You are on page 1of 14

Nuclear Engineering and Design 197 (2000) 155 – 168 www.elsevier.


Fracture resistance of cracked duplex stainless steel elbows under bending with or without internal pressure
P. Se ´ me ´ te ´ a,*, P. Le Delliou a, S. Ignaccolo b
Electricite ´ De France (EDF), Direction des Etudes et Recherches, De ´ partement MTC, Les Renardie ` res, 77818 Moret s /Loing Cedex, France b Electricite ´ De France (EDF), Direction de l’Equipement, SEPTEN, 12 -14, a6enue Dutrie6oz, 69628 Villeurbanne Cedex, France Received 17 August 1998; received in revised form 14 July 1999; accepted 13 September 1999

Abstract Electricite ´ De France (EDF), in co-operation with Framatome, has conducted a large research programme on the fracture behaviour of aged cast duplex stainless steel elbows. One important task of this programme consisted of testing three large diameter (580 mm O.D.) thermally aged cast elbows containing either one or two semi-elliptical notches on the outer surface of the flank. The first two elbow tests (called SEM1 and SEM2) were conducted under in-plane closure bending at 320°C and the third one (called SEM3) under constant internal pressure and in-plane closure bending at 60°C. Those tests were carried out with 2/3-scale models of PWR primary loop elbows. This paper presents the results of the experiments and the results of the fracture mechanics analyses, based on finite element calculations. © 2000 Elsevier Science S.A. All rights reserved.

1. Introduction Some components of the primary loop of Pressurized Water Reactors (pump casings, some elbows, pipes, fittings and valve casings) are made of cast duplex stainless steels. The fact that this kind of steel may age even at relatively low temperatures (under 400°C, i.e. in the temperature range of PWR service conditions) has been documented since the early 1980s (Chopra and Ayrault, 1985). The ageing is caused by a microstructural evolution of the ferritic phase. An
* Corresponding author. Tel.: + 33-1-60-737220; fax: + 331-60-736559.

important consequence of this ageing process is the decrease in the toughness properties of the duplex material. It is feared that an embrittlement, associated with the occurrence of casting defects, may increase the risk of failure. Therefore, an extensive programme has been launched in order to determine acceptability criteria for operating cast stainless steel components of the primary circuit. An overview of the research programme on the fracture behaviour of aged cast duplex stainless components conducted by EDF, in co-operation with Framatome, is presented in Massoud et al. (1997). This programme included tests on 2/3 scale mock-ups of the primary circuit elbows (Eripret et al., 1990; Le Delliou et al.,

0029-5493/00/$ - see front matter © 2000 Elsevier Science S.A. All rights reserved. PII: S 0 0 2 9 - 5 4 9 3 ( 9 9 ) 0 0 2 6 3 - 0

Their dimensions were as follow: Table 1 Chemical composition of the elbow steels (by weight%) Material I.6 33. One end of the elbow is fixed to a rigid welded frame embedded in the ground.7 35. The corresponding results are recalled here and are complemented by the results obtained with the third test for which the elbow was loaded with internal pressure and in-plane bending at 60°C. SEM1 SEM2 SEM3 SEM3 a Notch location Flank Near flanka Flank Near flanka Notch orientation Longitudinal Circumferential Longitudinal Circumferential Length 2c (mm) 210 88 210 88 Max.5 14. The notches were manufactured by electric discharge machining and were not fatigue-sharpened. while the third one was aged over 1000 h at 400°C. 2. and the fracture mechanics analyses that were performed for the three tests.3 l Ferrite 34. For each elbow.5 25. at 320°C) were already presented in a previous paper (Le Delliou et al.5 14. The characteristics of the notches are given in Table 2 and are represented in Fig. called  SEM  . Prior to the tests. For the third test. The first two tests (under in-plane closure bending only.72 2.8 26. the crack was located so that it was submitted to tensile stresses. .13 1. the numerical computations.30 Table 2 Characteristics of the machined notches Test I. 2.20 Cr The elbows of the first two tests contained a semi-elliptical notch on the outer surface of one flank and were loaded under in-plane closure bending. SEM1 SEM2 SEM3 Si 1. Se ´ me ´ te ´ et al. loaded under in-plane closure bending. 2.7 c /a ratio 10 3 10 3 15° towards extrados. / Nuclear Engineering and Design 197 (2000) 155–168 1996).D. Test description 2. Description of the elbows The elbows were made of Z3 CND 19-10M duplex stainless steel (French standard equivalent to CF8M). the elbow contained both cracks previously described (one on each flank) and was loaded with constant internal pressure of 15. the equivalent chromium content Cr* (Cr* = Cr + Si + Mo).73 Cr* 25. the first two elbows were thermally aged over 3000 h at 400°C. is schematically represented in Fig. as well as the l ferrite content. 1996).95 22. This paper presents experimental results. depth a (mm) 10.2. with and without internal pressure. The moment loading is generated by pulling on the arm pipe with a long stroke Mo 2.1. 1. outer diameter thickness bend radius bend angle 580 mm 44 mm 900 mm 90 degrees.23 Ni 10..70 21.00 10. Table 1 shows the important elements in the thermal ageing phenomenon of duplex stainless steels.5 MPa and under in-plane closure bending.15 10.7 10.19 1.5 21.71 2. Description of the test conditions The test facility used by EDF.156 P. while the other end is fitted with a 6 m long straight pipe used as a moment arm.D.

