Available online at wwwsciencedirect.com . Nuclear sorenoe Ghoineor Engineering and Design ‘Nuclear Enginering and Design 235 (2005) 575-596 \worw.clseviersomTocstelnucengdes Hydro-thermal-mechanical analysis of thermal fatigue in a mixing tee S. Chapuliot®, C. Gourdin®*, T. Payen®, J.P, Magnaud®, A, Monavon! * DEN/DMBSISEMMLISN, CBA Saclay, F-91191 GifsSu-Tverte Cedex, France ® IRSN/DSRISAMS, BP 17 F-92262 Fontenay auc Roses Cedex, France © DEN/DM2S/SFMELTMR, CEA Saclay, F-91191 Gif Sur Yeete Cedex, France 4 UnverstéPlere et Marte Cure, 4 Place Jusiew 75005 Parts, France Received 2 Apil 2004; received in revised form 30 August 2004; accepted 20 September 2004 Abstract This paper covers work carried out by the CEA to study the mechanisms leading to cracking of piping as @ result of thermal Toading in flow mixing zones. The main goal of the work is to analyse, by calculation, the thermal loading caused by turbulent ‘mixing in tees and to understand the mechanism of initiation and propagation of cracks in such components. ‘This work is supported by IRSN. This thermal fatigue phenomenon is still no fully understood. One ofthe main obstacles to its understanding resides in the multi-domain nature of the loading and associated damage, involving three complementary seiemtiie disciplines thermal-hydraulic field, thermo-mechanical field and materials science This paper describes the approach adopted by the CEA to establish natural mechanisms (turbulence, pulsing and instability) which might be the cause of any substantial thermo-mechanical loading in the piping, Although turbulence may be the cause ofthe thermal stripping (presence of high-frequency thermal fluctuations on the inner surface of the component), it cannot alone explain the propagation of deep cracks. The main reason is the “high-pass filter” effect of convection, The wall cannot be subjected to convection-related thermal fluctuations and frequencies less than the inverse of the turbulence transit ime. A straightforward frequency-based analysis of the loading, carried out as a first stage, made it possible to establish the limits of the loading ereated by these high-frequency events, However, turbulence can give rise to flow instability (such as pulsing) of lower frequency. But this eannot explain everything, The geometry upstream of the tee, particularly the sequence of straight sections and bends can, in certain cases, damp the pulses or greatly amplify them, The use of suitable thermal-hydraulic modelling is discussed in the second part ofthis article The final result of the thermo-hydre-mechanical link-up on application to the complex 3D geometry of the Civaux unit 1 case (which includes a mixing te, bends and straight sections) enabled the observations made inthis plant ease to be highlighted and correlated One of the originalities ofthis study isto carry out the overall analysis (thermal-hyraulie and thermo-mechanical) with a single computer code, the CAST3M code developed by the CEA, © 2004 Elsevier B V. All rights reserved * Corresponding author. Tel; +33 169 0845 86; fx +53 169008 87 84 F-mail address: gourding@semt2 sms cea fr (C. Gaudin) (0028-5493/8 ~see from matter © 2004 Elsevier B.V. Allright reserved oi'10 1016) nucengdes 2004 09.011576 S. Chapulio etal. / Nuclear Engineering and Design 235 (2005) 375-596 1. Introduction ‘The problem of fatigue in mixing areas, which is the subject of this paper, occurs in pipes where flows at different temperature mix. If the rates of circula- tion of the fluids are high, the mixing is turbulent and results in temperature fluctuations that can be local or global, depending on the form of the flow. When such thermal fluctuations are transmitted to the wall of the structure, a mixing tee ot a simple branch, the strain variations result in fatigue damage and erack- ing, The problem is complex and involves four scientific disciplines: «© thermal-hydraulic field, as concerns the thermal loading (turbulent mixing of two fluids), « thermal field, as concems the initial interface of the problem, as a result of heat exchange between the fluid and the wall, © mechanical field, as concerns the response of the structure to the complex thermal loading tothe inner surface, « the science of materials, as concerns the strength of the materials making up the structure subjected to the loading Each domain has its own difficulties and scientific limitations. 11, In the hydraulic field ‘There are difficulties with the experimental qualifi- cation and the modelling of turbulent mixing (temporal and spatial multi-scale phenomena): «© as concerns modelling, progress has been made with the LES and VLES models, but these are still in the development and validation stage, « the qualification of these models is extremely diffi- cult as a result of measurement difficulties and un- certainty about the validation of the numerical and experimental results, «the trend towards to build outsized numerical models that current computers can barely handle but which remain incapable of modelling the entire range of fluctuations. 