Gilman2015 Kinematic - Model

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Proceedings of the ASME 2015 Pressure Vessels and Piping Conference

PVP2015
July 19-23, 2015, Boston, Massachusetts, USA

PVP2015-45674

USING NONLINEAR KINEMATIC HARDENING MATERIAL MODELS


FOR ELASTIC-PLASTIC RATCHETING ANALYSIS

Tim Gilman Bill Weitze


Structural Integrity Associates, Inc. Structural Integrity Associates, Inc.
San Jose, CA, U.S.A. San Jose, CA, U.S.A.

Jürgen Rudolph Adrian Willuweit Arturs Kalnins


AREVA GmbH AREVA GmbH Lehigh University
Erlangen, Germany Erlangen, Germany Bethlehem, Pennsylvania,
U.S.A.

ABSTRACT
• Discussion of using monotonic and cyclic data for
assessment of the (non-stabilized) cyclic deformation
Applicable design codes for power plant components and
behavior
pressure vessels demand for a design check against progressive
plastic deformation. In the simplest case, this demand is • Number of backstress terms to be applied for
satisfied by compliance with shakedown rules in connection consistent ratcheting results
with elastic analyses. The possible non-compliance implicates
the requirement of ratcheting analyses on elastic-plastic basis. • Consideration of the temperature dependency of the
In this case, criteria are specified on maximum allowable relevant material parameters
accumulated growth strain without clear guidance on what • Consistency of temperature-dependent runs in
material models for cyclic plasticity are to be used. This is a ANSYS® and ABAQUS®
considerable gap and a challenge for the practicing CAE
(Computer Aided Engineering) engineer. • Identification of material parameters dependent on the
number of backstress terms
As a follow-up to two independent previous papers
PVP2013-98150 ASME [1] and PVP2014-28772 [2] it is the • Identification of material data for different types of
aim of this paper to close this gap by giving further detailed material (carbon steel, austenitic stainless steel)
recommendation on the appropriate application of the nonlinear including the appropriate determination of the elastic
kinematic material model of Chaboche on an engineering scale limit
and based on implementations already available within • Quantification of conservatism of simple elastic-
commercial finite element codes such as ANSYS® and perfectly plastic behavior
ABAQUS®. Consistency of temperature-dependent runs in
ANSYS® and ABAQUS® is to be checked. All three papers • Application of engineering versus true stress-strain
together constitute a comprehensive guideline for elasto-plastic data
ratcheting analysis.
• Visual checks of data input consistency
The following issues are examined and/or referenced:
• Appropriate type of allowable accumulated growth
• Application of monotonic or cyclic material data for strain.
ratcheting analysis based on the Chaboche material
model