Notch location and shape. 1. 2.P. . Schematic drawing of the test frame (SEM3 test configuration). Se ´ me ´ te ´ et al. / Nuclear Engineering and Design 197 (2000) 155–168 157 Fig. Fig.

The bending load was applied at a quasistatic rate. 2 and 3 Flange-SEM1 & 2 Flange-SEM3 Loading saddle SEM1.4 460 – – – – Ultimate tensile strength (MPa) 670 662 743 545 – – – – – – SEM1 cast elbow SEM2 cast elbow SEM3 cast elbow Connecting pipes SEM1 and 2 Connecting pipes SEM3 Flat heads-SEM3 Arm pipe SEM1. Young’s modulus (MPa) 176 500 175 500 188 000 188 000 201 000 201 000 204 000 183 000 201 000 204 000 0.5 321 223 278. The length of the arm pipe reTable 3 Mechanical properties of the three test materials Material I. 2 and 3 Fig.2% Offset yield strength (MPa) 258 259. True stress-true strain curves for SEM1.158 P. / Nuclear Engineering and Design 197 (2000) 155–168 ram. SEM2 and SEM3 tests. with a displacement rate of the ram at 2 mm min − 1. . 3. Test temperature (°C) 320 320 60 320 60 60 20 320 60 20 mained constant during the tests because the loading saddle is allowed to slip under the mobile frame. Se ´ me ´ te ´ et al.D.

1796 0. / Nuclear Engineering and Design 197 (2000) 155–168 Table 4 Charpy energy and toughness properties of the three elbows.D.2497 0. the elbow.2 66. two flat heads were inserted: one between the elbow and the flange (replacing the lower connecting pipe) and the other one between the upper connecting pipe and the arm pipe. then the bending moment was applied in the same way as described before. after ageing Material I.3. Material data The material characterisation programme included chemical analysis. Se ´ me ´ te ´ et al. Charpy impact tests and J -resistance curves determined on . Applied load versus ram displacement curves. an internal pressure was applied. while for the third test. For the third test. the elbow and the connecting pipes were heated at 320°C. Test temperature (°C) 320 320 60 KCV energy (daJ cm−2) 2.4 2. and the connecting pipe were maintained at 60°C. two connecting pipes were inserted: one between the elbow and the arm pipe and the other one between the elbow and the flange. 2.1 2. was at first pressurized up to 15.P. tensile tests.6 87.13 n mean value (Da in mm) 0. During the first two tests.6 J0. 4. The water.2698 159 SEM1 SEM2 SEM3 Fig. the flat heads. where.5 C mean value (kJ m−2) 97. For the first two tests. in addition to the moment loading.5 MPa.2 mean value (kJ m−2) 68 51 58. inside the elbow.

