Professional Documents
Culture Documents
1989-116-199-213-K Thomas-Elbow, Creep, Cyc
1989-116-199-213-K Thomas-Elbow, Creep, Cyc
North-Holland, Amsterdam
Kevin THOMAS
Westinghouse Electric Corporation, Advanced Energy Systems, PO Box 158, Madison, PA 1,5601, USA
This paper presents the results of an inelastic analysis of a type 304 stainless steel elbow subjected to alternating in-plane
closing and opening bending moments and internally pressurized to 2.76 MPa (400 lb/in2). The elbow diameter was 406 mm
(16 in) with a wall thickness of approximately 12.5 mm (0.5 in). The purpose of the analysis was to validate the use of an
inelastic finite element program (MARC) for the prediction of plastic and creep behavior of LMFBR components at elevated
temperature (593°C). The approach used was to compare the analysis results with detailed measurements of strain and
deformation obtained from a test elbow subjected to the same cyclic bending and internal pressure. The results confirmed the
ability of the MARC program to reproduce, in detail, the complex elevated temperature inelastic response history which
occurred in the test. In particular, the analysis confirmed test results which indicated that creep ratchetting of the elbow would
occur only with respect to local hoop strains at mid-elbow under conditions of reversed cyclic loading and internal pressure.
The analysis was apparently conservative in predicting significantly more ratchetting than was observed in the test. The
analysis also confirmed that cyclic elbow ovaliTstion was stable, i.e., did not show any progressive ovalization (ratchetting).
~ 16.4}
Fig. 2. 406 mm diameter test elbow.
40 ~ SECONDMLTFELaOW ~j~
304SS@1100°F .~
Oy = 23.0KSI ~ ~ -- 250
35 Opl= 11.6KSI ~' ~ ~ "
2O0
lSO
s.o,o..... _ Ioo
I1 I o , ~ , , o, + , . , . . s ,.,,o , ~ ,
Ir I o ...,. ..., .p.S.TA.O,
10 ~'- / ~ -- LINEAREXTRAPOLATION
]l J I sr~- st.,.
Ill / . POINT I (KSI) (%) --
Ill i~=~1.,.1o+,.s,.,.oo.,, ~, I o o
s li- I A I ..s o.om
II J e I ..o O.lO
!1 I c I z,.s o.',s
II I o I zg.o I.OO
U J E I 3xs 3.00
o ,[,~, I I I J o
0 0.2 1 2
STRAIN (%)
Fig. 3. Tensile stress-strain curves at 593°C for the 304 stainless steel elbow material.
202 K. Thomas / Creep ratchetting tests
l
sion and can therefore be considered to be displacement
controlled. Accordingly, the loading cycle was chosen to
be cyclic end rotation with a hold period at the point of
maximum rotation. It was also decided to differentiate
between cycles involving only closing rotations and
~ 0.0
those going from an opening state to a closing state.
Three sets of cyclic loadings were therefore selected and o
are shown in table 1.
ideal case, one should know the variation of those Elapsed Time (HR)
properties throughout the actual elbow geometry. Such Fig. 4. Applied moment history for the elbow creep ratchetting
detailed information could obviously not be obtained test.
prior to the test, neither could it be obtained in a timely
fashion for analysis subsequent to the test. As a practi-
cal compromise, the best data available consisted of 1 and 2 pressures). Fig. 4 shows a time-history of end
stress-strain curves obtained from test coupons moment applied to the elbow. After the initial loading,
extracted from previously tested elbows. Since all the a return to zero rotation required an opening moment
elbows were made from the same heat of material, this due to the generation of both plastic and creep strains.
was considered to be reliable data. The test coupons The moment relaxation for each hold period is dearly
themselves were obtained from locations in the elbows visible and amounted to about 13% per cycle. The gap
which were believed to have remained elastic during the in the middle of the time history was due to a planned
elbow tests. Fig. 3 shows stress-strain curves computed shutdown during the test. The second gap was due to an
as the average of data obtained from several coupons unplanned excursion which automatically shut down
for the first and second MLTF test elbows. For com- the test. Fig. 5 shows the time history of hoop strain at
parison, an average stress-strain curve for Type 304 the mid-elbow crown. The strain excursions during the
stainless steel customarily used for design purposes in shutdown periods were due to instrumentation anoma-
the US [2] is also shown. For the creep ratchetting test lies and do not signify real events. The strain history
analysis it was decided to use the lower of the two indicates some incremental growth in hoop strain dur-
measured curves as a compromise with the design data. ing the opening-closing cycles with internal pressure.