1 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


This way, a more accurate inelastic analysis methodology 2. ACQUISITION OF INPUT DATA FOR THE
for direct practical application to real world examples in the CHABOCHE MODEL
framework of the design code conforming elasto-plastic
2.1 General Remarks on the Application of Monotonic
ratcheting check is proposed.
and Cyclic Data
Keywords: Ratcheting check, Elastic-plastic analysis, This section constitutes a discussion of using monotonic
Maximum allowable accumulated growth strain, Locally and cyclic data for assessment of the (non-stabilized) cyclic
accumulated strains, Nonlinear Kinematic Hardening Model deformation behavior.
(NLK) of Chaboche, practical guideline The practical engineering application of the Chaboche non-
linear kinematic material model raises the question of what
material data to use for the calibration of the parameters.
Ratcheting itself is a phenomenon of transient plasticity
1. INTRODUCTION behavior and takes place in the transition region between
monotonic (e.g. quasistatic) and cyclically stabilized material
Ratcheting is progressive distortion of a component under behavior. The complex transitional plastic deformation behavior
cyclic duty. Taken to the extreme, it can lead to an unstable cannot be simulated based on the non-linear kinematic material
component geometry and subsequent collapse. Commonly used model of Chaboche. The set of parameters is calibrated either to
design codes for power plant components and pressure vessels the monotonic or the cyclically stabilized stress-strain behavior.
include rules for ratcheting analysis on elastic-plastic basis. The Furthermore, another problem is to find the non-stabilized
design is considered to be acceptable if either plastic hysteresis loop which is usually not available.
shakedown (accommodation) occurs after a few cycles or the The standard available data set is restricted to the
maximum accumulated local strain at the end of the service life monotonic and cyclically stabilized stress-strain curve. That’s
does not exceed specified limits. For instance, according to why a practical application provides the choice between the
ASME Code Section III [3] the design is considered acceptable monotonic and the cyclic curve and the user has to make a
if either shakedown occurs after a few cycles, or the maximum decision on which one to use.
accumulated local strain does not exceed 5% (for certain In addition, experimental data from ratcheting experiments
materials only) [3, NB-3228.4(b)]. The design rules [e.g. 3] do (either available or generated in dedicated experiments) may be
not provide guidance on the cyclic-plasticity material models used to calibrate the gamma-parameters of the Chaboche non-
that could be used in the analysis to satisfy these limits. This linear kinematic material model. After all the input parameters
paper is a direct follow-up to the previous papers PVP2013- are obtained, they can be used for any geometry and cyclic
98150 [1] and PVP2014-28772 [2]. Further reference can be loading under consideration.
found in PVP2011-57229 [4], which reviewed the local In [1] and [2] monotonic stress-strain data, presented in
ratcheting rules of various design codes, the applicability of the section 4.1, were used as input for the material model. The
NLK models to elastic-plastic ratchet analysis, and the reason for selecting the monotonic data for calibration was that
extrapolation methods to estimate the allowable service cycles it was expected to give conservative predictions of ratcheting
to reach the specified strain limits. All four papers together that would be considered acceptable for design purposes. This
constitute a comprehensive guideline for elasto-plastic is because for the cyclically hardening material considered (SA-
ratcheting analysis and determination of the allowable number 312 TP304 stainless steel) the flow stress is higher for each
of service cycles. The Chaboche model is chosen as it is widely additional cycle than that of the initial half-cycle generally for
available in common Finite Element software codes, the materials that harden with cycles. In other words, the cyclic
identification of the material parameters can be done based on a hardening that occurs with stainless steels was not considered
manual calibration scheme or based on an existing module because its results for ratcheting analysis are less certain to be
within the Finite Element code and it meets the needs of the conservative than those of the monotonic curve. Cyclically
design code requirements regarding elastic-plastic ratcheting softening materials may require the application of the cyclically
analysis. stabilized material model.
Furthermore, the given limit for cyclically accumulated
Additionally, in the framework of this paper the
strains shall be considered. High ratcheting rates and steep
performance of a more simple inelastic analysis approach
gradients are a results from using the monotonic curve. A high
assuming elastic-perfectly plastic behavior is discussed.
number of cycles before reaching the given limit for
accumulated strains may require the application of the cyclic
stress strain curve.
In any case of doubt the ratcheting analysis shall be carried
out both based on monotonic and cyclic stress-strain data
accepting the smaller number of cycles before reaching the
given limit on the accumulated strains.

2 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


2.2 Calibration Basis for Monotonic Data spreadsheet without any help from outside software. It depends
on matching the α and α NLK backstress curves as described
Material data for calibration have to be obtained by
above. The second option are calibration modules of the
experiment (including results from ratcheting tests) or by
commercial Finite Element Codes or any other external
consulting appropriate sources. One possible source of material
calibration tools. Both calibration methods are described
data is Section VIII-Division 2, Annex 3-D, paragraph 3-D.3 of
comprehensively in [1]. In this paper, the first (manual)
the ASME B&PV Code [3].
calibration option is readdressed and extended to the
As already explained in [1] and [2] the stress-strain curve in application on cyclically stabilized stress-strain data.
Section VIII-Division 2 is given in terms of true stress and true
In the first step of manual calibration, a given stress-plastic
strain. For the calibration, they are translated to engineering
stress ( σ ) and engineering strain ( ε ) using equations (1) and
strain data curve has to be divided into M segments inside the
chosen calibration strain range. The number of needed segments
(2). The engineering stress and strain will be used.
M is one more than the number N of Chaboche components
=ee
exp( true ) − 1 (1) used (e.g. 2 or 3), so that M=N+1 holds. Figure 1 shows the
division of a generic stress-plastic strain curve into M segments.
=σ σ true / (1 + e ) (2)

The backstress determination and the calibration process


are based on the stress-plastic strain curve. To get this curve, the
decision has to be made at which point yielding begins. In [1]
an engineering method for the determination of the Elastic
Limit was explained in detail and also applied in [2]. According
to this method the Elastic Limit is estimated at the stress that is
equal to 0.55 times the 0.2% proof stress. For more details see
[1].
The key in the calibration process is the backstress, which
denotes the translation of the yield surface in stress space. For a
one-dimensional representation of an NLK model, the α
backstress is calculated from equation (3).