1.2, Inthe thermal field ‘There are still difficulties in measuring and mod- elling the heat exchange between the uid and the wall In addition, there are theoretical problems associated with lack of understanding of the boundary layer phe- 1.3, Inthe mechanical field ‘There are no real theoretical problems, but issues associated with pipelines are not well understood. The ‘mechanical aspect relates primarily to the response of | the structure to the thermal loading, and itis therefore necessary to progress beyond the two previous steps to be able to make a precise evaluation of the mechanical loading. In view of the problems mentioned earlier, the matter is generally eluded and the response of the structure is not considered, the problem simply being analysed through the thickness (ID analysis) for an idealized thermal loading (considered to represent the worst case). 1.4, Inthe field of the science of materials ‘The resistance of the materials to cyclic thermal loading is not well understood either. ‘The material is locally subjected to a load consisting ofa large number of multi-frequency multisaxial cycles, under constant average tensile stress and/orstrain. Fortis type ofload, there 1s actually very little data on material fatigue re- sistance (resistance to cracking), or the evaluation of | damage, kinetic initiation and crack propagation. The various difficulties associated with theory, ex- perimenting and modelling result in the fatigue prob- lems of the mixing areas observed in the facilities are not being well understood: ‘» The appearance of a network of cracks is known but not properly controlled. The phenomenon of thermal stripping has long been known and analysed, partic- ularly for the fast breeder reactor structures vulner- able to this type of damage (Gelineau et al., 199), Its still not fully understood how to determine the conditions of initiation of elephant skin crack from a temperature load (temperature difference between two fluids) as concerns location, position, and vari- ation over time.'S-Chapuliot ta, / Nuclear Engineering and Design 235 (2005) $75-596 sm ‘« Propagation through the wall of macro-cracks re= sulting from the thermal stripping and the pres- ‘ence of discontinuities (discontinuities of geometry and/or material, such as a weld) is far rarer (in most ‘cases, the cracks remain atthe surface) and this phe- ‘nomenon is even less well understood. The problems ‘encountered in a mixing tee in the residual heat re- ‘moval system of the Civaux unit 1 (Faidy etal ,2000) is an example of this type of damage that is not ex- plained by the 1D analyses that cannot demonstrate through-wall penetration, It is to clarify this point, particularly by research into loading that can result in through-wall cracks, that mechanical studies have been initiated to gain a fuller understanding of the problem. The goal of these stud- ies isto supply the information that cannot be provided via ID analysis by seeking to determine the global re- sponse of the structure to the thermal loading. These are of the 3D type and are necessarily dependent on precision and knowledge of the thermal loading. As investigation ofthe thermal load imposed on the struc ture remains difficult, a special approach was devel- oped, which is described in this paper. Furthermore, this work was supported by IRSN. 2. The approach ‘The first stage of our approach consisted of a math- ematical analysis ofthe thermal loading, followed by a mechanical analysis of the associated mechanical load- ing, This was based on experimental temperature read- ings with the most realistic smoothing possible between them. This thermal loading obviously had little phys- ical reality as it was a mathematical construct, but it offered the advantage of being rapidly setting up and independent of successful thermal-hydraulic calculae tion or link-up between the different computer codes Furthermore, the parameters are more easily set, mak- ing it possible to more rapidly begin the study of the local and global responses of the structure. ‘ond stage essentially consisted of acomplete hydro-thermo-mechanical (HTM) numerical analysis carried out with the CAST3M computer code. This used the existing computing resources (LES simula- tion) and was thus suiject to their limitations and un- certainties. The main advantage of this approach was that it provided an example of mixing fluctuations in a tee, enabling subsequent deduction of the mechanical response of the structure (spatial distribution, through. the thickness, average stress, frequency response, etc). Although it may have been approximate, the solution offered the advantage of providing initial results for orientating the subsequent work. In addition, this ap- proach made it possible to implement the entire cal- culation sequence, to identify the problem areas and to initiate study of the tools and requirements for the interfaces between the different computing resources (particularly between the thermal-hydraulic caleula- ‘tons and the thermal-mechanical calculations) The third stage consisted in applying and using the knowledge acquired to a ease encountered in the field The incident that occurred in the residual heat removal system of the Civaux unit 1 constituted an excellent example of the phenomena associated with thermal fa- tigue, 3. Description of the Civaux I case Following the discovery of a leak due to a longitu- dinal crack at the outer edge of bend in a mixing zone of the residual heat removal system of the Civaux NPP. Unit | in 1998, metallurgical expertises were carried ut and highlighted that the origin of this degradation phenomenon was the cracking by thermal