- Email: [email protected]
structure analyses of the MEGAPIE spallation source target during transients
Nuclear Engineering and Design 237 (2007) 1656–1667
Coupled fluid/structure analyses of the MEGAPIE spallation source target during transients B.L. Smith a,∗ , W. Leung b,1 , A. Zucchini c,2 a
Thermal-Hydraulics Laboratory, Nuclear Energy and Safety Department, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland b Spallation Neutron Source Division, Paul Scherrer Institute, CH-5232 Villigen PSI, Switzerland c UTS Tecnologie Fisiche Avanzate, Sezione Metodologie e Diagnostiche, ENEA, Bologna, Italy Received 25 January 2007; received in revised form 1 February 2007; accepted 5 February 2007
Abstract The paper describes some R&D activities undertaken in support of the design and safe operation of MEGAPIE (MEGAwatt PIlot Experiment) spallation source target, which is scheduled to be irradiated by a proton beam in the SINQ facility at the Paul Scherrer Institute in 2006. The target material is lead bismuth eutectic (LBE), which also acts as the primary coolant. As a consequence of the spallation reactions, about 600 kW of heat would be deposited in the target during operation, and considerable R&D effort is being expended to demonstrate continuing coolability and structural integrity under a variety of operational and abnormal conditions. The paper gives three examples of transient analyses carried out as part of the safety assessment of the target: (1) a beamline trip and recovery; (2) failure of the primary electro-magnetic pump (EMP); (3) failure of the secondary EMP (used to cool the base of the target). The study involves the simultaneous application of a system-analysis code, in our case a version of RELAP5, a computational fluid dynamics (CFD) tool (CFX-4), and a structural analysis code (ABAQUS). The RELAP5 code is used to provide transient boundary conditions for a localized conjugate heat transfer analysis of the lower target region, undertaken using CFD, and includes the feed-back effects arising from the secondary cooling and control systems. A conjugate heat transfer problem is then solved using CFD, which provides time-dependent thermal and flow data within the LBE, together with the thermal and mechanical loads to the target structures. Finally, an in-house interface program is employed to transfer mesh geometry, model topology and (time-dependent) thermal/mechanical data to enable stress analysis of the principal lower-target structural components to be performed. It is demonstrated that none of the transients considered result in critical stress conditions occurring in the target components, but that further operation is not recommended unless both pumps are fully operational. © 2007 Elsevier B.V. All rights reserved.
1. Introduction In the ADS (Accelerator Driven System) concept, neutrons generated from spallation reactions are used to initiate and maintain a continuous chain reaction in a sub-critical nuclear core (Rubbia et al., 1995; Bolderman, 1997). In some ADS designs, the spallation neutrons are generated from proton bombardment of liquid lead bismuth eutectic (LBE), which then acts both as the target material and the means by which heat is removed from the spallation region. A pilot facility at the Paul Scherrer Institute in Switzerland, and supported by an international ∗ 1 2
Corresponding author. Tel.: +41 56 310 27 26; fax: +41 56 310 44 81. E-mail address: [email protected] (B.L. Smith). Tel.: +41 56 310 40 52; fax: +41 56 310 44 81. Tel.: +39 051 609 82 56; fax: +39 051 609 80 62.
0029-5493/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.nucengdes.2007.02.020
partnership, aims to demonstrate the feasibility of the LBE target concept (Bauer et al., 2001). Efforts are coordinated under the project title MEGAPIE, and groups from Belgium, France, Germany, Italy, Japan, South Korea, Switzerland and the USA are involved in the design, manufacture, safety assessment and, finally, operation of the target. As a consequence of the spallation reactions, about 600 kW of heat is deposited (Foucher, 2003), and it is necessary to demonstrate that the target would continue to be cooled adequately, and that structural integrity would be maintained, under a variety of operational and abnormal conditions. It is not possible to fully verify the cooling principle by means of mock-up experiments, because of the difficulty in reproducing the volumetric heating at the appropriate power density (at maximum, 1 GW/m3 for a beam power of 1 MW). Consequently, heavy reliance is placed on detailed numerical analyses. Such analyses
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
are reported here, the procedure utilizing a three-step modelling approach. The system behaviour is first determined using a version of the RELAP5 (1995) code specially modified for use with LBE. This provides transient boundary conditions for a subsequent CFD analysis, in our case using CFX-4 (CFX-4.3, 2000), to determine the thermal and structural loads to the structures. Finally, a stress analysis calculation is performed, here using ABAQUS (2001), to estimate the structural response to these loads. 2. MEGAPIE target assembly The MEGAPIE target, shown in cutaway and schematic form in Fig. 1, consists in the lower section of a cylindrical steel vessel about 2 m high and internal diameter 0.18 m, closed at the bottom by a hemi-spherical cap, or window, and flanged at the top to a larger vessel of 0.22 m diameter housing the annular target heat exchanger (THX). The coolant is LBE, and flow is driven by the main electro-magnetic pump (EMP), located between the THX and a cylindrical guide tube separating the hot and cold streams. LBE at about 230 ◦ C is discharged from the lower end of the THX to the annular space between the guide tube and the target hull, where it descends to the window and rises in the space inside the guide tube towards the main EMP entrance. Here, the now hot LBE is pumped through the inlet manifold of the THX, and down over 12 cooling pins (arranged in a ring)
1657
to complete the LBE circuit. A bypass pump is situated beneath the main pump assembly, and this is used to provide, via suitable ducting, a secondary coolant flow to the window. Down the centre of the target there is an instrumentation/control rod, which also contains an electrical heater for maintaining the LBE temperature during shutdown periods. The LBE target is surrounded by a D2 O-cooled safety hull, which protects the beam tube from possible LBE leaks. The gap between the target and safety hulls is filled with helium at a pressure of 0.5 bar. The entire assembly is mounted vertically, with protons entering the target from below. The protons pass through the safety and target windows and react with the LBE, where the bulk of the spallation neutrons are produced. These are collected in collimators placed around the target and subsequently utilized, particularly for materials research. Detailed CFD calculations (Dury et al., 1999) have shown that the bulk of the heat generated in the target is in the LBE itself (98%), though there is some deposition in the window (0.78%) and guide tube (0.89%) structures, which is of importance locally. The heating of the hull (the cylindrical part of the target outer boundary) due to the beam is much less important (0.18%), and there is no heat deposition in the instrumentation/heater rod, which is situated above the active spallation zone. Very important is the heat transfer from the window to the LBE (about 4.2 kW). Adequate cooling must be guaranteed under all conditions in which the beam is turned on to prevent structural failure. Significant too is the heat transfer from the hot riser to the cool annulus section through the non-insulated guide-tube wall (about 200 kW), since this influences the overall LBE operating temperature. The numerical simulations reported here all follow a threestep analysis strategy: (1) The transient system response is first determined using a RELAP5 model of the target and auxiliary equipment (Leung, 2004; Leung and Smith, 2005), the model including the primary/secondary/tertiary circuits and their control systems. Among other data, the system simulation supplies time-dependent flowrate and temperature data for the LBE entering the annulus and bypass channels to the lower target during transient events. (2) These data are tabulated, and then used as boundary conditions to an accompanying, detailed CFD simulation of the LBE flow in the lower target, and the heat transfer with nearby structures. (3) Geometry, topology and mechanical/thermal loads pertaining to the structures are transferred, via an interface program (Smith et al., 2004), to become input data to the structural analysis code ABAQUS, with which the local stress fields in the components are computed.