α= σ − σ 0 (3)

where σ is the stress in a uniaxial stress state of a tension Figure 1: Generic stress-plastic strain curve divided into M
specimen and σ0 is the initial yield stress at the Elastic Limit. segments [5]
For a Chaboche material model with the material parameters The parameters CK and γ K are then determined for all N
CK and γK , the α NLK backstress is described for uniaxial Chaboche components from equations (6) and (7). Each set of
tension by equations (4) and (5). ( CK , γ K ) controls its Chaboche component α K of equation
CK (4).
α K= ( )[1 − exp(−γ K e p )] (4)
γK  σ − σ ( K −1) σ ( K +1) − σ ( K ) 
=CK  ( K ) −  (6)
ε
N  p ( K ) − ε p ( K −1) ε p ( K +1) − ε p ( K ) 
α NLK = ∑ α K (5)
K =1 1
γK = (7)
Again, for more details see [1]. One open question is the ε p(K )
appropriate number of backstress terms N to be applied in
equation (5). In [1] and [2] the application of 3 terms and 2 K is the number of the Chaboche component, σ (J ) and
terms, respectively, was explored. It has been shown that the
application of more than 3 terms does not increase accuracy
ε p( J ) are the stress and plastic strain components selected from
significantly. the stress-plastic strain curve at the J-th point on the segment
boundaries shown in Figure 1. J is set to 0 at the first selected
Two options for determining the parameters of Chaboche point and J=M at the last selected point, as in Table 1 (in the
NLK models for ratcheting analysis are applicable. The first present example of SA-312 Grade 304 as in [1]).
option is a manual calibration which can be performed in a

3 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


Table 1: Stress-strain data points used for manual calibration
For Calibration up to 5% Plastic
Strain
Points Stress Plastic Strain
J [MPa] [%]
0 76 0.000
1 103 0.038
2 132 0.144
3 200 2.264
4 226 5.000

For the Chaboche model, the calculated total backstress Figure 2: Cyclic hardening curve [6]
curve α NLK using the parameter sets ( CK , γ K ) determined by
equations (6) and (7) generally lies below the given α The resulting function
backstress curve, so the parameters have to be further calibrated
CK  ∆ε p 
in a second step. a KMax + σ 0 = tanh  γ K  + σ 0 (9)
The second step of calibration consists of a manual γK  2 
calibration of the parameter sets ( CK , γ K ). The predicted total describes the corresponding cyclic stress-plastic strain
backstress curve (see equations 4 and 5) given by curve predicted by the K-th Chaboche component. Because
each component K can be integrated independently the total
N
CK
α=
NLK ∑(γ
K =1
)[1 − exp(−γ K e p )] (8)
cyclic stress-plastic strain curve is given by the sum of eq. (9)
over the total number N of backstresses used as
K

has to be compared with the α backstress curve and the ∆σ N


C   ∆ε p 
= a Max + σ 0 = ∑  K  tanh  γ K  +σ0
parameter sets have to be refined iteratively until the deviation 2 K =1  γ K   2 
between the calculated and the given curve is considered as
acceptable. (10)
Note, that available ratcheting experimental results may be and represents the total stabilized cyclic stress-plastic strain
curve predicted by the Chaboche model. Furthermore, the
used for further calibration of the γ K parameters. relationship
See [1] for a comprehensive explanation of the calibration
process for a Chaboche model on monotonic data with three ∆ε ∆σ ∆ε p
backstress terms. = + (11)
2 2E 2
2.3 Calibration Basis for Cyclic data between total, elastic and plastic strains remains valid for
cyclic stress-strain curves and may be used to extract the cyclic
In general, metallic materials subjected to a cyclic uniaxial plastic strains from a given cyclic stress-total strain curve, for
load (stress or strain controlled) tend to stabilize after a certain example.
number of cycles. For many materials a cyclic hardening or
The method for manual calibration of the parameter set
softening can be observed within the first 10 to 1000 loading
(CK; γK) to a cyclic stress-plastic strain curve is similar to the
cycles which tends to stabilize afterwards.
manual calibration method described in section 4.3 of
The so called cyclic stress-strain curve represents a relation
PVP2013-98150 [1] used for monotonic data curves.
between the peaks of the stabilized loops, corresponding to
In the first step of manual calibration, a given cyclic stress-
different stress or strain levels as shown in Figure 2, taken from
plastic strain data curve has to be divided into M segments
Lemaitre-Chaboche 1990 [6].
inside the chosen calibration strain range. This method of
As written in Lemaitre-Chaboche 1990 [6], section 5.4.4,
parameter estimation was intended originally for monotonic
the relationship between the stress amplitude ∆σ / 2 and the stress-strain data curves (see Jiang and Kurath, 1996 [5]) but
plastic strain amplitude ∆ε p / 2 for a stabilized cycle works for cyclic data curves as well.
predicted by the Chaboche NLK-model can be obtained by Using this methodology, a given cyclic stress-plastic strain
integrating on two half cycles of tension and compression for data curve has to be divided into M segments inside a chosen
each Chaboche component. calibration strain range. The number of needed segments M is