Creep data for the actual elbow material could not This did not occur in the initial series with closing
be obtained in a timely fashion for purposes of the rotations only, neither did it occur when the pressure
analysis and therefore design curves for 304 stainless was removed (Series 3 cycles). The growth was of very
steel were used. These were, the so-called Blackburn limited duration and appeared to saturate at about the
equations [6] with either a single or double exponential sixth cycle.
primary creep term. The hoop strain at the crown is the maximum strain
developed in the elbow and peaked at about 0.36%
2.3. Test results which corresponds to a stress of about 154 MPa (22.3
KSI).
The test, as performed, consisted of five cycles of The maximum moment measured during the first
closing rotation (Series 1), six cycles of opening-closing loading sequence was 90.5 kN m (803 kip-in). Sobel [7]
rotation and about eight cycles of opening-closing ro- has shown that an extension of the Southwell plot
tation at greatly reduced internal pressure (1% of Series method for elastic-plastic deformation can be used to
K. Thomas / Creep ratehetting tests 203
7ooo.0L, IIIIII I il IIIIIIIII IIILILi'-' with exact rigid body mode representation and C 1 con-
tinuity. The finite element mesh had also been opti-
6000.0
mized for the M L T F elbow geometry with respect to the
circumferential distribution of elements. This model
5000.0
was therefore selected as the basis for analysis of the
creep ratchetting test. Fig. 6 shows the mesh used in this
4000.0
application. Because the test elbow was subject to in-
plane loading only, it was possible to model only one
half of the elbow assembly with symmetric boundary
conditions being applied to the edges lying in the plane
of symmetry. Based on the work of Dhalla, seven d e -
I11 ments were used around the elbow half-circumference
°.°III
1oOO.O
with eight sets of elements axially. The total number of
I
-I000.0
-2oo0.o 'll I
III Ili Ilill II III I IILI
I1
I!1 LIII
elements was 140 giving a gross total of 2208 degrees of
freedom. Special constraint relations were used to con-
nect the singly curved elements of the pipe to the
doubly curved elements at the elbow-pipe junctions. In
addition, constraint conditions were imposed at the
0.O 1000.0 2000.0 3OO0.O 4000.0 5000.0 loaded end to enforce circularity and planarity and to
Elapsed Time (HR) permit loadings to be specified in a simple way at a
Fig. 5. Measured hoop strain history at the crown of the single node. These conditions were nonlinear to account
mid-elbow section. for finite rotations of the loaded end.
Temperatures were specified at every node in the
model and were assumed to be uniform through the
wall. They were developed by interpolation from a total
estimate a nonlinear buckling load from test (or analyti- of 54 thermocouple measurements of which 34 were
cal) data. Using this method, the buckling moment for located on the elbow. Temperatures were essentially
the elbow under test was estimated as 117.7 kN m (1042 constant throughout the test and were taken as such in
kip-in) (see Appendix for details). Thus, the maximum the analysis.
moment in the test was 77% of the estimated buckling
load.
~
MEAN THICKNESS z 0.414S IN,
MEAN RADIUS : 7.|11E IN.
HTSD program is the validation of detailed and sim- SENO RADIUS : 24.0 IN.
Ideally, analysis of a multiple cycle loading history As mentioned in § 2.1, pre-test analysis was per-
with plasticity and creep should be based on a material formed using the finite element model in order to
model which simulates the cyclic stress-strain char- establish test conditions most likely to induce ratchet-
acteristics of the material. The constitutive models ting. The results of this analysis showed that ratchetting
available in the MARC program included classical iso- did not occur in the absence of internal pressure. With a
tropic and kinematic hardening and also the ORNL pressure of 2.76 MPa (400 psi) there was evidence of
Tenth Cycle Hardening Rule [9] for cyclic strain ratchetting in the analytical results and pressure was
hardening. Since only monotonic stress-strain data had therefore specified for the test.
been obtained from the coupon tests, the choice was A further question, which could not be answered by
therefore restricted to classical bilinear kinematic pre-test analysis, lay in which creep equation to select.