3 (Longitudinal notch) 2. the connecting pipes and the flat heads. Electric potential drop versus ram displacement curves (SEM3 test). For the first two elbows. which remained elastic during the tests.6 (Circumferential notch) SEM1 SEM2 SEM3 SEM3 . the flange. J-resistance tests (at 320°C for SEM1 and 2 and 60°C for SEM3) on aged CT specimens gave the values of J0. Crack initiation Fy (kN) 363/378 336/353 365/391 (Longitudinal notch) 405/426 (Circumferential notch) End of test Fy (kN) 448 483 531 531 End of test moment (kN. three kinds of CT specimens were used: specimens without fatigue-precracking sampled on elbow C -A orientation.160 P. For the other components (flange. the arm pipe. J = C (Da )n. conventional tensile properties are given in Table 3 and the corresponding true stress-true strain curves are shown in Fig. The chemical composition chosen for the elbows leaded to a severe thermal ageing and low toughness properties.2 mm of crack extension) and a power-law fit of the J -R curve. fatigue-precracked specimens with C -R orientation.D.2 (value of J for 0. For the elbows. the flat heads. 3. and the loading saddle were made of carbon steel. specimens without fatigue-precracking with C -R orientation and finally. The connecting pipes. It was Table 5 Main results of the tests Test I. 5. only the Young’s modulus is indicated in Table 3. The specimens were manufactured from a straight pipe cast and aged at the same time than the elbow. / Nuclear Engineering and Design 197 (2000) 155–168 Fig. load- ing saddle and arm pipe). 1T-CT specimens (thickness: 25 mm). Se ´ me ´ te ´ et al.m) 2705 2900 3150 3150 End of test Da (mm) 8 12 6.

specimens without fatigue-precracking with C -A orientation were used. .P.4. The scatter was related to the coarse solidification macrostructure of cast duplex steels. Se ´ me ´ te ´ et al. structure rotation (with inclinometers). A statistical analysis of the data was used to give the results provided in Table 4. / Nuclear Engineering and Design 197 (2000) 155–168 161 Fig. CMOD in several points along the crack length. elbow diameter variations (for ovalization). ram displacement. and strains in the elbow mid-plane. The direct-current electric potential drop method was used to detect crack initiation and to interrupt the test when a large amount of stable crack growth was reached. 2. These measurements included applied load. not possible to show either notch radius effect or orientation effect because of the wide scatter observed on this type of material. Final crack shapes (SEM3 test). All the specimens were tested by the multiple specimen technique (interrupted tests). 6. For the third test. Test results Several measurements were made during the tests.

. View of the finite element mesh (SEM3 test). Applied load versus ram displacement curves are shown for the three tests in Fig. longitudinal notch). 6 (for SEM1 and SEM2 tests. This proves that the effect of the notch on the load versus displacement curve is quite limited. so that they initiated and grew subsequently by ductile tearing. circumferential notch) and 6 mm (SEM3 test. the notch area was cut out from the elbow. The final crack shape was determined optically and. Fig. The moment is evaluated in the mid-section of the elbow taking into account the moment due to the dead weight of the structure. / Nuclear Engineering and Design 197 (2000) 155–168 The machined notches were submitted to tensile stresses due to the applied loading. Table 5 summarises the main results of the three tests. 4. it was recharacterised as a semi-ellipse for the purpose of the analyses. The final shape of the longitudinal crack of the SEM3 test is shown in Fig.6 mm (SEM3 test. 13 mm (SEM2 test). 1996). 2. Fig. 5 shows the evolution of center crack electric potential drop versus ram displacement and the location of the crack initiation point for the SEM3 test. Se ´ me ´ te ´ et al. For this last test. 7. The final crack extension reached 8 mm (SEM1 test). then thermally marked. For the post-test examination of the machined notch. 3). see Le Delliou et al. where F y and U y are respectively the applied load and the displacement of the loading point normal to the moment arm. because the internal pressure and the flat heads limited the ovalization of the elbow and also because the lower test temperature (60°C instead of 320°C) increased the elbow stress-strain curve. the longitudinal notch initiated before the circumferential one. . because of its irregular geometry. cooled in liquid nitrogen to ensure brittle fracture. and finally broken open. The third test curve is stiffer than the others. due to the similarity of the stress-strain curves (see Fig. First two tests curves are very close together.162 P.

8.5 14. Description of the FE models The aim of the computations was to show their ability to simulate accurately the tests by conducting systematic comparisons between experimental Table 6 Crack depths used for the calculations Test I. 3. SEM1 SEM2 SEM3 (Longitudinal notch) SEM3 (Circumferential notch) Initial depth a (mm) 10.P.7 10. crack initiation and crack propagation).6 . Numerical analysis 3. These results were related first to the overall behaviour of the structure (load versus ram displacement curves) and secondly to the local behaviour of the notch (CMOD. Intermediate growth Da (mm) 4 6 3 1.5 14. Se ´ me ´ te ´ et al. The calculations were made with the FE code named Code – Aster® developed by EDF (Vautier and Billet.D.7 and numerical results.1.3 and 2 Final growth Da (mm) 8 12 6 2. G versus applied load curves (SEM3 test). / Nuclear Engineering and Design 197 (2000) 155–168 163 Fig.

