O - i . g f e . o l ! UT, ¿ A - Bib l ic seca | 7 s . ' u.,VO FUBICACIONES Int. J. Pres. Ves. & Piping 23 (1986) 149-161 Some Aspects of Thermal Fatigue in Stainless Steel A. F. lo r io a n d J. C . C resp i D ep ar ta m e n to Materiales, Gerenc ia de Desarrollo , Com isión Naciona l de Energía Atóm ica , 1429 Buenos Aires, Argentina (Received: 10 O c tober , 1985) A B S T R A C T This paper is concerned w ith the ana lysis o f fa ilu re s in the m o d era to r circu it branch p ip in g o f the A T U C H A -1 p re ssu r ized heavy w ater reactor ( P H W R ), which is m ade o f a u sten itic s te e l to D I N 1-4550 specifica tion (s im ila r to A I S 1 347). These fa i lu r e s arc considered to result f r o m th e rm a l fa tig u e p rocesses induced by flu c tu a tio n s in a zone where stra tified tem p era tu re layers occurred , the flu c tu a tio n s being associa ted w ith varia tions in m o d era to r flow . The first section eva lua tes the p o ss ib il ity o f cra ck in g due to th erm a l fa tig u e phenom ena and concludes tha t under service cond itions a crack m a y in itia te and g row th rough 7 m m th ickn ess o f the branch p ipe . In la b o ra to ry th e rm a l fa tig u e tes ts tha t s im u la ted the th erm om echan ica l cond itions fo r such a co m p o n en t , the n u m b er o f cycles requ ired to in itia te a th erm a l fa tig u e crack in a n o tch ed m o d ified s ta n d a rd fa tig u e specim en was abou t 10*. Th is value m a y be used to g ive a conserva tive p red ic tio n o f the num ber o f th e rm a l cyc les f o r cra ck in itia tio n in ac tu a l branch p ipes , including those sub jec t to the co ld p lu g cond ition w hich is p ro d u ced in som e em ergency shu t-dow n an d valve te s tin g s itua tions. It was also d em o n s tra ted th a t beyo n d a crack dep th o f 7 m m stress corrosion cra ck in g is the m a in p ro cess in fu r th e r crack p ro p a g a tio n . The relevance o f th is p red ic tio n is co n firm ed by m icro fra c to g ra p h ic observa tions, since the b r ittle na tu re o f the fr a c tu r e su rfaces under service cond itions appears very d iffe ren t fr o m the tra nsgranu la r ductile str ia tio n s fo u n d in bo th th e rm a l an d m echan ica l fa tig u e test specim ens as a result o f in tera c tin g en v iro n m en ta l e ffects. 149 Int. J. Pres. Ves. & Piping 0308-0161 86 $03-50 < Elsevier Applied Science Publishers Ltd. England, 1986. Printed in G rea t Brita in t N O M E N C L A T U R E a Crack length b Tube wall thickness E Young’s modulus K' Cyclic strength coefficient K ei( Effective stress intensity factor ^max Stress intensity factor at peak stress Kt Theoretical stress concentration factor K e Strain concentration factor ( =Afi/Aen) K a Stress concentration factor ( = Act/ActJ m Empirically determined exponent dependent upon material and temperature n' Cyclic strain-hardening exponent N ( N um ber o f cycles to fracture R Stress ratio a Coefficient o f linear thermal expansion AÂ lh Threshold stress intensity factor range A T Tem perature difference across the wall thickness Ae Cyclic strain range Aen Nominal cyclic strain range Act Cyclic stress range Aern Nominal cyclic stress range v Poisson’s ratio £ Ratio o f inner to outer radii ( = R .JR 0) CTmax Peak stress ct, Thermal stress CTy Yield stress IN T R O D U C T IO N Since the first failures occurred in 1977 in the ATUCHA-1 PH W R m odera tor circuit branch piping, which is made o f stainless steel to DIN 1.4550 specification (similar to AISI 347), it has been shown that mechanical fatigue was not the only mechanism that caused crack p ro p ag a tio n .1 ~3 F rom those early investigations it was concluded that the pressurized heavy water environment itself also had some influence on crack propagation, since there was agreement between the nature of the 150 A. F. Iorio, J. C. Crespi Thermal fatigue in stainless steel 151 fracture surfaces examined and those reported in some other relevant works.45 