4 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


one more than the number N of Chaboche components used, so • For each curve, the initial values of C1, γ1, C2, and
that M = N + 1 holds. γ2 are chosen as a set of values that yields a curve
The parameters CK and γK are then determined for all N that is reasonably close to the αNLK curves shown
Chaboche components from equations (12) and (13). Each set in Figure 3 and Figure 4.
of (CK; γK) controls its corresponding Chaboche component • An additional restriction is that the same values of
Max given by equation (9). γ1 and γ2 must be used for both the 70°F (21.1°C)
and the 400°F (204.4°C) model when used in a
 ∆σ ( K ) − ∆σ ( K −1) ∆σ ( K +1) − ∆σ ( K ) 
CK =   (12) temperature-dependent analysis. This is a

 ∆ε − ∆ ε ∆ ε − ∆ε  recommendation from ANSYS® personnel to
 p(K ) p ( K −1) p ( K +1) p(K ) 
prevent unreasonable interpolation on the αK
2 terms [8].
γK = (13)

∆ε p ( K ) Each backstress curve is entered into the ANSYS®
curve-fitting module (under Preprocessor,
Material Models, Structural, Nonlinear, Inelastic,
In equations (12) and (13), K is the number of the Curve Fitting).
Chaboche component, ∆σ ( J ) and ∆ε p ( J ) are the stress and • For each curve, the initial values of C1, γ1, C2, and
the plastic strain range components selected from the cyclic γ2, as well as a yield strength of zero (appropriate
stress-plastic strain curve at the J-th point on the segment since these are backstress curves), are entered in
boundaries. J is set to 0 at the first selected point and equals M the curve fitting module (under Plasticity,
at the last selected point. Plasticity, Kinematic Hardening, 2 Term
The total backstress curve given by eq. (10) using the Chaboche).
determined values of CK and γK will generally not match the • For γ1, γ2, and yield strength, the Fix box is
given cyclic stress-plastic strain data curve, so the parameters checked to prevent these parameters from varying.
have to be further calibrated in a second step. • The Solve button optimizes the C1 and C2 values,
As with the parameter sets determined for monotonic and the Plot button allows a comparison of the
curves (see [1]) the second step of calibration consists of a input backstress curve with the calculated curve.
manual calibration of the parameter sets (CK; γK). The predicted
total cyclic stress-plastic strain curve given by equation (10) has Table 2 shows the resulting parameters, and Figure 3 and
to be compared with the given cyclic stress-plastic strain data Figure 4 show the comparison of the backstress curves with the
curve and the parameter sets have to be refined iteratively until ANSYS®-generated Chaboche models. These figures show that
the deviation between the calculated and the given curve is the two-term Chaboche models appear to be reasonably close to
considered as acceptable. The iterative determination of the the backstress curves. In Figure 3 and Figure 4, αNLK is
parameter sets (CK; γK) is simplified by the fact that each backstress, α1 is the first Chaboche term, α2 is the second
Chaboche component in equation (10) has an initial slope of CK Chaboche term, and α (sum) is the sum of the Chaboche terms.
and some tend toward their saturation value of CK/γK inside the These values differ somewhat from previously derived values
plastic strain range that is used for calibration. In fact, the initial [2, Table 4] because the previous values were based on
slope and the saturation value of each component curve given independently determined backstress curves that were
by equation (9) is comparable to the corresponding value of the somewhat different from Tables 13 and 14 of PVT-14-1036 [7].
predicted monotonic curve (compare with section 4.3 of The current values should allow a more "apples to apples"
PVP2013-98150 [1]). comparison to the N=4 models from ABAQUS®.
Usually several components of the predicted total curve Table 2: Chaboche Parameters from ANSYS®
given by equation (10) converge fast toward their saturation 70°F 400°F
values CK/γK in the lower strain range and do not influence the N C, ksi γ C, ksi γ
further increase of the total curve any more, which greatly 1 13445 1129 10648 1129
simplifies the manual calibration procedure. 2 805 52 673 52