Crack growth analysis (SEM3 test-longitudinal crack). in fact. The number of nodes ranged from 12 000 to 16 000 nodes. This general purpose program is mainly devoted to industrial applications.2. The investigations that were carried out during the first calculations enabled us to improve the model. with 32 elements along the crack front. 9. 3. the overall behaviour of the structure was assessed in terms of load versus ram displacement curves. The meshes of the tested structures were built up using solid elements (15 and 20 node elements) and.e. according to the structure analyzed. . two symmetrical cracks). So the FE model was limited to the half-structure containing the notch (it means that the corresponding full model had.164 P. Due to the small size of the notch. Se ´ me ´ te ´ et al. because it took into account Fig. in addition for the third test. The mesh of the SEM3 mock-up included the flat heads as shown at the top of Fig. / Nuclear Engineering and Design 197 (2000) 155–168 1996). They were particularly refined in the crack area. The meshes were constructed from a cracked block (see Fig. in terms of load versus displacement curve). O6erall beha6iour of the tested structure In this section. 7. This was verified by an elastic-plastic computation of the uncracked structure and by the test results (see previous paragraph). A normal integration scheme was used. with shell elements (6 and 8 node elements) on the inner surfaces in order to apply internal pressure. We verified than the large displacement formulation gave better results than the small displacement one. it was assumed that its presence had no influence on the overall behaviour of the structure (i. 7) which was developed for fracture mechanics analyses (global approach).

For the third test. the crack initiation is either correctly predicted or predicted in a slightly conservative manner using the mean value of the parameter J0. A new method was developed to calculate the energy release rate G in accordance with the  large deformation  option (Mialon and Visse. Crack initiation e6aluation The crack initiation evaluation was made by comparing the material’s toughness J0. Se ´ me ´ te ´ et al. For the three tests. the softening effect due to the elbow ovalization.2. Those results are presented in Fig. G was computed with the small displacement formulation. Crack growth analysis (SEM3 test-circumferential crack). the application of internal pressure as a  following load  provided another improvement for the simulation of the overall behaviour. 1992). 1981. Finally. as shown in Fig. 10. Wadier and Malak. The 3-D energy release rate was calculated using the G -Theta method developed by EDF (Destuynder et al. For the SEM3 test. The dead weight of the moment arm was also included in the loading conditions because its contribution to the moment is significant (about 160 kN. However. 8 for the longitudinal and the circumferential . 4. the agreement between experimental and numerical load versus ram displacement curves is excellent for the three tests. 1989.2 to the curve of the energy release rate G at the deepest point of the crack versus load. in this particular case. the internal pressure and the flat heads limited the elbow ovalization so that the small displacement formulation also gave very good results. 1995) and was used for the first two test computations. 3. Suo and Combescure. / Nuclear Engineering and Design 197 (2000) 155–168 165 Fig.P.m)..3.

166 P. the material J -R curves were extrapolated beyond 3 mm (which is the validity limit for CT25 specimens) with the power-law fit determined within the validity range. The results of the crack growth evaluation are shown in Fig. For each material. For the first three cases in Table 6. Se ´ me ´ te ´ et al. . Fig. (1996). 3. Crack growth e6aluation This evaluation is made on a J -Da diagram by comparing the applied-J curves (obtained from the calculations) with the material’s J -R curve (obtained on CT specimens). For each crack. Formally. 8 that. 9 for the longitudinal notch of the SEM3 test and in Fig. three curves were assessed: a mean curve. the figures relative to SEM1 and SEM2 can be found in Le Delliou et al. G versus applied load curves for several crack depths (SEM3 test-longitudinal crack). the final depth measured at the end of the test and an intermediate value (Table 6). until the initiation load is reached. 10 for the circumferential one. 11. this calculation with the initial crack size is no longer valid once the crack begins to propagate. The applied-J curves were obtained for different levels of loading from calculations of the parameter G made on meshes with different crack depths. the figures relative to SEM1 and SEM2 can be found in Le Delliou et al.4. we choose. the G value is higher for the longitudinal crack than for the circumferential one. / Nuclear Engineering and Design 197 (2000) 155–168 cracks of SEM3 test. This is quite consistent with the test results where the longitudinal crack actually initiated first. (1996). a minimun one ( − 2| ) and a maximum one ( + 2| ). besides the initial depth. and the curves should not be used beyond this point. We can see on Fig.