hardening. Fig. 7 shows the stress-strain data and the In order to obtain the best choice, the first loading cycle
bilinearization selected for the analysis of the test. The with 160 h creep hold time was analyzed twice using
bilinearization followed the recommendations of RDT both the single and double exponential Blackburn equa-
Standard F9-5T by selecting the plastic slope so that it tions. Fig. 8 shows a comparison of the analysis and test
passed through points A and B at 50% and 100% of the data for Cycle 1 for both creep equations. It is clear that
maximum anticipated strain. This gave a yield stress of the so-called 1974 (double exponential) equation gave
118 MPa (17.1 KSI). far more relaxation than was obtained in the test whereas
The creep model used was classical strain hardening, the 1971 (single exponential) equation gave better re-
based on the well-known observations [9] that it gives suits. Fig. 9 shows the end moment relaxation history in
better correlations with test data under conditions of more detail. The single exponential equation under-
varying stress. estimated the primary creep but gave a slightly higher
/
28000
175
24000
/ O.T. ,,,.',,--
2OOOO /I
t ~P
150
125
16000
,.j
lzooe E
t5
8000
N
4000
o| ° I I L I 1 1 l I I o
0.0 0,1 0.2 0.3 0.4 0.5 0.6 0,7 0.8 0.9 1.0
STRAIN (Percent)
F i g . 7. S t r e s s - s t r a i n curve bilincarization.
K. Thomas / Creep ratchetting tests 205
1200
t O CYCLE1WITH1974CREEPEQ. 120
1000 •% CYCLE1WITH1971CREEPEQ.
["1 CYCLE1 TESTOATA lOO
8OO
80
60
40
i 400 g~
200 ~ _
20
,,~ 0
-20
-200
(~0 MLTFTEg,T 4 -CYCLEI
-40
-400
-00
.,-600 --
-80
I I I I ] [
-800
0.0 0.2 G.6 0.9 1.2 1.5 1,8 2.1
END ROTATION(Degrees)
Fig. 8. Cycle I - comparison of test and analysis results.
secondary rate. Overall, the relaxation was better pre- zero rotation and this was also reproduced in the analy-
dicted by the 1971 equation. sis. Cycle 2 and subsequent cycles then followed a
The detailed procedure for incrementing loads in the similar pattern with return to a 'zero' rotation under
analysis is illustrated in table 2. This shows the history pressure of -0.015 °. At a number of points in the
of the first and second cycles with the end rotations loading-unloading sequence, a zero load increment was
applied as they were observed in the test. The initial applied in order to ensure that any residual loads due to
pressurization to 2.76 MPa produced a slight opening nonlinearities were eliminated. The cyclic analysis was
rotation ( - 0 . 0 1 5 ° ) . The temperatures were then im- performed for the first fourteen cycles only. This was
posed which produced a slight change in rotation of because the unplanned load excursion (which shut down
0.001 °. Finally, the end rotation was increased to 1.415 ° the test after Cycle 14) made the continuity of test
which equalled to the total change of 1.431 ° (closing) results somewhat unreliable after this point. Fig. 10
achieved in the test. A creep hold time of 159.2 h was shows the moment-rotation curves obtained from the
then imposed while maintaining the end rotation fixed analysis for the fourteen cycles analyzed. The steps in
at 1.431 °. Cycle 1 in the test did not return to exactly the plots are due to the residual load correction incre-
206 K. Thomas / Creep ratchetting tests
9OO
800
TEST
A
Q.
ANALYSIS -
1971 CREEPEQ.
I
,ofo ANALYSIS -
1974 CREEPEQ.
3OO
g J I I i I I z I I
O 20 40 60 80 100 120 140 160 180 200
CREEP HOLD TiME (Hours)
Fig. 9. Cycle 1 - creep relaxation comparison.
ments. A striking feature of these plots is the change in Overall, this was judged to be acceptable. From the
slope for the hysteresis loops corresponding to Cycles economic viewpoint, this was a costly analysis (see the
12-14. This effect is due, of course, to the reduction in Appendix for details), but it was felt that adequate
stiffness arising from the removal of the internal pres- results would not have been obtainable with a less
sure. elaborate procedure.
The degree of refinement in loading increments was
dictated by the need to obtain reliable equilibrium
throughout each loading cycle and creep period. A 4. Comparison of test and analysis
measure of the adequacy of the refinement was given by
the corrections observed at each of the zero load steps.
4.1. First load cycle
40O
hysteresis loop for the complete cycle where the test
I f -- 4o
data shows marked nonlinearity in the test data as
ZOO
unloading begins. This could not be reproduced by the
analysis because the kinematic hardening model un-
20O
loads elastically. No obvious explanation exists for the
test behavior. Conceivably, a complex system of resid-
100
ual stresses could produce an effect of this kind but this
can only be speculative based on the current work.
o
Nevertheless, fig. 8 shows good overall agreement be-
tween test and analysis and establishes the ability of the
-100 analysis to model the cyclic behavior in a credible
~ - -2o manner.