fatigue, The incident that occurred in this system was caused by thermal fatigue associated with fiuctuations inthe tem- perature of the fluid at the mixing tee, However, the mechanisms that caused the transition from crack ini- tiation to the development of a set of cracks of a sig- nificant depth in a relatively short length of time have not been precisely established The overall system is shown in Fig. 1. The section of interest is that beginning at Valve 101 VP (hot leg) and Valve 071 VP (cold leg) up to the tee and downstream oft (framed). The system is ata pressure of 36 bar. The hot leg contains water at 180°C with a flow rate of Qc and the cold leg contains water at 20°C with a flow rat of Qr. The total flow rate Qr-+ Qe is 550m*/h. The flow ratio Qrl(Qr + Qs), which was analysed, corresponds t0 20%, ‘A number of analyses were made to understand how, under the effect of thermal fluctuation loading, cracks could appear and develop through the wall of a bendsm S. Chapultot etal. / Nuclear Engineering and Design 235 (2005) 375-596 Fig 1, Overall view of the loop, in such a short length of time (1500). Metallurgical ‘examination also revealed substantial cracks and net- works of small thermal fatigue cracks in the vicinity of the welds, in the absence of fabrication defects. The damage was mainly located (see Fig. 1) (a) on the outside of the bend downstream of the mix- ing tee, (b) in the mixing tee, (©) at the bottoms of the weld beads and at the circum ferential weld roots, (@) in the straight sections of piping (after longer pe- riods of operation). 4. Frequential study of propagation 4.1. Introduction Different types of thermal-hydraulic loading give rise to the appearance of a network of cracks and to the propagation of a macroscopic crack. At high frequen- cies, temperature penetration is low as the conduction phenomenon filters out the temperature variations and they remain localized at the surface of the component. ‘The region affected is therefore smaller and mechanical loading outside the area is too small to cause significant crack propagation. However, at very low frequencies, the penetration of a thermal shock is greater, which, gives rise to greater stress variations at a depth in the ‘component. This type of low frequency loading can ‘cause rapid propagation (in terms of number of eycles) ‘of a macroscopic crack. The goal of the frequential study was to determine ‘what frequency range should be taken into account in studying the major mechanisms involved in the rapid propagation of eracks in the mixing tee. 4.2, The analytical model and its subsequent results The analytical model was based on exact solution of a problem corresponding to the conduction of an infinite plate subjected to sinusoidal thermal loading, Allowance for and representation of the phenomenon of heat exchange between the fluid and the wall of the plate was added by Kasahara et al, (1999), A formal system for the mechanics of fracture was analysing by Jones and Lewis (1995). And the model was identified and applied it to the case encountered at Civaux and to enable study of propagation of a crack subjected to random loading, By investigating a wide range of frequencies extending from quasi static (0,001 Hz) to high (100 Hz), it was possible to narrow the range to be studied. Dividing the stress into different terms such as peak stress, membrane stress and bending stress made it possible to allow for different constraints, and boundary conditions (rigid or pin-jointed ete.) in a structure idealized by combining the three terms (see Fig. 2), For different locations in the mixing tee and for each type of stress, the analytical mode! could determine how stress varied as a function of the lrequeney of the load. ‘The results are shown in Figs. 3 and 4 for the caseSeg 2) ess Linearized sara | Fig 2, Suess breakdown z 5 a z i & Frequency (he ig. 3. Change in stress varistion asa function of frequoney atthe outside surfae of she bend (lation () in Fig, 1) of the outside of the bend and for the thick part of the mixing tee. For the thickness corresponding to the outlet bend (9.3mm), the peak and bending stress variations were at maximum at 0.04 and 0.4 Hz, respectively. And from. 10 Hz, the total stress variation thereafter represented only 22% of the maximum value obtained at 0,001 Hz ‘o_ Stress variation (MPa) Frequency Fig. 4 Change of stess variation asa funetion of frequency «case ofthe tee thick pat) (Location (b) in Pig. 1. 'S-Chapuliot ta, / Nuclear Engineering and Design 235 (2005) $75-596 3% In addition, the bending and membrane stresses cane celled each other out (see Fig. 3) For the thickness corresponding to the tee, the stress variations or frequencies were at a maximum at 0.004 Hz for bending and at 0.1 Hz for peak stress. Beyond 10 Hz, the total stress variation no longer rep- resented more than 20% of its maximum value (see Fig. 4), 4.3. Conclusions of the frequential analysis Rapid crack propagation necessitates substantial variation of membrane bending and peak stresses However, it was observed that there was a substantial reduction in the peak stress and null variations for the bending and membrane stresses at frequencies greater than 10 Hz. Tn the light of these results, the frequeney of 10 Hz, represented an upper limit for variations ofthe thermal stresses to be analysed to investigate the propagation