Fig. 1. Cutaway and schematic views of MEGAPIE target (configuration with a flat-ended guide tube).
Note that a particular advantage of the present approach is that there is no need to recalculate the heat transfer problem within the structural analysis code. Point-wise thermal data are transferred directly from the CFD simulation, avoiding the need to impose average heat transfer coefficients or heat fluxes.
1658
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
3. The modelling stages We give here three examples of transient analyses which have carried out as part of the safety assessment of the target: (1) a beamline trip; (2) failure of the primary electro-magnetic pump (EMP); (3) failure of the secondary EMP. Beamline trips are expected to occur frequently as a consequence of accelerator problems and/or beam instabilities during the operational lifetime of the target (up to 1 year). The target design (optimized for steady-state operation) is expected to be sufficiently robust to resist over-stressing of components during such incidents. In particular, adequate cooling of the window must be maintained at all times, and under all conditions, including during abnormal operation, and for (postulated) accident sequences. 3.1. System behavior: RELAP5 The MEGAPIE cooling system is of a triple-loop type, employing three working fluids: LBE (primary coolant), Diphyl THT (intermediate oil loop) and light water. The oil loop is necessary to avoid any possible LBE/water contact in the (unlikely) event of cooling pin rupture. The circuits are thermally connected by heat exchangers. A schematic of the system layout used as the basis of the RELAP5 model, showing the target and bypass circuits, the target heat exchanger (THX), intermediate cooling loop, intermediate heat exchanger (IHX), water cooling loop (WCL), and the water heat exchanger (WHX) is shown in Fig. 2. The original model was assembled by Petrazzini and Alemberti (2002), and extended by Leung et al. (2003). CFD analysis of the lower target response to the beamtrip and pump-failure incidents cannot be considered in isolation, as there is thermal coupling with the ancillary circuits, and transient feedback effects resulting from programmed responses of the control system. Results from the RELAP5 simulation are provided in the form of time-dependent mass flowrate and temperature boundary conditions for the LBE inlets to the detailed, stand-alone CFD study. Lack of computer power precludes the option of
Fig. 2. Schematic representation of the system layout used as the basis to the RELAP5 simulation.
using CFD to represent the entire system. Further details of the RELAP5 models are given by Leung and Smith (2005). 3.2. CFD model: CFX-4 Under normal, steady-state operating conditions, the proton beam is at full power, and the LBE pumps and all auxiliary circuits are functioning normally. Thermal-hydraulics simulations are being undertaken to ensure there is adequate cooling in the lower target region where the heat is deposited (Dury, 2004). The lower target geometry is shown in Fig. 3. (The outer, water-cooled safety container seen in Fig. 1 is not modelled.) The simulations being undertaken are aimed at optimizing the geometrical layout, including the shape of the guide tube (flat or slanted at the bottom), and the bypass nozzle area, shape and position. The CFD model for the slanted option is also shown in Fig. 3, extending up to a level just below the exit from the heat exchangers: i.e. to 2.15 m above the lowest point on the internal surface of the target window. The window, made from T91 steel, is 1.5 mm thick at its lowest point, increasing uniformly to 2 mm at the junction with the cylindrical part of the hull. The central rod contains thermocouples and other instrumentation, and heaters to maintain LBE above its melting temperature when the target is in hot standby condition with the beam off. The window, hull, guide tube and central rod are all defined as conducting solids in the CFX4 input data. The model used in this study consists of 113,000 hexahedral fluid cells and 62,000 hexahedral structural elements, the mesh layout in the lower section being reproduced in Fig. 3. The target hull, annulus, riser and instrumentation/heater rod are clearly discernible. The ducting to the bypass nozzle is also modeled, though cannot be seen in this picture, since the duct walls coincide with mesh lines. The bypass nozzle itself is rectangular, of dimensions of 10 mm × 20 mm, and is orientated to provide a stable wall jet over the hottest part of the window. A CFD conjugate heat transfer problem was set up within CFX-4 and run to steady-state to provide the starting conditions for each of the transient simulations. The heat deposited in the LBE and structures is defined in terms of volumetric sources, as described by Dury (2004). The flow of LBE is downwards in the annulus, over the window, and upwards in the central region or riser, with additional window cooling being provided by the bypass jet. The inlet boundary conditions are the mass flowrates of the LBE entering the top of the annulus and the entrance to the bypass duct (37.5 kg/s and 2.5 kg/s, respectively), and the temperatures (230 ◦ C and 242 ◦ C, respectively). (To avoid introducing an excessive number of meshes, heat transfer across the wall of the bypass duct was not taken into account. The increase in temperature of the LBE in the duct from 230 ◦ C at the THX outlet to the 242 ◦ C at the nozzle exit has been estimated using 1D correlations.) The outlet is defined as a constant-pressure surface at the top of the riser. The inlet Reynolds number at the top of the annulus is 81,600, so the flow is turbulent. To take account of the turbulence effects, a high Re k–ε model, with wall functions, has been employed. Care was taken to ensure legitimate use of the logarithmic law of
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
1659
Fig. 3. Lower target geometry, hexahedral mesh layout, and steady-state temperature distribution (configuration with slanted guide tube end).
the wall near non-slip boundaries; that is, by choosing the centre of the cell next to the boundary to be in the range 30 < y+ < 100, expressed in wall units (Launder and Spalding, 1974). Because of the low Prandtl Number for LBE (Pr ≈ 0.02), it may not be appropriate to apply the logarithmic profile for the near-wall temperature field (Kader, 1981). Automatic checks are made in the code, and a linear profile is employed if the first mesh cell is located within the laminar part of the thermal boundary layer; see (CFX-4.3, 2000) for details. The temperature distribution in the target in the vertical plane of the bypass jet is shown in Fig. 3. The spallation region extends from the window up the riser to a height of 27 cm, and the figure shows that the maximum temperature of 382 ◦ C occurs just below this level. The distribution is very non-uniform, and one can expect that the structures nearby will become thermally stressed as a consequence of strong, local temperature gradients. Of special interest is the structural integrity of the window, since this is where the beam enters the target, and which must remain adequately cooled under all conditions to prevent loss of LBE. It can be seen from Fig. 3 that, in the plane of the bypass, supplementary window cooling is being provided by the jet flow to this region. 3.3. Structural solver: ABAQUS ABAQUS (2001) is a commercial, general-purpose Finite Element Method (FEM) program for the solution of many different linear and non-linear structural problems, including stress analysis, heat transfer, mass diffusion and acoustics. In stress analysis, FEM is applied to the principle of virtual displacements, with the stress/displacement elements based on the Lagrangian description of deformation. The degrees of freedom, calculated at the nodes of the element, are the fundamental
variables of the analysis. In ABAQUS, the material response (stiffness) at the integration points within the elements is evaluated for most elements by full or reduced integration (Gaussian quadrature) over the volume of the element. In the present analyses, reduced integration has been used. From the element stiffness matrices, the global stiffness matrix is assembled, and a global system in the unknown degrees of freedom is generated. Linear equation solvers are used for both linear and non-linear analysis. In the latter case, Newton’s method (or a variant of it) is utilized, within which a set of linear equations are solved at each iteration. This is done using direct Gaussian elimination, coupled with a sparse solver based on a “multi-front” technique. All details are contained in the appropriate program documentation referenced above. Here, the mesh discretization used for the structural analysis is adopted directly from the CFD model. Consequently, use is made of linear hexahedral elements to represent the structures, the degrees of freedom being the translations at each of the eight corner nodes. Though most components consist of thin shells, there are five elements across the thickness to ensure adequate representation of the bending moments. Details of the material properties are given in Table 1. For convenience in defining the boundary conditions, each structure is suspended from its top axial section by imposing zero axial displacements to all nodes along the upper edge, and three anchor points on the top rim are constrained to move in the radial direction only. This prevents rigid body movements of the structures, and simulates well the flange attachment in the actual target. Local stress concentrations around the anchor points are ignored. From earlier studies, it was found necessary to have spacers between the guide tube and the target hull to limit lateral displacement of the guide tube resulting from the differential heating. The spacers are simulated in the model by
1660
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
Table 1 Steel material properties used for stress calculations Component Target hull Guide tube a
Material
Young’s modulusa
Poisson ratio
Expansion coefficient
T91 AISI 316
1.51–2.06 × 1011
0.3 0.3
11.5 × 10−6 K−1 16.0 × 10−6 K−1
Pa
2.0 × 1011 Pa
Young’s modulus is a bi-linear fit to temperature-dependent data (Zucchini and Smith, 2003).