-200
r -_3o
-300
4,2. Cyclic end-moment history
- - -40
-400 The cyclic variation of the end moment generated by
-0,02 -0,01 0,00 0,01 0.02 0.03
the imposed end rotations is shown in Figs. 10 and 11.
ROTATIONTHETA-Z(Rediins)
It is clear that both the maximum and minimum values
Fig. 10. End m o m e n t - r o t a t i o n curves (Cycles 1-14) - M A R C and the cyclic ranges were underpredicted by the analy-
analysis. sis. This is consistent with the observations already
made regarding the linear stiffness and the buckling
Appendix but are believed to be caused mainly by moment in which the analysis underestimated the test
testing and equipment uncertainties. results. However, both test and analysis show similar
Another point of comparison is the shape of the behavior with respect to the variation from cycle to
900
400PSI I 100
400PSi CLOSING& OPENING
800 L CLOSINGROTATION D~?A?IflN 4 ~1
700 8O
600
500
400
40
300
"~ 200 2O
100
~ o 0 z
5° -100
-2O
~ -200
-300
-4O
--400
-500 -60
-600
-700 -80
-000
-900 -100
200 400 600 800 1000 1209 1400 1600 1800 2000 2200 2400
CREEPTIME (Hours)
Fig. 11. Applied end m o m e n t history - test and analysis.
208 K. Thomas / Creep ratchetting tests
cycle of the end moment at the opening end of the than was available from the test.
cycle. The comparison between test and analysis is With regard to the creep behavior of the hoop strain,
better for the Series 1 and 2 tests with 2.76 MPa good correlation was obtained between analysis and
internal pressure than for Series 3 with essentially zero test. For the Series 1 cycles, the test results did not
pressure. These results support the applicability of the indicate any incremental growth in hoop strain whereas
kinematic hardening model with a stress-strain curve the analysis showed significant creep per cycle but also
based on the elbow coupon data and suggest that cyclic little, if any, incremental growth. In the second series,
hardening (or softening) did not play a significant role where the elbow is opening to 0.75 °, the test showed
in the inelastic behavior of the elbow. significantly more creep and even more incremental
growth which appeared to saturate by the sixth cycle.
4.3. Hoop strain history The analysis results showed similar behavior but to a
more marked degree. When the pressure was removed
The point of maximum hoop strain in an elbow is for the third series, the test data showed a tendency for
close to the crown at mid-elbow. Fig. 12 shows the the hoop strains to diminish during the hold period. The
comparison between test and analysis at this location. It analysis, on the other hand, predicted creep growth (per
is clear that maximum hoop strains were greater in the cycle) but no incremental growth in total strain.
analysis than in the test (by more than 50%). However, These results are regarded as the most significant
if a comparison is made between cyclic changes in because they show that limited creep ratchetting may
maximum hoop strain, the results are quite close (within occur if the elbow has internal pressure and if the end
about 10%). The difference in the maximum values may rotations are cycled from opening to closing. If either of
be due, in part, to the difficulty of positioning the strain these conditions is absent, then ratchetting does not
gauges accurately at the location of maximum strain. seem to occur.
Another contributing factor is the uncertainty in the
actual yield behavior of the test elbow, which is further 4.4. Axial strain history
compounded by the known variability of the properties
within the elbow. A full explanation of these differences Fig. 13 shows the cyclic behavior of the maximum
requires a more detailed set of strain measurements axial strain at mid-elbow. This occurs at an angular
7000
6000
O MARC ANALYSIS
-o--~,.o..ff.~
-- ~ ~ TEST DATA .....<:r-- ..O'-- ."(>'-
..O"
5000
F
4000
.~j. -.O--
3000 m -
== lOOO O
0
I
I
-1000
-2000 I I i I I L I i I I I
200 400 600 800 1000 1200 1400 1600 1900 2000 2200 2400
CREEP TIME (Hours)
Fig. 12. History of maximum mid-elbow hoop strain on outer surface - test and analysis.