The identical Bree cylinder problem from PVP2013-98150


[1, Section 5] shown in Figure 9 is analyzed herein.
3. CONSISTENCY OF TEMPERATURE- Specifically, this is a cylindrical shell with a mean radius of
® ®
DEPENDENT RUNS IN ANSYS AND ABAQUS 3.740 in (95 mm) and a thickness of 0.394 in (10.01 mm) under
constant pressure of 1595 psi (11 MPa), with an inside surface
The backstress curves from Tables 13 and 14 of PVT-14- temperature that varies between 21°C (69.8°F) and 200°C
1036 [7] are converted to Chaboche models with two (392°F), and the outside kept at 21°C.
components (N=2) using the ANSYS® program as follows:

5 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


Table 3: Comparison of Results
Software, Software, Parameter
model finite element tempera- εpeq after
development analysis N ture, °F 60 cycles
ABAQUS ABAQUS 4 70 0.003156
ABAQUS ANSYS 4 70 0.003035
ANSYS ANSYS 2 70 0.002920
ABAQUS ABAQUS 4 * 0.005186
ABAQUS ANSYS 4 * **
ANSYS ANSYS 2 * 0.005603
ABAQUS ABAQUS 4 400 0.02608
ABAQUS ANSYS 4 400 0.02589
ANSYS ANSYS 2 400 0.02803
* Temperature-dependent parameters
** Since the γ values were not constant, this model could not be used
in a temperature-dependent ANSYS® analysis.
Figure 3: Comparison of Backstress and Chaboche Model,
70°F (21.1°C) The updated N=2 results demonstrate accuracy to within
about 8% of the ABAQUS® results, and this is true even when
using temperature-dependent properties. This improved
accuracy compared with previous results [2, Table 5] is due to
the modifications to the models in Table 2 as compared with the
previous models [2, Table 4]. The authors believe that this level
of accuracy is sufficient given the conservatism involved in
generating the backstress curves [7].

4. IDENTIFICATION OF MATERIAL DATA FOR


DIFFERENT TYPES OF MATERIAL (CARBON
STEEL, AUSTENITIC STAINLESS STEEL)

4.1 Stainless Steel

Monotonic data can for instance be obtained from the


Figure 4: Comparison of Backstress and Chaboche Model,
formulas in ASME Code [3] Section-VIII Div. 2, Annex 3-D.3
400°F (204.4°C)
for materials listed in Table 1 below.
The temperature distributions are determined in ANSYS®
Table 4: Material identification for obtaining the data
using steady state analysis; that is, no transient effects are
considered. A coefficient of thermal expansion of 1.84x10-5 Material Temperature Limit
1/°C (1.022x10-5 1/°F) is used. Ferritic Steel 480°C (900°F)
As in PVP2013-98150 [1], equivalent plastic strain (εpeq) is
used to measure locally accumulated strain [1, Section 5]. In Stainless Steel and Nickel
480°C (900°F)
ANSYS®, this is the EPPLEQV parameter, which is the Base Alloys
equivalent strain based on the plastic strain tensor. The Duplex Stainless Steel 480°C (900°F)
appropriateness of this parameter is confirmed by comparison
with results from ABAQUS®.
Precipitation Hardenable
540°C (1000°F)
Nickel Base
Table 3 includes previous analytical results using N=4
Aluminum 120°C (250°F)
models [2, Table 5], and updates the N=2 results from the
current analyses. It was previously shown that when using the Copper 65°C (150°F)
N=4 Chaboche model without temperature dependency,
ANSYS® results matched those from ABAQUS® quite well, Titanium and Zirconium 260°C (500°F)
demonstrating that a four-term Chaboche model based on a
bounding stress-strain curve (that is, without material Elastic modulus, tensile strength, and yield strength are
temperature dependency) is adequate for modeling ratcheting required to generate the data for the material and temperature
behavior [2]. under consideration. Elastic modulus can be obtained from

6 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


ASME Section II Part D Table TM-1, tensile strength from Table 5: For SA-312 Grade 304 at 204°C (400°F)
Section II Part D Table U-1, and yield strength from Section II
MPa ksi
Part D Table Y-1.
An example for SA-312 Grade 304 stainless steel at 204°C Elastic Modulus 182,028 26,400
(400°F) is given by Kalnins, Rudolph, and Willuweit [7] in the
PVT-14-1036. For this example, the material identification of Tensile Strength 441.3 64.00
the third row in Table 4 was selected. The material properties
are listed in Table 5 below, and the resulting monotonic Yield Strength 142.7 20.70
engineering stress-strain curve is shown in Figure 5 below.