However. it is well known that this curve shows a significant geometry dependence of crack growth resistance.6 mm is obtained for a calculated load Fy (491 kN) lower than the experimental one (531 kN). even with the maximum curve. which is associated with different levels of crack tip constraint under large scale yielding conditions. 12. / Nuclear Engineering and Design 197 (2000) 155–168 167 Fig. the through-wall stress distribution being mostly a bending one for the SEM1 test (like in CT specimens) and a membrane one for the SEM2 test. for the circumferential notch. G versus applied load curves for several crack depths (SEM3 test-circumferential crack). As an example. Conclusions EDF performed three bending tests on large diameter aged cast duplex stainless steel elbows . the calculated Fy value is lower than the experimental one) with the mean J -R curve and agreed well with the maximum curve. For the SEM2 test. The reason of this difference between both tests is probably related to the constraint effect on fracture behaviour.P. considering the other uncertainties which come from the scatter of the raw J -R data and the power-law extrapolation of the J -R curve. For both notches of the SEM3 test. the numerical prediction was conservative (i. respectively for the longitudinal and the circumferential notch. The curves of G (at the deepest point) versus applied load for different crack depths are shown in Figs. Se ´ me ´ te ´ et al. These curves are in accordance with the test results where the longitudinal crack grew faster than the circumferential one.e. the J -R curve was often tacitly regarded as a material property. 4. This explanation by the constraint effect must remain qualitative. the prediction was conservative with the mean J -R curve. 11 and 12. For the SEM1 test. the final crack propagation of 2. In the past. the numerical prediction was very conservative.

J. Three reasons are proposed to explain this fact: the scatter of the J -R data. Owing to the good efficiency of the d-c electric potential method. Eng. O.. 1997. Masson.. X. References Chopra. 1996. May 26 – 30.. 1995. Sur une interpre ´ tation mathe ´ matique de l’inte ´ grale de Rice en the ´ orie de la rupture fragile. Mialon. Ignaccolo.. 13 – 18.. Nucl. 135. 1. S. S. Nucl. Appl. On the application of G u method and its comparison with De Lorenzi’s approach. Code – Aster® Version 3. I. pp. / Nuclear Engineering and Design 197 (2000) 155–168 containing either one or two semielliptical notches on the outer surface of the flank. 207 – 224..5 MPa. 1990. Des... In: Proc. et al.Z. This document is available in English and French at the Internet Address http://www. Wadier. 3.. Y. The finite element analysis of the three tests simulated accurately the overall behaviour of tested structure and gave in each case a good prediction of the crack initiation. Overview of the research program conducted by EDF. In: Proc. Eripret.. Billet. Thermal ageing of cast duplex stainless steel primary components. . 1981. Anaheim Vol.C. 1996. 1997. The first two elbows were loaded under in-plane closure bending while the third elbow was loaded by an in-plane closure bending moment combined with an internal pressure of 15. Ph. vol. Se ´ me ´ te ´ et al. O. 70 – 87. An energy release rate formulation for large deformation problems (in French). The crack growth analysis was conservative. the crack initiation was accurately detected and the final crack extension was correctly estimated. pp. 1996 PVP Division Conference. In: Proc. Visse.P. E.. L. 69 – 77. Information Brochure in French. pp. 1985. 86. A. Nice. compared with the experimental results.. Study of cast duplex stainless steel elbow under closure bending. EDF Report HI-75/96/073. Meth. Eng. Ch. 1992.. P. . the extrapolation of J -R curves from standard CT specimens to longer crack extension.. G. 1570 – 1575. Djaoua. Massoud.. The Theta method applied to the analysis of 3-D elastic-plastic cracked bodies. Aging degradation of cast stainless steel: status and program. Malak. Combescure. G. 1995. Ayrault. Proc.168 P. ECF8 3. Math.. 117 – 123. Suo. P. P. 1989. M....K. P. Sci. In: Proc.. SMIRT 10 1989. J. Se ´ me ´ te ´ . Vautier. Vol. These tests and their detailed analyses contribute to the integrity assessment of cast duplex stainless steel components from the primary circuit of PWR units. Destuynder.. Icone-5. Fracture mechanics analysis of cast duplex stainless steel elbows containing a surface crack. Jayet-Gendrot.edf. Le Delliou. Des. 89 – 94. 323. 2e ` me Colloque National en Calcul des Structures. Giens. Le Delliou. and a constraint effect in the The tests showed that it was possible to obtain a large amount of stable crack growth (up to 13 mm) despite the low toughness properties of the aged material.