More recent experimental w ork6 has confirmed the presence of fluctuations in a coolant zone where stratified temperature layers occurred, these fluctuations being associated with variation in m oderator flow, so allowing thermal fatigue to be induced. The stratification process is due to the ingress of relatively cold heavy water at 140 °C into the primary cooling circuit, which is at 280 °C, as shown in Fig. 1. The cold water is derived from the m odera tor circuit and a cold plug condition is produced in some emergency shut-down and periodic valve testing situations. QHOI Pump Th e rm o l s t r a t i f i c a t i o n ( c r a c k e d zon e) Fig. 1. M o d e ra to r circuit b ranch piping scheme showing the stratified layer location. Although a considerable am ount of work on thermal fatigue has been done in the past and part o f it shows that the appearance of the fracture surface is transgranular brittle .7 there is not much published evidence on this aspect. The first section of the present paper evaluates the possibility of cracking due to thermal fatigue phenomena. The following section describes a test which simulated the thermomechanical conditions to which the m oderator circuit branch piping is subject, in order to clarify whether an environmental effect interacts with thermal fatigue phenomena. The resulting fracture surfaces under different rupture conditions are compared. 152 A. F. lorio. J. C. Crcspi POSSIBILITY O F T H E R M A L F A T IG U E C R A C K IN G As indicated above, the thermal fatigue process is considered to originate in the fluctuation of a stratified coolant layer; the related conditions are that the temperature o f the inner surface o f the pipe is 280 °C and every 20 s6 a ‘strip’ o f water at 140°C moves cyclically on this surface. This implies a quasi-stationary process governing the phenomenon in which a thermal shock condition appears to be of minor relevance, involving only a very thin layer o f material thickness; nevertheless, such a thermal shock effect is o f particular concern for a s tationary regime because stresses produced by this effect reinforce stresses coming from fluctuations o f the stratified layer. S tationary thermal stresses were given by Zudans and Y en8 as a function of pipe radius, r (R- < > < R J , where the internal surface hoop stresses are tensile and the external surface hoop stresses are compressive: E a A T 2(1 — v)ln £ 1 + In R, + c 2 In c 1 - c 2 ( 1) These pseudo-elastic stresses, which also include the hoop stresses due to pressure through the wall thickness, are shown in Fig. 2. The solid lines represent these stresses once they have been transformed into Ri Ro [mm] Fig. 2. Stress levels th rough the pipe wall thickness due to therm al fatigue plus ho o p stresses. Thermal fatigue in stainless steel 153 elastic plastic stresses. As can be seen in Fig. 2, up to 7 mm depth from the internal radius the stresses are higher than the material yield stress. Using the values of these cumulative stresses in conjunction with the fatigue design curves from A SM E Code, Section III, it can be predicted that a fatigue crack will develop in about 104 cycles. Considering that in low-cycle fatigue the num ber of cycles for crack initiation is about 10 20 % of the number of cycles for fracture, then in the present case such initiation should occur after 1000-2000 cycles. The fatigue crack propagation rate is governed by the effective stress intensity factor, Ke„, which is given by the following expression:9 Ke(i = K mJ \ - R r (2) where K mdX has been estimated from the following re la tionship:10 KmM= ( ° m.MsJna)F(alb\ R J R 0) (3) where F(a/b: R J R 0) is a non-dimensional function. Figure 3 shows the profile o f K eff through the wall thickness, calculated with pseudo-elastic stresses after eqn (1). An elastic-plastic criterion would be more appropriate for the first 6 -7 mm of wall thickness but it was considered that the elastic solution used should bring a conservative result. The inflection point shown in Fig. 3 indicates a change in the crack growth process, since at that point the cyclic stress frequency varies from seconds (due to the thermal fatigue phenomenon) to months (due to R j R0 (m m ) Fig. 3. Kc„ profile th rough the pipe wall thickness. 