Fig. 4. Conditions over the window external surface in steady-state.
constraining three pairs of nodes around the bottom edge of the guide tube and the hull to have the same radial displacement. Fig. 4 shows the temperature and von Mises stress distributions over the external surface of the window. The bypass flow is from left to right, and in this configuration is aligned with the minor axis of the elliptical beam footprint. Maximum temperatures occur near the centre of the window, slightly displaced as a consequence of the cross-flow. The stress distribution cal-
culated using the ABAQUS code is also shown in the figure, with the point of maximum stress also displaced from the centre point of the target. The mesh layout derives from the CFX-4 simulation, and has been carried over, together with the thermal and mechanical (surface pressure) data, to ABAQUS via the interface program, as previously explained. The associated temperatures and stresses on the guide tube inner and outer surfaces are shown in Fig. 5, looking from the lower edge upwards. Again, the temperature distribution is seen to be rather non-uniform. The hot spot at the lower end lies just above the bypass nozzle, and derives from a small vortex in the flow, which becomes established in this region. This is also the point of maximum stress in the guide tube. The differential heating effect up the length of the riser section (also seen in Fig. 3) results in elongation and bending of the guide tube and instrumentation rod, the relative magnitudes of which are also given (in exaggerated form) in Fig. 5. These findings confirmed the need for spacers to be installed, to secure the target components, and became a feature of the design thereafter. Full details are given by Zucchini (2002). 4. Analysis of transients
Fig. 5. Conditions on the guide tube (in steady-state), and relative deformations of the structural components (magnified).
Each of the transients analyzed begins at time t = 0, starting from steady-state conditions. The calculation procedure is the same in all cases. First, a RELAP5 system simulation is carried out for the particular transient under consideration. Histories of the mass flowrates and temperatures at the THX and bypass nozzle exits are tabulated from the run, up to such times as steady operating conditions have been re-established. These become, respectively, time-dependent inlet boundary conditions for the annulus and bypass duct for the transient CFD simulation, which is carried out next. Time-dependent thermal/mechanical data from the CFD run are stored at time-steps of interest, and
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
1661
Fig. 6. Input conditions to the CFX-4 simulation of the lower target derived from the RELAP5 system analysis for the beamtrip transient: (a) beam power, (b) inlet temperatures and (c/d) annulus/bypass mass flowrates.
these become input data for a series of static, stress analysis calculations carried out using ABAQUS. 4.1. Beamtrip and restart Although the SINQ facility at PSI is meant to be a continuous spallation source, the proton beam does not operate continuously because ultra-sensitive detectors employed at the injector from the cyclotron, and along the beam lines, can trigger a shutdown of the beam whenever a fault signal is detected. Two types of beam outages may be discerned: a beamtrip and a beaminterrupt. In the former case, the beam is shut off instantly, followed by a 10 s down time, and a ramp up to full power over 20 s. This event occurs quite frequently—perhaps 50 times a day. In contrast, for a beaminterrupt, the beam is shut off instantly, but recovery is delayed for an extended period, from 30 s, to several minutes, or for a few days. Beam-interrupt events occur much less frequently, perhaps two or three times a week, and may be instigated by the operators: for example, during regular maintenance periods. In such circumstances, the target has to be in a hot standby condition, ready for restart. The present study focuses on the response of the lower target components to the beamtrip and restart event. Details of conditions elsewhere in the cooling system have been described at length in Leung and Smith (2005). Details from the RELAP5 simulation are shown in Fig. 6. For any beamtrip event, the loss of beam power is instantaneous, but the recovery to full power is a controlled ramp over a 20 s period. This is an enforced operating practice, to avoid thermal shocks. The inlet temperature histories (Fig. 6b) generally follow that of the power, but are smoothed by the continuing heat transfer processes taking place between the LBE, the still hot structures, and the cooling pins in the THX. Though the pumps remain in operation, the loss of the buoyancy-driven part of the flow circu-
lation is seen in the reduced flowrates to the annulus (Fig. 6c) and bypass (Fig. 6d) inlets. As a consequence of the thermal inertia of the system, full recovery to steady-state operation takes altogether about 60 s, to be compared with the power transient of 30 s. The transient mass flowrates and temperatures given in Fig. 6 are used as boundary conditions for the accompanying CFD simulation of the lower target region. Fig. 7 illustrates the temperature response at the centre of the window and at the point of maximum temperature on the guide tube, obtained from the CFX-4 simulation. Following beam shut-off (at t = 0), the window, which is still being cooled by the bypass LBE flow, loses heat very rapidly. The minimum temperature occurs 10 s into the transient, at the time the beam power starts to be ramped up. Because of the slow thermal response of the target system, about 60 s are required for conditions to return to steady-state. The guide tube, which is the most stressed component in the lower target region, is not subjected to such strong thermal transients due to the heat capacity of the LBE.
Fig. 7. Temperature responses in the lower target region for the beamtrip transient, as determined using CFX-4.
1662
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
Fig. 8. Time histories of (a) the maximum temperature and (b) the stress history in the lower-target structures for beamtrip transient.