issue. It was considered that phenomena occurring at high frequencies could not explain the rapid propaga- tion of the eracks observed in the former configuration of the mixing tee 5. Thermal-hydraulic analysi 5.1. Description of the system ‘The first mesh (RRA2C, see Fig. 5) was limited to the first two upstream bends. The second mesh. (RRAAC, see Fig. 6) included the three bends upstream of the hot leg between the tee and the valve as well as the bend downstream of the cold leg, The eflect of the valves on the flow was not taken into considera- tion, The fineness of the meshes was adapted in each case to obtain approximately the same number of nodes, (200,000), ‘Three cross-sections and eight points regularly dis- tributed over the circumference were chosen for mon itoring the water temperature over time (see Fig. 7) 5.2. Physical analysis 5.2.1. Flow regimes ‘The mean flow velocities (Uc, Ua) were 0.59 and 2.97 mfs, respectively, in the cold leg and the outlet80 ‘S Chapuliot eral. / Nuclear Engineering and Design 235 (2005) 575-596 section (combining the two flows). The correspond ing Reynolds numbers Re= UDI (where D was the pipe diameter and U was, respectively, Ue and Ua) ranged from 1.5 x 105 to 3.9 x 10°, The flow is un- steady and turbulent in the two legs, with statistically stationary boundary conditions. The buoyancy effects could be assessed by calculating a Richardson num- ber (Ri=gAoD/p(Uc)?) that varied between 0.9 and 2.6 x 10. The effects of natural convection were of ‘minor importance without being negligible at the point ‘of actual mixing in the tee. These were modelled using the Boussinesq approximation 5.2.2, Timescales Upstream of the tee (hot and cold legs), the straight tube lengths were not sufficiently long for a stable regime to be established (as they only corresponded to six diameters). The model of flow in a pipe could nevertheless be used to estimate the characteristies of the turbulence and the inlet of the tee 5.2.2.1. Twn over time. The turn over time is the longest period that can be associated with a turbulent som = i smh | sa me" | Fig. 5. Mesh of system (RRA2C), structure. It is of the order of T= L/Us where L is the characteristic size of the energetic turbulent structure in a duct and Us the friction velocity. L was taken to be equal to D. the largest size for a turbulent structure liable to pass through the duct. At the hot and cold flow intersection (in the tee), an energy structure could be associated with a temperature variation of the order of | the temperature difference between the two currents. In ‘ther words, it was considered that the two flows only mixed slightly at this level. ‘The timescale associated with these turn over times characterized the persistence time of a temperature fluctuation due to turbulence ‘To evaluate Us, standard friction laws in a duct, (Schlichting (1968, p. 412) and the following) were used. Here 10 = pUZ= 5 ApU? with A =0.316Re~* (Blasius). Fig. 6. Mesh of system (RAC),'S Chapuliot ta, / Nuclear Engineering and Design 235 (2005) $75-596 se Fig. 7. Cross-section instrumented Hence the relationship between Us and U(U,/U? = Als. The friction velocity Us thus characterized the level of the turbulent fluctuations and we used it to estimate the amplitude of the velocity fluctuation for the inlet boundary conditions 5.2.2.2. Transit time, At a given point, a set of tur bulent structures pass along the wall at a velocity of U. The order of magnitude of the transit time of the largest turbulent structure is: Fr = (LU). From this it is possible to deduce the relationship between transit time and turn over time (independent of the size of the turbulent structure): T/T =Us/U~ 0.05, The transit times are considerably shorter than the turn over times. The wall can thus experience the passage of turbulent structures as though they were frozen. ‘This is a con- ventional approximation in the case of high-velocity flow. The wall could not be influenced by convected tur- bulent structures of characteristic time greater than the transit time It would only see a small proportion of it, as convection acts as “high-pass” filter. The lower fre- quencies that the wall was exposed to and origi nate from turbulent structures corresponded to the in- verse of the transit times, Table I shows the esti- mates that could be made, The flow velocity distri- bution actually remained very heterogeneous in the outlet section (as will be seen). The frequencies as- sociated with the transit times ranged from 2.5 to 12H. Transit Frequeney Tum over Frequeney Ouletsection 008s 125Mz 16s O628Tkz Us=2.83 mis Cold eg 4s -25H2 Bs ORSHe Uc-O4T nissa S. Chapultot etal. / Nuclear Engineering and Design 235 (2005) 375-596 5.3. Analysis in greater depth It is possible to make this line of argument more ‘general by assuming a power law velocity profile: = (F)' wheren = and0.2R < y= R Toa ‘The ratio between the transit time in a region of length L(i¢. Ty=L/u) to the turn over time imposed by the av- ‘erage local strain rate (ie. Tx= Iu’, where 1/ = away) is equal to: Tr _ Lu’) _ nk Tu) and, accordingly, for the two times to be equal, y must be equal tom. As under steady-state flow assumptions, 5 ig. 10, Isothermal in the cross-sections (RRA2C) in °C Fig. 11 Flow iso-velocty (RRA2C) in mis'S Chapulio 1 a, / Nuclear Engineering and Design 235 (2005) $75-596 Fig. 