154 A. F. lo rio , J. C. Crespi variation o f internal pressure produced by reactor start-ups and shu t­ downs). It follows that at such a point the crack will not grow by a fatigue process. The conditions are then especially set up for stress corrosion cracking to drive the crack growth process because the length of the crack may produce a stagnant environment with a high stress intensity factor at the crack tip. E X P E R IM E N T A L P R O C E D U R E A notched modified s tandard fatigue specimen was designed, which is shown in Fig. 4. Specimens were machined from the failed wall pipe, the specimen axis following the longitudinal axis of the tube. já 5/ 8 " UNF 4 , 6 5 S e c tio n A Fig. 4. Notched modified s tandard fatigue specimen. The chemical composition of the steel agreed with the standard specification for D IN 1.4550 (AISI 347) and its microstructure in the transverse section is shown in Fig. 5. The distribution o f the inclusions was normal but a few large inclusions (up to 103 /mi long) were aligned in the longitudinal direction o f the tube. The temperature cycle was achieved by means of a 10 kW (450 kHz) induction heating furnace, using 30 s for increasing the tem perature from 100°C to 300°C and 17s for cooling it down to 100°C. The total thermal cycle was accomplished in 47 s. The cooling process was carried out by means of copper blocks fixed to the specimen with an internal injection o f cold water flow. Simultaneously Thermal fa tigue in stainless steel 155 Kig. 5. D IN 1.4550: austenitic stainless steel m icrostructure with partial recrystall i­ zat ion ( x 100). a jet of cold air was applied in order to accelerate the cooling. Before the heating cycle was reinstated the cooling was stopped and the copper cham ber emptied. The temperature loop system was closed by a 0-3 mm diameter Chromel Alumel thermocouple located directly at the specimen notch root. By keeping the stroke constant, stiff restraints were imposed on the specimen by means of a servohydraulic universal testing machine Fig. 6. Experimental therm om echanical cycle imposed on the specimen. 156 A. F. lo rio , J. C. Crespi (lOOkN). The stress resulting from this restraint was increased by an additional m onotonic tensile stress. The applied net stress at 100°C was then 210 M Pa, which was the stress estimated to cause crack initiation in approximately 103 cycles. The complete thermomechanical cycle is shown in Fig. 6. Due to the stress relaxation dependence11 a semi-continuous adjustment was made to the applied load in order to maintain a constant stress on the specimen during the complete cycle. In order to predict the num ber o f cycles required to initiate a crack about 0-2 mm long it is necessary to know the stress and strain fields at the notch tip. This requires the local strain at the notch tip to be related to the cyclic behaviour of a specimen without notch which is subject to the same cyclic s tra in .12 It is assumed that the num ber of cycles for crack initiation is the same for similar strain values in both specimen types. The notch tip material can be characterized by the stationary cyclic stress -strain response. This relationship was derived from data obtained by Kanazawa and Y osh ida13 for stainless steel type AISI 347 at 4 50°C and strain rate 0 = 6-7 x 10 3 s “ 1: where K' and n' had the values 6000 M Pa and 0-542 respectively. Furtherm ore, the Neuber ru le14 was considered to apply in order to obtain the relationship between the notch tip stress and strain concentration factors: C R A C K IN IT IA T IO N P R E D IC T IO N Act = K'{As)n (4) (5) ( 6 ) Combining eqns (4) and (6): (7 ) Thermal fa tigue in stainless steel 157 With K' = 6000 M Pa, «' = 0-542, KT = 2-74, A