e-4, which is about 0.1% of the ratio of the moment
variation (100kN-m) and the yield moment (300kN-m). (The apparent convergence of the results for a
relatively high tolerance (1.0) may be coincidental and not representative of the convergence
characteristics for different structures and ground motion records.) Based on this investigation, a value of
0.05% of the calculated moment and the yield moment ratio can be taken as a basis for selecting the
overshoot tolerance for the IP movement.

CONCLUDING REMARKS

Two new elements that were implemented in DRAIN-2DX are introduced. Like for all new elements, first
time user’s are encouraged to get familiar with the features and capabilities of the elements through
simple test analyses. A set of example input files are included with the element files and can be
downloaded from the NISEE’s web site.

ACKNOWLEDGEMENTS

Element 10 was developed under the SAC Steel Project with funding from the Federal Emergency
Management Agency.
REFERENCES

1. Allahabadi, R.; Powell, G. H., “DRAIN-2DX: user guide,” UCB/EERC-88/06, Berkeley:


Earthquake Engineering Research Center, University of California, Mar. 1988.

2. Erbay O. O., Aschheim, M. A., “User’s manual for DRAIN-2DX Element 07,” University of Illinois
at Urbana-Champaign, Urbana, IL 2003. (Also provided in the downloadable version of DRAIN-
2DX from the NISEE’s [5] web site)

3. Hueste, M. B. D. and J. K., Wight, "Nonlinear Punching Shear Failure Model for Interior Slab-
Column Connections," Journal of Structural Engineering, ASCE, Vol. 125, No. 9, September 1999,
pp. 997-1008.

4. Kanaan, A. E.; Powell, G. H., “General purpose computer program for inelastic dynamic response of
plane structures,” UCB/EERC-73/06, Berkeley: Earthquake Engineering Research Center,
University of California, Apr. 1973.

5. NISEE, National Information Service for Earthquake Engineering, “Drain-2DX, Static and Dynamic
Analysis of Inelastic Plane Structures,” http://nisee.berkeley.edu/software/drain2dx/.

6. Park, Y. J., Ang A. H. S., “Mechanistic seismic damage model for reinforced concrete,” Journal of
Structural Engineering, ASCE, Vol. 111, No. 4, 1985.

7. Powell, G. H., “Drain-2DX element description and user guide for element type01, type02, type04,
type06, type09, and type15: version 1.10,” UCB/SEMM-1993/18, Berkeley: Department of Civil
Engineering, University of California, Dec. 1993.

8. Prakash, V., “Dynamic response analysis of inelastic building structures: the DRAIN series of
computer programs,” UCB/SEMM-1992/28, Berkeley: Department of Civil Engineering, University
of California, Dec. 1992.

9. Prakash, V.; Powell, G. H.; Campbell, S., “DRAIN-2DX base program description and user guide:
version 1.10,” UCB/SEMM-1993/17, Berkeley: Department of Civil Engineering, University of
California, Nov. 1993.

10. Shi S., “Evaluation of connection fracture and hysteresis type on the seismic response of steel
buildings,” Ph.D. thesis, University of Illinois at Urbana Champaign, Urbana, IL, 1997.

11. Takeda, T., M. A., Sozen, and N. Nielsen, "Reinforced Concrete Response to Simulated
Earthquakes," Journal of Structural Division, ASCE, Vol. 96 (ST12), 1970.

12. Tang, X., Goel, S. C., "DRAIN-2DM – Technical notes and user's guide," Research Report UMCE
88-1, Department of Civil Engineering, University of Michigan at Ann Arbor, MI, 1988.

You might also like