4.2 Setting the Elastic Limit


SA-312 TP304 at 204C (400F)
An excerpt from the book by Davis [9] explains why there is no 500
such thing as an exact Elastic Limit.
"It is tempting to define an elastic limit as the stress at which

Engineering Stress, MPa


plastic deformation first occurs and a proportional limit as the 400
stress at which the stress-strain curve first deviates from
linearity. However, neither definition is very useful, because
measurement of the stress at which plastic deformation first 300
occurs or the first deviation from linearity is observed depends
on how accurately strain can be measured. The smaller the
plastic strains that can be sensed and the smaller the deviations 200
from linearity can be detected, the smaller the elastic and
proportional limits."
100
Even if not exact, an Elastic Limit is needed to obtain the stress-
plastic strain and backstress curves for the Chaboche model.
These curves are shown in Figure 6 for a one-dimensional 0
representation of hardening in the model. The symbol σ |0 that 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
indicates initial yielding is the Elastic Limit. Engineering Strain
As stated in [10], it is the analyst's choice of setting the
Elastic Limit. Various methods can be used. One way is to Figure 5: Stress-strain curve of a stainless steel
define it by a fraction (say, 0.55) on the 0.2% Proof Stress or
the Yield Strength. The problem with that is that the 0.2% Proof Isotropic Hardening, omitted Uniaxial engineering stress-
Stress and the Yield Strength for the same material and for ratcheting analysis plastic strain curve from test
temperature may not be the same in design codes and users of
such codes (e.g., ASME [3] or KTA [12]) would get different
plastic strain curves. To avoid this possible conflict, the method
used in [6] was based on the calculation of the plastic strain by
equation (14)

s 
ep =e − + StrainOffset  (14)
E 
where εp is the plastic strain, ε is strain, σ is stress, and E Backstress
is the modulus of elasticity. A Strain Offset of 0.000157 was
assumed.

Figure 6: One-dimensional representation of hardening in


the NLK model

7 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


4.2 Carbon Steel strain curve for carbon steel does not fit the Chaboche NLK
model. A possible solution to this problem is given next.
Mild steel (carbon steel with up to about 0.2% C) consists
mostly of ferrite. If so, then the monotonic data could be tried
SA-516-70 by Test at Room Temp.
for carbon steel to be obtained from SC-VIII Div.2, Annex 3-
600 D.3 [3], using the second row in Table 4 (Ferritic Steel). The
required elastic modulus, tensile strength, and yield strength are
determined based on the data shown in Figure 7 and Table 6.
Engineering Stress, Mpa

500 The resulting stress-strain curve is shown in Figure 8, together


with that in Figure 7.
400
Table 6: For SA-312 Grade 304 at 204°C (400°F)
300 MPa ksi

Elastic Modulus 195,081 28,293


200
Tensile Strength 572.08 82.97
100
Yield Strength 425.63 61.73
0
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
SA-516-70
Engineering Strain
600
Figure 7: Stress-strain curve of a carbon steel
Engineering Stress, Mpa