Time-dependent thermal data for the lower-target structural components, derived by CFX-4, were passed to the ABAQUS code for the stress analysis. Results are presented in Fig. 8. As expected, minimum temperatures in the structures are seen at the end of the beam-off period (t = 0–10 s), and then gradually increase as the beam is ramped up to full power (t = 10–30 s). Under normal operation, the beam deposits 4.2 kW of heat in the window, which is transferred to the LBE at the inner, cooled surface; the outer surface is assumed adiabatic, because of the poor heat-transfer properties of the helium in the gap between the target and safety hulls. The temperature difference from the outer to inner surface of the window (about 21 ◦ C) sustains the heat flux at the inner surface necessary for this heat transfer process. Following the beamtrip, the heat source is lost, and the temperature becomes uniform through the window thickness, as seen in Fig. 8a at t = 10 s. During the subsequent 20 s ramp to full power, the temperature gradient is re-established monotonically. Fig. 8b shows that no stresses are created in the lower target structures, and conditions return to their equilibrium states without any transient stress crises.
LBE container during this transient, and to evaluate the potential for target operation by natural circulation only, perhaps at reduced power. The initial, steady-state condition is that pertaining to normal operation of the target. The pump failure incident begins at t = 0. According to the system calculation (Leung and Smith, 2005), the mass flowrate to the annulus falls sharply from the initial 37.5 kg/s, and then recovers partially to about 23 kg/s, driven by the buoyancy of the heated LBE in the riser (Fig. 9a). The bypass flow is hardly affected, under the assumption that the bypass pump continues to function normally (the double-pump failure scenario is of lower probability, and will be analysed at a later date, if requested by the safety authorities). The temperature at the top of the annulus falls, due to the enhanced cooling of the LBE by the THX resulting from the reduced flow in the primary circuit (Fig. 9b), while full flow is maintained in the intermediate cooling loop. The temperature drop triggers operation of the ICL three-way valve (Fig. 2) at t = 10 s, isolating the THX. This leads to temperature recovery in the primary LBE circuit (Fig. 9b). Further operation of the ICL control system leads ultimately to new steady-state operating conditions. The monotonically-increasing nozzle exit temperature (which in the RELAP5 calculation includes the effect of heat transfer to the bypass duct from the LBE in the annulus) reflects the establishment of generally hotter operating conditions overall in the target due the reduced LBE circulation rate. New equilibrium conditions (at the reduced flowrate) are established after about 150 s. Using these data, a transient, conjugate-heat-transfer, CFD simulation was performed to determine the flow and thermal
4.2. Failure of main pump In the event of failure of the main EMP, the flowrate in the primary circuit reduces very quickly, because of the low-inertia and high-resistance characteristics of the loop, but partial flow is maintained as a consequence of the buoyancy induced by the beam heating. For the lower target region, where the heating takes place, it is necessary to ensure that there is no risk to the
Fig. 9. Input conditions to the CFX-4 simulation for the main-pump-failure transient.
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
1663
Fig. 10. Velocities and temperature contours 5 s into the main-pump-failure transient.
conditions in the lower target region. At regular intervals (and other times of particular interest), thermal and mechanical load data were stored for the subsequent structural analysis. The simulation was run for 100 s, which is close to the new steady-state operating point. Fig. 10 shows conditions 5 s into the pump-failure transient. From the velocity field, one notices that the bypass jet has penetrated into the opposite annulus. This is because the main flow has been reduced, and flow conditions are no longer optimal over the window. However, the window itself remains adequately cooled, since the bypass flow has not been affected by the loss
of the main EMP. That is, the velocity of 1.43 m/s at the bypass nozzle exit remains as before, the wall jet being the controlling mechanism for regulating the window temperature. However, the non-optimized flow field has an adverse effect on the cooling of the target hull. The bypass flow picks up heat from the window, and directly from the beam, and transports it to the opposite annular gap (instead of to the riser region), producing a hot spot on the sidewall. This is expected to result in unwelcome local stresses in the material, as later confirmed by the structural analysis. Already, the maximum temperature in the LBE has increased by 70 ◦ C, and the proximity of the maximum to the
Fig. 11. Velocities and temperature contours 30 s into the main-pump-failure transient.
1664
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
Fig. 12. Maximum temperature histories in the LBE and on structural components for the main-pump-failure transient.
lower end of the heater rod will mean increased temperatures in the material, stronger temperature gradients, and increased thermal stresses. The situation 30 s into the transient is shown in Fig. 11. In the plane of the bypass jet, the upflow region in the annulus has reached a height of more than 27 cm, with accompanying increase in temperature. The region of strong beam heating is much narrower (cf. Fig. 3), and the maximum LBE temperature has now increased from the estimated 387 ◦ C under normal operating conditions to 600 ◦ C. Maximum temperatures in the window, guide tube, heater rod and LBE are given in Fig. 12 for the first 80 s of the transient. There is a strong heating up of the lower part of the target (except the window) over the first 15 s, followed by a more gradual return to steady-state operating conditions, but one in which the main flow is now buoyancy driven. The window remains cooled by the bypass flow, with the mean of the maximum temperature staying constant at about 375 ◦ C. However, the maximum is seen to fluctuate, with many short-lived peaks. It should be remembered that maximum window temperatures are plotted here rather than the
temperature evolution at a given point, and the erratic behaviour may not reflect any local physical development. The frequency of the fluctuations is in any case too high to be of consequence to the stress history in the window material. The maximum LBE temperature increases by about 180 ◦ C, and is also oscillatory after 35 s. Maximum heater rod and guide tube temperatures increase by about 100 ◦ C and 125 ◦ C, respectively, the latter displaying the same oscillatory behaviour as the LBE. In contrast to the seemingly random fluctuations over the window, however, the oscillations in the LBE and on the guide tube are more regular, and are most likely caused by instabilities in the downflow region on the narrow-gap side of the guide tube into which the bypass jet penetrates, though more detailed analysis would be needed to confirm this. The temperatures and stresses over the outer surface of the window are given in Fig. 13 at different times in the transient. The maximum window temperature is slightly displaced from the centre as a consequence of the cross-flow induced by the bypass jet (indicated by the arrow), which, by assumption, still functions normally. The von Mises stress pattern changes with time, but there are no severe stress concentrations. Maximum stresses for the lower-target structures at various times are summarized in Table 2. Stresses remain more-or-less constant over the window, as previously noted, but increase markedly for the guide tube (from 49 MPa at t = 0 to 181 MPa at t = 30 s) and for the cylindrical part of the target vessel (from 45 MPa at t = 0 to 90 MPa t = 20 s). Though large, these stresses do not pose threats to the structural integrity of the components: the yield stress for T91 steel (window and target hull) at 400 ◦ C is 460 MPa, and that for SS316 (guide tube) at the same temperature is 190 MPa. 4.3. Failure of bypass pump The bypass pump has a particular function in providing a continuous jet stream of LBE across the window, and stabilizing the flow in this region. Sudden failure of the pump is not expected to represent a catastrophic event, because the slanted design of
Fig. 13. Temperatures (top) and von Mises stresses (bottom) over the window outer surface at: (a) t = 0, (b) t = 5 s and (c) t = 30 s into the main-pump-failure transient; the arrow indicates the bypass flow direction.