K. "lhomas / Creep ratchetting tests 209
3000
l
2000 ~ - ~ MARCANALYSIS
L
~,--~-- TESTDATA
1000
.w
~ -lOOO
z_
_~ -2O0O
-3000_4000f
_moo I I I I I I I I I I I
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 24O0
CREEP TIME (Hours)
Fig. 13. History of maximum mid-elbow axial strain on outer surface - test and analysis.
circumferential position of about 80 ° from the ex- contrast to overprediction in the case of the hoop strain.
trados. The difference between the analysis and test is These differences are believed to be due, in part, to the
more marked in the case of the axial strain. The analysis high circumferential gradients in axial stress arising
underpredicted the measured strain by about 30% in from ovalization. The positions of the strain gauges in
0.0 -~ 20
0.6
0.4
0.2 ,t
~ 0.0
~ -0.2
f-+--Z-SS,
-0.4
- A
~ ~-1o
O MARC ANALYSIS
-0.6
m ~ ~ TESTDATA
-0.0
~ _ _ I J__,.,t__.__L l
200 400 690 800 1009 1200 1400 1600 1800 2N0 2200 2400
CREEP TIME (Hcm*+s)
the test and of the integration points in the analysis did Table 4
not exactly match which, in the presence of high gradi- Accumulated creep strain/cycle
ents, gave rise to significant differences.
With regard to the cyclic behavior, no evidence of Test series Creep strain/cycle
ratchetting of axial strains can be observed either from (1) 0° ~1.44 ° 360x l0 -6
the test data or the analysis. 2.76 MPa
(2) -0.75 ° ~ 1.44° 333 X 10 - 6
4.5. Mid-elbow ovalization history 2.75 MPa
(3) -0.75 ° ~ 1.44° 133 x 10-6
Perhaps the most significant of the comparisons
0.03 MPa
which can be made is for the mid-elbow ovalization. In
planning this particular test it was suspected that a
mechanism might exist for incremental ovalization and
that cycling of an opening-closing nature might induce
such behavior. Fig. 14 shows that for the conditions 4.6. Summary of results
investigated, no evidence of growing ovalization was
obtained. The presence of internal pressure during the Table 3 presents a summary of the comparisons
creep period can be expected to reduce the Ovalization between test and analysis for the maximum values of
(and also the reverse ovalization during the opening end moment, hoop strain, axial strain and ovalization
part of the cycle). Even during the third series of tests and their corresponding cyclic ranges. Quantitatively,
(approximately zero pressure) no evidence of increasing the agreement can only be regarded as fair. Test results
ovalization is apparent. and analysis agree best in the case of maximum mid-
Comparison between test measurements and analysis elbow ovalization, but the ovalization range is signifi-
for ovalization shows good agreement at maximum clos- cantly underpredicted by analysis. This must be con-
ing rotations, but much worse agreement at maximum trasted with the maximum hoop strain results where
opening. The ovalization range in the test is much larger analysis seriously overpredicts the test results but only
( > 30%) than the calculated range. This is the opposite slightly overpredicts the hoop strain range.
result from the case of the hoop strains where the Since the ovalization should be proportional to the
ranges were in good agreement and the maximum val- hoop strain, more consistent agreement is to be ex-
ues in poor agreement. pected. There is no clear explanation for this behaviour,
but it may be a result of unknown circumferential
variation of material properties in the test elbow.
Table 3 The comparison of predicted and measured strains
Summary of analysis and test comparisons was necessarily in terms of the total strain components.
The behavior of creep and plastic strains could only be
Quantity (analysis) - (Test) %
obtained from analysis. The analysis results for creep
(Test) strain can be summarized by creep strain rates on a per
Series I Series 2 Series 3 cycle basis. Table 4 shows the average rates for the three
cycles cycles cycles series of cycles. It is evident that the creep rate is largely
(2.76 MPa) (2.76MPa) (0.03MPa) dependent on the presence of internal pressure. The
cyclic loading with reverse plasticity, produces an offset
Max. end moment - 18 -16 -35
to the steadily accumulating creep strain which is what
Max. end moment
range - 14 - 12 - 23 limits the incremental growth.
Max. hoop strain 36 57 78
Max. hoop strain
range <5 11 16 5. Summary and conclusions
Max. axial strain - 35 -28 -15
Max. axial strain An in-plane cyclic bending test on a prototypic
range - 38 - 30 - 22 LMFBR stainless steel elbow was conducted at 593°C.