500

As seen from Table 4, the materials that apply to Section- 400


VIII Div.2, Annex 3-D.3 [3], do not identify directly carbon
steel. Other sources of monotonic stress-strain curves for carbon
300
steel that are available to analysts are not known to the authors.
To explore how carbon steels fit the Chaboche NLK
200
models for ratcheting analysis, a curve was taken from an actual
test that was conducted for WRC Bulletin 414 by Kalnins and
100
Rana [10]. The stress-strain curve is for low-carbon SA-516
Grade 70 steel (about 0.3% C) at room temperature. That curve
0
is shown in Figure 7. It is expected to be typical for low-carbon
steels. 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Comparison of Figure 7 with Figure 5 shows that the Engineering Strain
shapes of the monotonic stress-strain curves are different. That
Carbon Steel by Test Ferritic by SC-VIII 3-D.3
difference is significant for Chaboche NLK models. In Figure 7,
the stresses between the strains of about 0.5%-1.5% are almost Figure 8: Comparison of the Ferritic curve with that by test
constant with the average of 431.2 MPa (62.5 ksi). This is the
well known Lüders band (or "slip band") that is commonly As seen from Figure 8, the monotonic stress-strain curve by
observed for low-carbon steel. Particularly, the modelling of SC-VIII Div.2, Annex 3-D.3 for ferritic steel does not recognize
this slip band in monotonic data for carbon steels constitutes a the Lüders band and overstates the stress below the strain of 5%
challenge. This requires an assumption on the elastic limit. The that was obtained by test for the SA-570 Grade 70 steel
problem of using this particular curve for an NLK model is that specimen. For strains above 5% the two curves match pretty
the backstress within the band is close to zero and NLK well. Higher stress means higher resistance and lower strain.
hardening begins at the end of the band after about 1.5% strain. However, the difference appears to be small and the use of the
It contradicts the shapes for the stress-plastic strain and formulas of Annex 3-D.3 of [3] for carbon steel, selecting
backstress curves that represent the hardening in an NLK "ferritic steel", could be acceptable.
model, as shown in Figure 6 [11]. For these reasons, the stress-

8 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


5. QUANTIFICATION OF CONSERVATISM OF
SIMPLE ELASTIC-PERFECTLY PLASTIC
SA-312-304 at 204C
BEHAVIOR
0.08
In the context of the ratcheting check, the question is

Accumulated PEMAG
whether the application of simple elastic-perfectly plastic 0.06
behavior is a sufficiently conservative and simple option in
order to avoid the more cumbersome application of more
sophisticated nonlinear kinematic material models, such as the 0.04
Chaboche model. Besides principal reservations of determining
local strains using an elastic-perfectly plastic (EPP) material 0.02
model, the question of sufficient conservatism is discussed by
way of a comparative application example below.
0.00
The geometry consists of the shell sketched in Figure 9.
0 10 20 30 40 50 60
The cross-hatched object on top is a rigid, circular plate that
Cycles
keeps the upper edge of the shell from rotating. The mean shell
radius is assumed 95 mm (3.740 inches), and the thickness 10 Chaboche NLK EPP
mm (0.394 inches). The shell is subjected to a steady internal
pressure of 11 MPa (1.60 ksi). Temperature on the inside Figure 10: Both models are run with material parameters
surface is cycled very slowly between 204°C (400°F) and 21°C (Ci and γi of Chaboche and yield stress of EPP) at 204°C
(70°F), and the outside is kept at 21°C. The material is SA-312
Grade 304 stainless steel. For this case, the EPP model is conservative. As already
stated in [7] trials using Elastic Limits at smaller plastic strain
offsets [than 0.000157] have shown that the accumulated strains
change very little.

T-D 21C-204C by Chaboche and EPP


0.007
Accumulated PEMAG

0.006

0.005

0.004

0.003

0.002

0.001
0 10 20 30 40 50 60
Cycles
Figure 9: Bree Shell TD by Chaboche TD by EPP

60 temperature cycles are run by ABAQUS® for two Figure 11: Both models are run by accounting for
material models. One in Figure 10 is using 4 Chaboche NLK temperature dependency (TD) of the material parameters (Ci
components with Ci and γi (i=1…4) calibrated at 204°C. The and γi of Chaboche and yield stresses of EPP) of the support
other is EPP with Yield Stress of constant 142.7 MPa at 204°C. points at 21°C and 204°C
After 60 cycles, in the first run the accumulated multiaxial
equivalent strain (PEMAG in ABAQUS®) is 2.39%. In the The other run in Figure 11 uses the parameters with
second, PEMAG is 6.48%. temperature dependency of the support points at 21°C and
204°C. For this case the EPP model is not conservative. It
predicts shakedown instead of ratcheting.
The assumption of the elastic limit for the NLK model is
described in section 4.2.