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
1665
Table 2 Maximum von Mises stresses at various times during the pumptrip transient
Table 3 Changes in input conditions for bypass trip transient
Time (s)
Inlet to
0 1 2 3 4 5 10 15 20 30 40 60 80
Maximum von Mises stress (MPa) Guide tube
Window
Target vessel
49.1 57.5 54.2 59.0 65.0 70.8 118.8 149.1 163.0 180.9 140.0 155.1 175.5
44.6 47.5 43.9 43.2 43.0 47.1 41.6 44.8 45.1 46.4 44.3 45.3 47.8
44.6 47.5 43.9 43.2 43.0 47.1 41.6 61.7 89.7 55.4 53.0 45.3 60.2
the guide tube would still maintain a cross-flow over the hottest part of the window. However, the transient still requires analysis to give input to the decision whether to continue operation of the target. According to the RELAP5 simulation (Leung, 2004), in the event of failure of the bypass EMP, with the main EMP operating normally, the flowrate in the bypass duct decreases very quickly (low inertia), but maintains a bleed flow at the nozzle of about 0.9 kg/s. Due to the heating effect of the pump, LBE enters the top of the duct at around 260 ◦ C, cools as a result of heat exchange with the LBE in the annulus at 230 ◦ C, but then heats up again because heat transfer from the riser to annulus region increases the temperature of the LBE in the annulus gap with respect to that in the duct. According to the RELAP5
Annulus Bypass
Temperature (◦ C)
Mass flowrate (kg/s)
t<0
t>0
t<0
t>0
230 242
230 254
37.5 2.5
39.9 0.9
simulation, this complex interaction results in the exit temperature at the nozzle being raised to 254 ◦ C, and this value is used for the subsequent CFD simulation. Since the temperatures and flowrates for the main and bypass circuits change so abruptly, simple step-changes in the input conditions for the CFD calculation suffice to define a worst-case scenario. Details are given in Table 3. The CFD simulation has been run for 100 s, which is close to the new operating point. Fig. 14 shows the temperature field in the target 5 s after the start of the transient; the temperature scale is kept the same as that in Fig. 3 in order that comparisons may be made. The cross-flow over the window is still evident in Fig. 14a, which is the plot taken in the plane of the bypass flow; the flow remains more-or-less symmetric in the plane perpendicular to this (Fig. 14b). The higher main flow reduces the region of maximum LBE temperature almost to a point, just above the spallation region. In fact, the highest temperature in the target is now at the window, which peaks at this time, 140 ◦ C above normal, steady-state conditions. Time histories of the maximum temperatures in the LBE, and on the window and guide tube surfaces, are shown in Fig. 15. The maximum window temperature peaks at t = 5 s, then stabilizes after t = 10 s. The final temperature is 115 ◦ C above normal. The
Fig. 14. Temperature contours 5 s into the bypass-pump-failure transient (a) in the plane of the bypass jet and (b) perpendicular to it.
1666
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
Fig. 15. Time-plots of maximum temperatures for bypass-pump-failure transient.
maximum LBE temperature at first falls sharply, then also peaks at t = 5 s, stabilizing at t = 10 s. The final (maximum) temperature is the same as under normal operating conditions. The guide tube maximum temperature also falls initially, but then recovers, and thereafter remains constant, about 30 ◦ C above normal. There is evidence of oscillatory behaviour in the lower target region, with period of about 15 s, beginning 30 s after transient initiation. Careful inspection of the flow and temperature field over the window has revealed that the instability is most likely buoyancy-driven, due to overheating at the centre of the window
and periodic lifting off of the attached boundary layer, which the bypass flow is no longer able to stabilize. The principal results from the subsequent ABAQUS calculation are given in Fig. 16. During the first 5 s, the window temperature increases sharply by 180 ◦ C, due to the loss of the direct cooling effect of the bypass jet, and then stabilizes at about 140 ◦ C above nominal operating conditions. The maximum guide tube temperature does not peak, but is nonetheless elevated to 400 ◦ C, which is 60 ◦ C higher than normal. During the phase of rapid temperature increase in the first 5 s, the maximum von Mises stress on the window peaks at 118 MPa on the outer surface (in steady-state, the maximum stress was 44 MPa on the inner window surface), but then stabilizes at 90 MPa. The maximum stress on the inner surface exhibits similar behaviour, with a peak at 98 MPa, also stabilizing to 90 MPa. These stresses are well below the yield stress of T91 at 500 ◦ C of 400 MPa, so the structural integrity of the window is not threatened during this transient. The most dramatic change is the maximum stress in the guide tube, which increases from 49 MPa to 167 MPa, which is approaching the yield stress for SS316 of 190 MPa at 400 ◦ C. However, the stressed region is highly localized, and the overall integrity of the guide tube remains secure. 5. Conclusions CFD analyses for the lower part of the MEGAPIE spallation source target have been carried out using CFX-4 to simulate conditions resulting from one operational transient (beamtrip) and two abnormal transients (failure of main or bypass pump). In all cases, transient inlet conditions have been employed, these having been obtained from system analyses of the different incidents using a version of the RELAP5 code specially modified for use with liquid metals. The CFD model adopted is that used previously for steady-state analysis of the target, here used in transient mode. At regular time intervals, and at other times at which quantities were changing rapidly, structural thermal data were transferred to the ABAQUS code, and stress analysis of structural components was carried out. Main observations from the analyses are listed below:
Fig. 16. Time-plots of maximum temperatures and stresses in structures for bypass-pump-failure transient.
• The beamtrip and power-up transient does not lead to either excessive temperatures or excessive stresses in the lower target structures. Though the beam heating is lost instantaneously, there is a more moderate cooling of the structures, due to their heat capacities. In this scenario, the beam stays off for 10 s, and then full power is restored gradually over a 20 s period. No transient peaks occur during this process, and temperatures and stresses monotonically return to their equilibrium values. • For the main-pump-failure incident, mean temperatures over the window, through which the proton beam passes, remain constant, controlled by the bypass flow, which is assumed to be still functional. Elsewhere, temperatures rise rapidly: by more than 100 ◦ C for the heater rod and guide tube, and by nearly 200 ◦ C in the beam-heating region of the LBE. The flowrate in the annulus drops from the nominal value
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
of 37.5 kg/s to that driven by buoyancy alone: around 23 kg/s. The flowrates are no longer optimal for the target geometry: the bypass jet overshoots the window, penetrating into the annulus on the opposite side, and resulting in locally higher sidewall temperatures, and most likely initiating flow instabilities. The incident does not represent a threat to the structural integrity of the target, but further operation at full power warrants careful consideration, given the changes in flow conditions and the temperature increases seen in these calculations. • In the case of failure of the bypass pump, the RELAP5 calculation predicts that the velocity of the LBE jet at the nozzle would fall from the nominal 2.5 kg/s to a bleed flow of 0.9 kg/s. According to the CFD simulation, this is too weak to traverse the window and maintain adequate cooling, and the window temperature increases, finally stabilizing 115 ◦ C above normal. The maximum temperature in the LBE is little changed from that under normal operating conditions, though does peak during the transient. The detailed ABAQUS calculation reveals that though the maximum temperature on the guide tube increases by only 30 ◦ C, it occurs at the narrowgap end of the tube, and results in an undesirable increase in the maximum stress on the structure, from about 50 MPa to 170 MPa. There is also evidence of unstable flow behavior. Thus, although the slant on the guide tube ensures a good cross-flow over the window, even in the absence of the bypass flow, overall conditions are far from optimal, and any further irradiation of the target, even at reduced power, would not be advisable. References ABAQUS/Standard User’s Manual I, II, III, 2001. Hibbit, Karlsson and Sorenson Inc., Providence, Rhode Island, USA (www.abaqus.com). Bauer, G.S., Salvatores, M., Heusener, G., 2001. MEGAPIE, a 1 MW pilot experiment for a liquid metal spallation target. J. Nucl. Mater. 296 (1–3), 17–33.