14, Standard deviations Fig, 9). An initial explanation of this extremely high instability was provided by the iso-flow velocities of friction used to trace the flow in this isothermal sec~ tion (see Fig. 15). This plot showed the presence of twisting of the flow in the hot leg upstream of the tee, The excess flow on the outside of the bend reap- peared atthe inner profile ina position that was virtually symmetrical 5.7. Conclusions ‘The physical analysis enabled us to clarify the role that turbulence can play with respect to the thermal loading of the structure. Although it may have been the cause of the thermal stripping (high frequencies), it alone could not explain the crack propagation, The ‘main reason was the way in which convection acted as a highepass filter which prevented the wall from exp. riencing convected thermal fluctuations of frequencies greater than the inverse of the transit time (2.5 Hin the ullet section). These frequencies were still too high to explain the crack propagation. The calculations that we made indicated that bound- ary condition excitation with white noise triggered pulsation of relatively large amplitude. ‘The turbu- lence was therefore the probable natural origin of the low-frequeney instability in the decay heat removal system. However, the geometry of the upstream hot leg played a major role in either amplification or damping of the pulses. The RRA4C geometry, which was closest to the actual geometry (extending up to the valves), was the most destabilising, The sequence of three bends and straight sections resulted in substantial twisting of the flow at the inlet of the tee, ‘The RRA2C geometry, limited to the upstream bend at the hot leg inlet, actually substantially reduced the amplitude of the pulses at the tee Only thermal-mechanical analysis would make it possible to qualify how much of the damage related to cach of the three forms of loading 6. Linked-up analysis of the RRAIC calculation The goal of the hydro-thermo-mechanical link-up made with the RRA4C geometry was analysis of the areas in which cracks appeared, particularly the outside of the outlet bend for which no analysis could explainss S. Chapultot etal. / Nuclear Engineering and Design 235 (2005) 375-596 Fig. 15. Prietion the crack propagation which created a major leak ob- served in the field, 6.1. Description of the linked-up analysis Generation of the meshes of the structure was di- rectly based on the meshes used for the thermal- hydraulic calculations, the thickness of the tee being 30mm and the thickness of the bend and the pipes be- ing 9.3mm, The transition zone between the tee and the bend or the pipes was taken into account. The size of the first mesh in the tee was set at 0.75 mm, and the size of the mesh in the bend and the pipes was set at 0.28 mm. Thirteen elements were used in the thickness, with greater fineness towards the inner surface. The number of nodes for the entire structure was 44,562 and the number of elements was 40,768. The elements were of degree of interpolation one (cightenode cube of type PI). This discretisation enabled us to obtain all, the changes of the different variables at a frequency of 10Hz. Figs, 16 and 17 show the mesh structure for the linked-up thermal-mechanical calculations. ‘The material and the line modelled were considered to be thermally and mechanically linear. Properties of the material were those of an austenitic steel at room temperature. Thermal « linear conduction (A) of 14.7 Wim°C, ‘© specific heat (C,) of 480 J/kg/"C, ‘© density (p) of 7800 kg/m’ Mechanical ‘+ modulus of elasticity (E) of 177,000 MPa, ‘* Poisson's coefficient (v) of 0.3, ‘© coefficient of thermal expansion 164 x 10-5°C. of @ ‘The thermal loadings were derived from the RRASC thermal-hydraulic calculations. The physical time modelled corresponded to a sequence of the last 10's, and comprised 100 calculation steps. The fluid tem- perature maps were used as input data for the thermo= ‘mechanical calculations. ‘The heat transfer between the fluid and the structure took place without approxima- tion as the two meshes were created at the same time and the nodes on the inner surface of the structure coin- cided with those of the fluid model. The coeflicient of | heat exchange between the fluid and the inner wall was taken to be constant in the entire temperature (from 20'S-Chapuliot ta, / Nuclear Engineering and Design 235 (2005) 575-596 59 Fig. 16. Thermo-mechanical mesh of the structre to 180°C) and time range. Also, it was assumed that Displacement of the cold section was prevented ‘the structure was ideally thermally insulated onthe out- (lower bend inlet). This. section being the far- side and that there was no exchange of heat with the thest from the mixing zone, its influence would be exterior (null-flux condition) small The mechanical boundary conditions were set so. @ Rotation was prevented but not of displacement of that: the outlet and hot inlet sections, Fig. 17. Detail of mesh (transition zane),590 'S Chapuliot etal. / Nuclear Engineering and Design 235 (2005) 575-596 Establishing these conditions was a delicate matter as it was difficult to determine the actual stiffness of | the upstream and downstream lines 6.2, Determination of the temperature fields in the structure 6.2.1. Static or average calculations ‘The average temperature in the structure was deter- ‘mined by imposing the average temperature map for the fluid. This average temperature was imposed as the initial temperature for the transient thermal calculation. 