9 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo


6. APPLICATION OF ENGINEERING VERSUS REFERENCES
TRUE STRESS-STRAIN DATA
[1] Kalnins, A.; Rudolph, J.; Willuweit, A., "Using the
Application of engineering versus true stress-strain data Nonlinear Kinematic Hardening Material Model of
can't be resolved based on analysis results, because it will make Chaboche for Elastic-Plastic Ratcheting Analysis,"
too small a difference for the strains dealt with in the context of Proceedings of the ASME 2013 Pressure Vessels and
this paper. That's why the demand for true stress-strain data as Piping Conference, Paper No. PVP2013-98150
input to the commercial finite element software such as
[2] Weitze, W.F.; Gilman, T.D., "Additional Guidance for
ANSYS® and ABAQUS® is rather formal than practical.
Inelastic Ratcheting Analysis Using the Chaboche
Either way, the effect is fairly minor for strain ranges
Model," Proceedings of the ASME 2014 Pressure
allowed in engineering design. It only becomes significant with
Vessels and Piping Conference, Paper No. PVP2014-
large deformation and analyses accounting for geometric
28772
changes from the deformation. This situation should not occur,
at least not in a local ratcheting check [4]. [3] 2013 ASME Boiler & Pressure Vessel Code. Section
III Division 1 - Subsection NB and Section VIII
Division 2
7. APPROPRIATE TYPE OF ALLOWABLE
[4] Rudolph, J.; Kalnins, A.; Götz, A.; Hilpert, R., “Local
ACCUMULATED GROWTH STRAIN
Ratcheting by Elastic-Plastic FEA — Criteria and
Code Based Approaches,” Proceedings of the ASME
The appropriate growth strain used in this paper and in the
2011 Pressure Vessels and Piping Conference, Paper
previous papers [1] and [2] is the multiaxial equivalent plastic
No. PVP2011-57229
strain that is defined by equation (13) in PVT-14-1036 [7]. Note
that applicable design code requirements may differ from that [5] Jiang, Y. and Kurath, P. (1996), “Characteristics of the
choice. E.g., as we already explained in PVP2011-57229 [4] the Armstrong-Frederick type plasticity models”,
strain measure of the German Nuclear Safety Standard KTA International Journal of Plasticity 12 (3), 387-415.
3201.2 [12] is as follows: At the end of service life, the locally
[6] Lemaitre, J., and J.-L. Chaboche, (1990), Mechanics
accumulated principal plastic tensile strain shall not exceed, at
of Solid Materials, Cambridge University Press.
any point of any cross section, the following maximum value:
5.0% in the base metal, 2.5% in welded joints. [7] Kalnins, A.; Rudolph, J.; Willuweit, A., "Using the
The practical implementation of this rule is not simple in a Nonlinear Kinematic Hardening Model of Chaboche
cycle by cycle analysis as the principal strain directions are for Elastic-Plastic Ratcheting Analysis."
likely to change during the history. That’s why an incremental ASME J. Pressure Vessel Technol., PVT-14-1036,
method is to be used in order to sum up the strain increments in 2015, to be published.
that case. In other words, the accumulated strain parameter
[8] E-mail from Rajanikanth Jayaseelan (ANSYS) to
cannot be taken directly from the output of the FE-analysis after
XANSYS forum dated 10/11/2013, Subject: “Re:
the last analyzed cycle.
[Xansys] Xansys Digest, Vol 120, Issue 14”
[9] Davis, J. R. (Editor), 2004, "Tensile Testing", Chapter
8. CONCLUSIONS 1, Introduction to Tensile Testing, page 5, 2nd Edition,
Materials Park, OH, USA, ASM International.
As a follow-up to two independent previous papers
[10] Kalnins A. and Rana, M. D., 1996, "A New Design
PVP2013-98150 ASME [1] and PVP2014-28772 [2] it is the
Criterion Based on Pressure Testing of Torispherical
aim of this paper to close this gap by giving further detailed
Heads," Welding Research Council Bulletin No. 414
recommendation on the appropriate application of the nonlinear
kinematic material model of Chaboche on an engineering scale, [11] Dassault Systemes Simulia Corp., 2014,
based on implementations already available within commercial "ABAQUS/Standard Finite Element Program", 6.13-1,
finite element codes such as ANSYS® and ABAQUS®. The ABAQUS Analysis User's Manual, Subsection 23.2.2,
major aspects of this engineering application have been Providence, RI, by License to Lehigh University.
addressed. The application of the nonlinear kinematic material
[12] Nuclear Safety Standards Commission, 2013,
model of Chaboche is recommended for local elasto-plastic
"Components of the Reactor Coolant Pressure
ratcheting checks in connection with the described strain criteria
Boundary of Light Water Reactors", KTA 3201.2 Part
and appropriate accumulated strain extrapolation methods (see
2, Design and Analysis, Salzgitter, Germany
[4]) in order to minimize calculation efforts.

10 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86316/ on 02/15/2017 Terms of Use: http://www.asme.org/abo

You might also like