1667
Bolderman, J.W., 1997. Accelerator-driven nuclear energy systems. Australian academy of technical sciences and engineering. In: Academic Symposium: Energy for Ever: Technical Challenges of Sustainable Growth, November 1997. CFX-4.3, 2000. AEA Technology, Harwell, Oxfordshire, OX11 0RA, UK (www.ansys.com). Dury, T.V., 2004. CFD design support at PSI for the international MEGAPIE liquid-metal spallation target. J. Nucl. Sci. Technol. 41 (3), 285–295. Dury, T.V., Smith, B.L., Bauer, G.S., 1999. Design of the European spallation source liquid-metal target using computational fluid dynamics. Nucl. Technol. 127, 218–232. Foucher, Y., 2003. Etude et D´eveloppement d’une Cible de Spallation. Ph.D. Thesis. University of Nantes, France. Kader, K.A., 1981. Temperature and concentration profiles in fully turbulent boundary layers. Int. J. Heat Mass Transfer 24 (9), 1541–1544. Launder, B.E., Spalding, D.B., 1974. The numerical computation of turbulent flows. Comp. Meth. Appl. Mech. Eng. 3, 269–289. Leung, W.H., 2004. On the MEGAPIE target thermal hydraulics—a RELAP5 analysis. In: Proceedings of the 12th International Conference On Nuclear Energy (ICONE 12), Arlington, Virginia USA, April 25–29, 2004. Leung, W.H., Smith, B.L., 2005. The MEGAPIE spallation source target: analysis of operational and accident transients using RELAP5. In: Proceedings of the 11th International Conference on Nuclear Reactor Thermal Hydraulics (NURETH 11), Paper 509, Avignon, France, October 2–6, 2005. Leung, W.H., Petrazzini, M., Alemberti, A., 2003. RELAP5 analysis of the MEGAPIE target. In: Proceedings of the 4th MEGAPIE Technical Review Meeting, Paris, March 18–19, 2003 (FZK Report FZKA 6876, December 2003). Petrazzini, M., Alemberti, A., 2002. Input and RELAP5 Model Description. Ansaldo Report MPIE 1-TRIX 200, Genova, Italy, September 2002. RELAP5/MOD3 Code Manual, 1995. Code Structure, System Models, and Solution Methods, vol. 1. NUREG/CR-5535. Rubbia, C., et al., 1995. Conceptual Design of a Fast-Neutron Operated HighPower Energy Amplifier. CERN/AT/95-44(ET). Smith, B.L., Dury, T.V., Ni, L., Zucchini, A., 2004. A pragmatic coupling strategy between commercial CFD and structural analysis codes. In: ASME/JSME/PVP 5th International Symposium on Computational Technologies, San Diego, USA, July 24–30, 2004. Zucchini, A., 2002. MEGAPIE: Stress Analysis of Lower Target; Preliminary Design. ENEA Report No. RTI/FIS/MET/2002/1. Zucchini, A., Smith, B.L., 2003. Stress analysis of the LMC under accident conditions. In: Proceedings of the 4th MEGAPIE Technical Review Meeting, Paris, March 18–19, 2003 (FZK Report FZKA, 6876, December 2003).
Coupled fluid/structure analyses of the MEGAPIE spallation source target during transients B.L. Smith a,∗ , W. Leung b,1 , A. Zucchini c,2 a
Thermal-Hydraulics Laboratory, Nuclear Energy and Safety Department, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland b Spallation Neutron Source Division, Paul Scherrer Institute, CH-5232 Villigen PSI, Switzerland c UTS Tecnologie Fisiche Avanzate, Sezione Metodologie e Diagnostiche, ENEA, Bologna, Italy Received 25 January 2007; received in revised form 1 February 2007; accepted 5 February 2007
Abstract The paper describes some R&D activities undertaken in support of the design and safe operation of MEGAPIE (MEGAwatt PIlot Experiment) spallation source target, which is scheduled to be irradiated by a proton beam in the SINQ facility at the Paul Scherrer Institute in 2006. The target material is lead bismuth eutectic (LBE), which also acts as the primary coolant. As a consequence of the spallation reactions, about 600 kW of heat would be deposited in the target during operation, and considerable R&D effort is being expended to demonstrate continuing coolability and structural integrity under a variety of operational and abnormal conditions. The paper gives three examples of transient analyses carried out as part of the safety assessment of the target: (1) a beamline trip and recovery; (2) failure of the primary electro-magnetic pump (EMP); (3) failure of the secondary EMP (used to cool the base of the target). The study involves the simultaneous application of a system-analysis code, in our case a version of RELAP5, a computational fluid dynamics (CFD) tool (CFX-4), and a structural analysis code (ABAQUS). The RELAP5 code is used to provide transient boundary conditions for a localized conjugate heat transfer analysis of the lower target region, undertaken using CFD, and includes the feed-back effects arising from the secondary cooling and control systems. A conjugate heat transfer problem is then solved using CFD, which provides time-dependent thermal and flow data within the LBE, together with the thermal and mechanical loads to the target structures. Finally, an in-house interface program is employed to transfer mesh geometry, model topology and (time-dependent) thermal/mechanical data to enable stress analysis of the principal lower-target structural components to be performed. It is demonstrated that none of the transients considered result in critical stress conditions occurring in the target components, but that further operation is not recommended unless both pumps are fully operational. © 2007 Elsevier B.V. All rights reserved.
1. Introduction In the ADS (Accelerator Driven System) concept, neutrons generated from spallation reactions are used to initiate and maintain a continuous chain reaction in a sub-critical nuclear core (Rubbia et al., 1995; Bolderman, 1997). In some ADS designs, the spallation neutrons are generated from proton bombardment of liquid lead bismuth eutectic (LBE), which then acts both as the target material and the means by which heat is removed from the spallation region. A pilot facility at the Paul Scherrer Institute in Switzerland, and supported by an international ∗ 1 2
Corresponding author. Tel.: +41 56 310 27 26; fax: +41 56 310 44 81. E-mail address: [email protected] (B.L. Smith). Tel.: +41 56 310 40 52; fax: +41 56 310 44 81. Tel.: +39 051 609 82 56; fax: +39 051 609 80 62.