6.2.2. Transient analysis ‘The thermal calculation was cartied out in five sue- cessive 10 sequences to obtain a stabilised response for the structure, The temperature variation atthe inner surface is shown in Fig. 18. Itcan be seen that there is, a broad zone with a high AT, of around 119°C. This. value is substantial and represents 75% of the maxi- mum AT (Tz —1j). This zone extends from the outer surface to the inner side of the mixing tee opposite of | the hot inlet. This zone of extreme temperature varia- tion encompasses the tee to bend transition zone and corresponds to the oscillating motion of the cold front ‘observed in the average calculations Furthermore, in a small region downstream of the mixing tee a the intersection of the hot and cold pipes, there are substantial temperature variations, but at high frequency and extremely localised As concems the outside of the bend, temperature variations of around 50 °C also occur, as well as signif icant temperature variations (A7> 50°C) in the outlet tube. Fig. [8 Temperatre variation atthe inner surfice ofthe structure (°C).'S-Chapuliot ta, / Nuclear Engineering and Design 235 (2005) $75-596 59 ~N Fig. 19, Average equivalent sees atthe inner surface ofthe structure (MP), 6.3. Determination of the stress fields in the structure 6.3.1. Average calculations (On the basis of the average temperature determined in the structure, as indicated inthe previous section, the average thermal stresses were determined by a thermo- mechanical calculation, Fig. 19 shows the equivalent stress distribution maps (in the Von Mises sense) at the inner surface, The maximum value of the equivalent ‘Von Mises stress reached 358 MPa, The tee-to-bend transition zone and the inside of the bend were also places where these high stresses occurred, On the out- side of the bend, the average stress was very low. 63.2. Fluctuating calculations 63.2.1. Background on determination of equivalent stress variation. An important stage of analysis of the mechanical state consisted in determining the value of the equivalent variation of the stress in the Von Mises sense at all points in the structure. An itera- tive procedure was used for obtaining the equivalent variation. Determining the maximum variations of oq in- volved calculation of the stress variation between two instants, then determination of the maximum for the entire loading sequence. Between the two instants f and f (or calculation steps / and ), the variation of the equivalent stress was determined as follows 1 Aoll, = a[(Aon — Aayy)? + (Aoyy — o.:)? v2 2 + (Moe: — Ady)? + 6(Ao?, — Ac +A02,)]'? with Ae, = Aol, — Ao},, Aoy = Aoi, — Adj, ete The maximum values of all the possible pairs of calculation steps was therefore (for n calculated steps): AoMAX abet omy 2 MAX (Ao?!sn 'S Chapuliot etal. / Nuclear Engineering and Design 235 (2005) 575-596 Note: AAs the maxima did not appear at the same instants, the resulting field did not correspond toa state of equilibrium. Neither could it be used to determine the spatial averages (bending or membrane stress for instance). Fig. 20 shows the distribution of the variation of the ‘equivalent stress in the Von Mises sense at the inner surface for the fourebend configuration. The maximum value of the equivalent stress variation was 467 MPa ‘The area where the variation was the greatest was located in the tongue extending from the inside and ‘extending up to the tee on the side opposite the hot fluid inlet. Other zones were distinguished where the ‘equivalent stress variation was significant, particularly fon the outside of the bend with values of 180 to 225 MPa, as well as at the outlet of the straight sec- tion downstream of the mixing tee. It was noted that there were values close to 100 MPa on the inside of the bend. 67° 27° 64. Mechanical post-processing 64.1. The case of the outside of the bend ‘The temporal change of two circumferential and longitudinal stress points is shown in Fig. 21. Counting of the cycles by the Rainflow method (Norme AFNOR, 1996) made it possible to show three major eycles for the point located at 27° (values of major eycles: 215, 168 and 147 MPa) and two major eycles forthe point lo- cated at 67° (values of major eycles: 225 and 140 MPa). ‘The stresses in the bend were mainly peak stresses (stress variations of 150 MPa). The variations of bend ing stress were around 70 MPa and the variations of ‘membrane stress were less than 10 MPa It was therefore noted that there was substantial bending. However, the level reached was not normally sufficient to cause crack initiation, but allowance for the non-lush weld present in the field would introduce. a stress concentration coefficient of up to 1.5. Fig. 20, Bquivelent stess variation a he inner surface ofthe structure (MPa).2250 'S Chapulio 1 a, / Nuclear Engineering and Design 235 (2005) $75-596 593 750 l\ Time (s) Fig 21. Variation ofthe circumferential stress at two points on the outlet bend 64.2. The case of the transition zone ‘The tee-to-bend transition zone was the most highly loaded one. The level of variation reached 570 MPa in the 10's sequence for both longitudinal and circumfer ential tress. The stress state was equi-biaxial. Counting, the