Max. ovalization <5 10 18 Creep hold periods of approximately 160 h were im-
Max. ovalization
posed at the point of maximum bending for each cycle.
range - 23 - 26 - 23
The elbow was tested for a total of twenty cycles during
K. Thomas / Creep ratchetting tests 211
which an internal pressure of 2.76 MPa was sustained Regarding the likelihood of elbow ratchetting in
for the first eleven. Measurements of displacements and service, it must be concluded to be quite small. The
rotations, ovalization, applied loading, strains and tem- range of elbow rotations used in this study was rela-
peratures were made and recorded continuously tively high, loading the elbow well into the plastic zone
throughout the test. (hoop strains on the order of 0.35%) and with higher
A finite element, doubly-curved shell model of the internal pressure than would be typical for LMFBRs.
test elbow was analyzed using the test conditions for Even under these circumstances, ratchetting tended to
imposed end rotations, temperature, pressure and creep saturate quickly, essentially ruling it out as a failure
hold time. The model was analyzed using the MARC mode.
program. The same model was used also for analysis
prior to the test to establish appropriate test conditions.
The purpose of the testing and analysis was tO de-
termine the potential for creep ratchetting of elbows Appendix
under cyclic bending loads and to validate the use of a
detailed finite element model for analyzing the complex
elevated temperature behavior of a piping elbow.
1. Test anomalies
The tests demonstrated very little potential for creep
ratchetting behavior in elbows under cyclic bending.
Some growth in peak hoop strain in the elbow was Comparison of analysis and test results indicated a
observed for cycles in which the elbow was pressurized significant discrepancy between the measured and pre-
and also for which the end rotations ranged from open- dicted values of moment-rotation stiffness. For the
ing to closing. Both test and analysis showed that this creep ratchetting test, the analysis underpredicted the
growth was enhanced during each cycle by the plastic stiffness by 17.7%. This stiffness discrepancy also af-
loading and reverse loading which took place outside fected the nonlinear part of the response and hence the
the hold period. Thus, it may be concluded that ratchet- predicted buckling moment and creep behavior. Four
ting did occur but that the total strains tended to reach potential causes were identified:
a limiting value after a few cycles. With respect to (1) differences between the assumed values of the elas-
overall deformation, no progressive ovalization was ob- tic modulus for 304 stainless steel and the actual
served in either test or analysis. There appeared to be material data,
no mechanism for incremental distortion of the elbow (2) external restraint on the test elbow, e.g., the sodium
cross-section leading to flow restrictions. The presence drain fine,
of internal pressure clearly inhibits such an effect, but (3) added stiffness due to insulation on the test elbow,
in the absence of pressure there was still no progressive (4) basic errors in the finite element model.
ovalization. The possibility of basic errors in the finite element
Comparing the results of analysis and test validated model was checked by developing a simple beam model
the analytical model and procedure in a general sense. of the elbow with accurate representation of the elbow
The use of a relatively refined finite element model in flexibility including the effects of restrained ovalization
conjunction with simple constitutive laws (i.e., bilinear and internal pressure. The difference in stiffness be-
kinematic hardening and creep strain hardening) proved tween the two models was only 3.5% which was consid-
capable of reproducing the detailed cyclic behavior of ered to be excellent agreement.
the elbow in a realistic manner. The results were rea- With regard to the elastic properties, careful mea-
sonably accurate with respect to strain ranges and ovali- surements of elastic modulus were made on the elbow
zation. Precise agreement between measured and calcu- material at room temperature and 593°C in an effort to
lated response variables could not be obtained because resolve this question. Using coupon specimens, the mea-
the detailed inelastic material properties and their varia- sured moduli were from 4.4% to 8.8% higher than the
tion within the elbow were unknown. values used in the analysis. Applying these modulus
Clearly, for design purposes, detailed inelastic analy- values to the finite element stiffnesses reduced the dis-
sis of components is both feasible and capable of pro- crepancy by up to 50%.
viding reliable results. Inevitably the cost of such analy- It was therefore concluded that the balance of the
sis for cyclic loading is very high (see the Appendix) discrepancy was, in all probability, due to the combined
and can only be justified on a basis of need in specific effects of external restraint and added stiffness due to
situations. the insulation.
212 K. Thomas / Creep ratchetting tests
[8] A.K. Dhalla, Collapse characteristics of a thin-walled [9] C.E. Pugh and D.N. Robinson, Constitutive equations for
elbow: Validation of an analytical procedure, in: Pressure meeting elevated temperature design needs, in: Pressure
vessels, piping and components - Design and analysis, Vessels and Piping: Design Technology - 1982 - A Decade
PVP-Vol. 107 (American Society of Mechanical Engineers, of Progress, eds. S.Y. Zamrik and D. Dietrich (American
New York, 1986) pp. 95-104. Society of Mechanical Engineers, New York, 1982) ch. 3.3.