0029-5493/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.nucengdes.2007.02.020
partnership, aims to demonstrate the feasibility of the LBE target concept (Bauer et al., 2001). Efforts are coordinated under the project title MEGAPIE, and groups from Belgium, France, Germany, Italy, Japan, South Korea, Switzerland and the USA are involved in the design, manufacture, safety assessment and, finally, operation of the target. As a consequence of the spallation reactions, about 600 kW of heat is deposited (Foucher, 2003), and it is necessary to demonstrate that the target would continue to be cooled adequately, and that structural integrity would be maintained, under a variety of operational and abnormal conditions. It is not possible to fully verify the cooling principle by means of mock-up experiments, because of the difficulty in reproducing the volumetric heating at the appropriate power density (at maximum, 1 GW/m3 for a beam power of 1 MW). Consequently, heavy reliance is placed on detailed numerical analyses. Such analyses
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
are reported here, the procedure utilizing a three-step modelling approach. The system behaviour is first determined using a version of the RELAP5 (1995) code specially modified for use with LBE. This provides transient boundary conditions for a subsequent CFD analysis, in our case using CFX-4 (CFX-4.3, 2000), to determine the thermal and structural loads to the structures. Finally, a stress analysis calculation is performed, here using ABAQUS (2001), to estimate the structural response to these loads. 2. MEGAPIE target assembly The MEGAPIE target, shown in cutaway and schematic form in Fig. 1, consists in the lower section of a cylindrical steel vessel about 2 m high and internal diameter 0.18 m, closed at the bottom by a hemi-spherical cap, or window, and flanged at the top to a larger vessel of 0.22 m diameter housing the annular target heat exchanger (THX). The coolant is LBE, and flow is driven by the main electro-magnetic pump (EMP), located between the THX and a cylindrical guide tube separating the hot and cold streams. LBE at about 230 ◦ C is discharged from the lower end of the THX to the annular space between the guide tube and the target hull, where it descends to the window and rises in the space inside the guide tube towards the main EMP entrance. Here, the now hot LBE is pumped through the inlet manifold of the THX, and down over 12 cooling pins (arranged in a ring)
1657
to complete the LBE circuit. A bypass pump is situated beneath the main pump assembly, and this is used to provide, via suitable ducting, a secondary coolant flow to the window. Down the centre of the target there is an instrumentation/control rod, which also contains an electrical heater for maintaining the LBE temperature during shutdown periods. The LBE target is surrounded by a D2 O-cooled safety hull, which protects the beam tube from possible LBE leaks. The gap between the target and safety hulls is filled with helium at a pressure of 0.5 bar. The entire assembly is mounted vertically, with protons entering the target from below. The protons pass through the safety and target windows and react with the LBE, where the bulk of the spallation neutrons are produced. These are collected in collimators placed around the target and subsequently utilized, particularly for materials research. Detailed CFD calculations (Dury et al., 1999) have shown that the bulk of the heat generated in the target is in the LBE itself (98%), though there is some deposition in the window (0.78%) and guide tube (0.89%) structures, which is of importance locally. The heating of the hull (the cylindrical part of the target outer boundary) due to the beam is much less important (0.18%), and there is no heat deposition in the instrumentation/heater rod, which is situated above the active spallation zone. Very important is the heat transfer from the window to the LBE (about 4.2 kW). Adequate cooling must be guaranteed under all conditions in which the beam is turned on to prevent structural failure. Significant too is the heat transfer from the hot riser to the cool annulus section through the non-insulated guide-tube wall (about 200 kW), since this influences the overall LBE operating temperature. The numerical simulations reported here all follow a threestep analysis strategy: (1) The transient system response is first determined using a RELAP5 model of the target and auxiliary equipment (Leung, 2004; Leung and Smith, 2005), the model including the primary/secondary/tertiary circuits and their control systems. Among other data, the system simulation supplies time-dependent flowrate and temperature data for the LBE entering the annulus and bypass channels to the lower target during transient events. (2) These data are tabulated, and then used as boundary conditions to an accompanying, detailed CFD simulation of the LBE flow in the lower target, and the heat transfer with nearby structures. (3) Geometry, topology and mechanical/thermal loads pertaining to the structures are transferred, via an interface program (Smith et al., 2004), to become input data to the structural analysis code ABAQUS, with which the local stress fields in the components are computed.
Fig. 1. Cutaway and schematic views of MEGAPIE target (configuration with a flat-ended guide tube).
Note that a particular advantage of the present approach is that there is no need to recalculate the heat transfer problem within the structural analysis code. Point-wise thermal data are transferred directly from the CFD simulation, avoiding the need to impose average heat transfer coefficients or heat fluxes.
1658
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
3. The modelling stages We give here three examples of transient analyses which have carried out as part of the safety assessment of the target: (1) a beamline trip; (2) failure of the primary electro-magnetic pump (EMP); (3) failure of the secondary EMP. Beamline trips are expected to occur frequently as a consequence of accelerator problems and/or beam instabilities during the operational lifetime of the target (up to 1 year). The target design (optimized for steady-state operation) is expected to be sufficiently robust to resist over-stressing of components during such incidents. In particular, adequate cooling of the window must be maintained at all times, and under all conditions, including during abnormal operation, and for (postulated) accident sequences. 3.1. System behavior: RELAP5 The MEGAPIE cooling system is of a triple-loop type, employing three working fluids: LBE (primary coolant), Diphyl THT (intermediate oil loop) and light water. The oil loop is necessary to avoid any possible LBE/water contact in the (unlikely) event of cooling pin rupture. The circuits are thermally connected by heat exchangers. A schematic of the system layout used as the basis of the RELAP5 model, showing the target and bypass circuits, the target heat exchanger (THX), intermediate cooling loop, intermediate heat exchanger (IHX), water cooling loop (WCL), and the water heat exchanger (WHX) is shown in Fig. 2. The original model was assembled by Petrazzini and Alemberti (2002), and extended by Leung et al. (2003). CFD analysis of the lower target response to the beamtrip and pump-failure incidents cannot be considered in isolation, as there is thermal coupling with the ancillary circuits, and transient feedback effects resulting from programmed responses of the control system. Results from the RELAP5 simulation are provided in the form of time-dependent mass flowrate and temperature boundary conditions for the LBE inlets to the detailed, stand-alone CFD study. Lack of computer power precludes the option of
Fig. 2. Schematic representation of the system layout used as the basis to the RELAP5 simulation.