eycles with the Rainflow method made it possible to extract two important major cycles of 489 and 327 MPa, as well as four cycles extending from 140 to 100 MPa In addition, the average stress reached significant val- ues (up to 350 MPa) at a number of points. The total, membrane and bending stress variations were, respec tively, $36 MPa, 7 and 73 MPa (circumferential) and. 527 MPa, of and 41 MPa (longitudinal). Peak stresses were therefore predominant Note: The stresses obtained in the transition zone were obtained with a very coarse representation of the transition between the tee (thick) and the bend (thin) Finer discretisation of the thickness change would have resulted in higher values, particularly atthe level of the thickness change on the inner surface, Once again, this level of stress could therefore explain the presence of | cracking, 64.3. The case of the outlet pipe Variations of equivalent stress corresponded to the 250 MPa sequence. The area in which the variations of | the values were significant was located two diameters downstream of the bend, in the straight section. ‘The extent of the loaded zone was therefore correctly rep- resented, However, the level was too low to explain the cracking. 6.5. Study of the propagation 65.1. Methodology (Drubay et al., 2002) ‘The material of which the bend and the mixing tee were made is 304. steel, with a Paris law of the fol- lowing form da He = claKeal Cand nare material coefficients as defined in document RCC-MR (Drubay etal, 2002) equating to 7.5 x 10-1 for C and 4 for n ‘The effective stress intensity factor was determined by caleulating AX; and function q(R) such that: AKar=q(R)AK, Function 4(R) was given by the following formula 1-05R IER < Othereforeq andifR > Otheng94 S. Chapultot etal. / Nuclear Engineering and Design 235 (2005) 375-596 where pe Kain Kas Stress intensity factor Ky was determined by analy sis ofthe stress profile through the thickness of the com- ponent. For each point treated, the stress profile in the thickness was derived for the duration of the sequence (105). By identification, at each instant of calculation, the values of coefficients ao, o1, a2, 03 and as were determined by smoothing the stress profile through the thickness with a fourth degree polynomial. ‘Then the influence coefficient values (io, /1, ia, is and ig) were used, being tabulated as a function of the structure and the defect (Drubay et al., 2002), the stress intensity factor being expressed as follows: (oo ean (2) ven 2) #0) 4") ie + oui(2)') va where “a” comesponds tothe depth ofthe defect and D"to the thcknes ofthe component AK; canbe determined by counting the eels with the Ranviow metho, for msance 65.2. Application to the case with four bends 65.2.1. The case of the outside of the bend. On the basis of the circumferential stress, the stress intensity factor was determined. Then, function q(R) was de- termined, depending on load ratio R. For each crack advance inerement, the corresponding number of ey cles was obtained, then, by integration, the number of | eycles to progress from a crack of afh=0.1 to alh=0.8 (6.92 mm), i. 372.720. By considering the two major cycles per sequence of 10 (see Fig. 21), a propagation time of $17h was obtained. The influence funetions taken into account were those for an intemal axisym- metrical through-wall circumferential defect (Drubay etal, 2002) 65.2.2. The case of the tee-to-bend transition. 65.2.2.1. The thick part of the transition. ‘The first point (location b in Fig. 20) treated was located in the thick part of the mixing tee. On the basis of the lon- sgitudinal stress profiles, the stress intensity factor over time was determined, for different crack depths. The type of defect treated was an internal axisymmetri= cal through-wall circumferential delect. Fig. 2 shows, that the intensity factor was negative from a crack depth of 6mm (1. alh of 0.2). The propagation then stopped. 6.5.2.2.2. The thin part of the transition. The see- cond point (location a in Fig. 20) treated was located in the thin point of the mixing tee, in the close vicinity of the preceding point, in order to qualify a possible thickness effect. The two points were subjected to sub- stantially equivalent surface stress (locations a and b in Fig. 20). The type of defect treated was an internal axisymmetrical through-wall circumferential defect. A number of eycles to reach a crack length of 6.92 mm. (alh=0.8) was 21,588. The extremely rapid propaga- tion of the erack can be explained by allowance being, made for an axisymmetrical defect (see Fig. 23). 7. General conclusions This paper contains a complete analysis of the mix- ing tee of the Civaux unit 1. Metallurgical examinations of the residual heat removal system mixing tee revealed the presence of substantial cracks as well as local ther- ‘mal fatigue cracks: « in the mixing tee, ‘» on the outside of the bend, « in the straight sections of the line, ‘The cracks observed in the mixing tees and the con- nection with the piping were substantially explained by the various calculations made, as this area is shown to be strongly loaded and subjected to strong temper ature fluctuations, In the case study described in this, paper, the stress variation levels are in the region of 570MPa (in a relatively small area) in the four-bend configuration, Allowance made for the coarse descrip tion used for this arca, these levels suffice to explain the appearance of cracks. ‘As concerns the thermal fatigue cracks downstream of the bend in the straight section of the line, the cal- culations showed that they could be due to twisting of the velocity of the fluid flow in this duct. The case considered in this paper showed that the variations of equivalent stress were of the order of 250 MPa in the four-bend configuration. The location of these zones: ‘where the thermal fluctuation was substantial was morei daPain) i 'S Chapulio 1 a, / Nuclear Engineering and Design 235 (2005) $75-596 595 Time (10 5) ig. 22. Vacation ofthe stress intensity facto in the thick prt ofthe tee (e= 30 mm). Time (105) Fig. 23, Variation ofthe tess intensity factor inthe thin part ofthe te (¢=8 65 mm). than two diameters downstream of the bend, as was ob- served in practice at Civaux unit 1 ‘The critical point of the Civaux unit 1 incident was the appearance of a crack on the outside of the bend. and its rapid propagation through the wall. Before this study, no load had been identified that could explain the cracking in this zone. ‘The calculations described in this paper show the importance, in understanding, this phenomenon, of making allowance for a substan- tial length of line with the various bends upstream. ‘The reference calculation (RRA2C) only included two bends and no significant thermal or mechanical load ‘was found. When allowance was made for a greater length of line upstream, with the various bends present in it (RRAAC), evidence was found of large-scale i stability of greater amplitude. This large-seale insta- bility gave rise to pulses which were characterised by the presence of significant thermal fluctuations. It is to be noted that this instability was triggered by high- frequency random fluctuations. ‘These were therefore ‘equivalent to average flow oscillation “specific modes” ttiggered by turbulence. ‘The mechanical variations induced were around 225MPa for the circumferential stress. For stainless596 S Chapulot etal / Nuclear Engineering and Design 235 (2005) 575-596 values should not have caused initiation (vhich remains to be confirmed), but it should be bore in mind that the values were obtained without allowance for the presence of non-flush weld bead on the outside, which would have substantially increased the local stress variation values Furthermore, the calculations described in this paper correspond to a particular mixing factor, flow velocity, uration and turbulence combination. ILis unlikely that it corresponded to the worst case for the line Inaddition, on the basis of the analysis of the surface stress, @ study of crack depth was made. It was then ‘observed that, forthe thin sections, the outside and the outlet of the tee, the calculated loadings resulted in very rapid propagation until the wall was breached. In particular, on the outside of the bend, once a crack was initiated at the surface, it propagated through the wall in $00. Thisresult is importants it represents the first evidence for a thermal load strong enough to cause a through-wall crack. For the thick parts, the values of AK cancelled out, resulting in stopping oferack growth, at 6mm ‘Asa general rule, the four-bend configuration made it possible to identify the main flow phenomena which, could have given rise to sufficient thermal fluctuations for damage to the structure to occur. Even though the thermal-hydraulic models used for the purpose are still being finalised and validated, the results of the calcu lations show how certain types of flow can become ‘unstable and very harmful for the structure. For an industrial-scale operator, itis clear that this type of un- stable mixing must be eliminated. Itremains to be seen ‘whether it can be predicted and avoided at the design stage Acknowledgements ‘The authors would like to thank the Cast3M team consisting of T. Charras, A. Millard and P. Verpeaux for the assistance they provided in setting up the coupled thermal-hydraulic calculations. References ABNOR, 1996, Fatigue sous solicitations amplitude vanable “Méthode Railow de eomptage de eycles. Norme AFNOR, A3- 406 (Cast3M Documentation, 2004. hip www-cast3m cea fe Drubay,B ,etal,2002, Al6: Guide pour lanase de a nocivite des <éfaus et Fuite Avant Rupture. RCC-MR, AFCEN, Pais. Fm, A., Guermond, J, 2001. Eléments fini théore, applica: jon, mise en euvte. Springer: Mathematiques et Applications 3 Faidy, C., ct al, 2000, ‘Thermal fatigue in French RNR sys- ims In: Int Conf on Fatigue of Reactor Components. Vapa, fornia Gelineu, © , Simoneau, , Sperandion, M, etGuinovat J, 1999 Review of predictive methods applied to thermal stripping prob- lems and recommendations. In. Proceedings ofthe SMIRT 15, Paper F06!3, Seoul Guermoné, 51, 1998 Sur''approximation des uations de Nave ‘Stokes par une méthode de projection. C.R. Acad. Sel. Pati 319, 7-492 Jones, [S., Lewis, M.W4 ‘conditions in he frequency response model of thermal stripping Fatigue Fact. Engng. Mater Struct 18 4), 489-502 Kasahara, N., Yacumpai, A, Takasho, I, 1999. Structural re- sponse diagram approseh for evaluation of thermal sipping phe= ‘nomenon Trans. On SmiRT-15 Schlichting 1, 1968, Boundary-ayers Theory: MeGraw-il Smagorisky, J, 1963. General eicelation experiments with the primitive equations. Mon. Weath. Rev. 91 (3), 99-164 995, ‘The effet of various constraint