using CFD to represent the entire system. Further details of the RELAP5 models are given by Leung and Smith (2005). 3.2. CFD model: CFX-4 Under normal, steady-state operating conditions, the proton beam is at full power, and the LBE pumps and all auxiliary circuits are functioning normally. Thermal-hydraulics simulations are being undertaken to ensure there is adequate cooling in the lower target region where the heat is deposited (Dury, 2004). The lower target geometry is shown in Fig. 3. (The outer, water-cooled safety container seen in Fig. 1 is not modelled.) The simulations being undertaken are aimed at optimizing the geometrical layout, including the shape of the guide tube (flat or slanted at the bottom), and the bypass nozzle area, shape and position. The CFD model for the slanted option is also shown in Fig. 3, extending up to a level just below the exit from the heat exchangers: i.e. to 2.15 m above the lowest point on the internal surface of the target window. The window, made from T91 steel, is 1.5 mm thick at its lowest point, increasing uniformly to 2 mm at the junction with the cylindrical part of the hull. The central rod contains thermocouples and other instrumentation, and heaters to maintain LBE above its melting temperature when the target is in hot standby condition with the beam off. The window, hull, guide tube and central rod are all defined as conducting solids in the CFX4 input data. The model used in this study consists of 113,000 hexahedral fluid cells and 62,000 hexahedral structural elements, the mesh layout in the lower section being reproduced in Fig. 3. The target hull, annulus, riser and instrumentation/heater rod are clearly discernible. The ducting to the bypass nozzle is also modeled, though cannot be seen in this picture, since the duct walls coincide with mesh lines. The bypass nozzle itself is rectangular, of dimensions of 10 mm × 20 mm, and is orientated to provide a stable wall jet over the hottest part of the window. A CFD conjugate heat transfer problem was set up within CFX-4 and run to steady-state to provide the starting conditions for each of the transient simulations. The heat deposited in the LBE and structures is defined in terms of volumetric sources, as described by Dury (2004). The flow of LBE is downwards in the annulus, over the window, and upwards in the central region or riser, with additional window cooling being provided by the bypass jet. The inlet boundary conditions are the mass flowrates of the LBE entering the top of the annulus and the entrance to the bypass duct (37.5 kg/s and 2.5 kg/s, respectively), and the temperatures (230 ◦ C and 242 ◦ C, respectively). (To avoid introducing an excessive number of meshes, heat transfer across the wall of the bypass duct was not taken into account. The increase in temperature of the LBE in the duct from 230 ◦ C at the THX outlet to the 242 ◦ C at the nozzle exit has been estimated using 1D correlations.) The outlet is defined as a constant-pressure surface at the top of the riser. The inlet Reynolds number at the top of the annulus is 81,600, so the flow is turbulent. To take account of the turbulence effects, a high Re k–ε model, with wall functions, has been employed. Care was taken to ensure legitimate use of the logarithmic law of
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
1659
Fig. 3. Lower target geometry, hexahedral mesh layout, and steady-state temperature distribution (configuration with slanted guide tube end).
the wall near non-slip boundaries; that is, by choosing the centre of the cell next to the boundary to be in the range 30 < y+ < 100, expressed in wall units (Launder and Spalding, 1974). Because of the low Prandtl Number for LBE (Pr ≈ 0.02), it may not be appropriate to apply the logarithmic profile for the near-wall temperature field (Kader, 1981). Automatic checks are made in the code, and a linear profile is employed if the first mesh cell is located within the laminar part of the thermal boundary layer; see (CFX-4.3, 2000) for details. The temperature distribution in the target in the vertical plane of the bypass jet is shown in Fig. 3. The spallation region extends from the window up the riser to a height of 27 cm, and the figure shows that the maximum temperature of 382 ◦ C occurs just below this level. The distribution is very non-uniform, and one can expect that the structures nearby will become thermally stressed as a consequence of strong, local temperature gradients. Of special interest is the structural integrity of the window, since this is where the beam enters the target, and which must remain adequately cooled under all conditions to prevent loss of LBE. It can be seen from Fig. 3 that, in the plane of the bypass, supplementary window cooling is being provided by the jet flow to this region. 3.3. Structural solver: ABAQUS ABAQUS (2001) is a commercial, general-purpose Finite Element Method (FEM) program for the solution of many different linear and non-linear structural problems, including stress analysis, heat transfer, mass diffusion and acoustics. In stress analysis, FEM is applied to the principle of virtual displacements, with the stress/displacement elements based on the Lagrangian description of deformation. The degrees of freedom, calculated at the nodes of the element, are the fundamental
variables of the analysis. In ABAQUS, the material response (stiffness) at the integration points within the elements is evaluated for most elements by full or reduced integration (Gaussian quadrature) over the volume of the element. In the present analyses, reduced integration has been used. From the element stiffness matrices, the global stiffness matrix is assembled, and a global system in the unknown degrees of freedom is generated. Linear equation solvers are used for both linear and non-linear analysis. In the latter case, Newton’s method (or a variant of it) is utilized, within which a set of linear equations are solved at each iteration. This is done using direct Gaussian elimination, coupled with a sparse solver based on a “multi-front” technique. All details are contained in the appropriate program documentation referenced above. Here, the mesh discretization used for the structural analysis is adopted directly from the CFD model. Consequently, use is made of linear hexahedral elements to represent the structures, the degrees of freedom being the translations at each of the eight corner nodes. Though most components consist of thin shells, there are five elements across the thickness to ensure adequate representation of the bending moments. Details of the material properties are given in Table 1. For convenience in defining the boundary conditions, each structure is suspended from its top axial section by imposing zero axial displacements to all nodes along the upper edge, and three anchor points on the top rim are constrained to move in the radial direction only. This prevents rigid body movements of the structures, and simulates well the flange attachment in the actual target. Local stress concentrations around the anchor points are ignored. From earlier studies, it was found necessary to have spacers between the guide tube and the target hull to limit lateral displacement of the guide tube resulting from the differential heating. The spacers are simulated in the model by
1660
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
Table 1 Steel material properties used for stress calculations Component Target hull Guide tube a
Material
Young’s modulusa
Poisson ratio
Expansion coefficient
T91 AISI 316
1.51–2.06 × 1011
0.3 0.3
11.5 × 10−6 K−1 16.0 × 10−6 K−1
Pa
2.0 × 1011 Pa
Young’s modulus is a bi-linear fit to temperature-dependent data (Zucchini and Smith, 2003).
Fig. 4. Conditions over the window external surface in steady-state.
constraining three pairs of nodes around the bottom edge of the guide tube and the hull to have the same radial displacement. Fig. 4 shows the temperature and von Mises stress distributions over the external surface of the window. The bypass flow is from left to right, and in this configuration is aligned with the minor axis of the elliptical beam footprint. Maximum temperatures occur near the centre of the window, slightly displaced as a consequence of the cross-flow. The stress distribution cal-
culated using the ABAQUS code is also shown in the figure, with the point of maximum stress also displaced from the centre point of the target. The mesh layout derives from the CFX-4 simulation, and has been carried over, together with the thermal and mechanical (surface pressure) data, to ABAQUS via the interface program, as previously explained. The associated temperatures and stresses on the guide tube inner and outer surfaces are shown in Fig. 5, looking from the lower edge upwards. Again, the temperature distribution is seen to be rather non-uniform. The hot spot at the lower end lies just above the bypass nozzle, and derives from a small vortex in the flow, which becomes established in this region. This is also the point of maximum stress in the guide tube. The differential heating effect up the length of the riser section (also seen in Fig. 3) results in elongation and bending of the guide tube and instrumentation rod, the relative magnitudes of which are also given (in exaggerated form) in Fig. 5. These findings confirmed the need for spacers to be installed, to secure the target components, and became a feature of the design thereafter. Full details are given by Zucchini (2002). 4. Analysis of transients
Fig. 5. Conditions on the guide tube (in steady-state), and relative deformations of the structural components (magnified).
Each of the transients analyzed begins at time t = 0, starting from steady-state conditions. The calculation procedure is the same in all cases. First, a RELAP5 system simulation is carried out for the particular transient under consideration. Histories of the mass flowrates and temperatures at the THX and bypass nozzle exits are tabulated from the run, up to such times as steady operating conditions have been re-established. These become, respectively, time-dependent inlet boundary conditions for the annulus and bypass duct for the transient CFD simulation, which is carried out next. Time-dependent thermal/mechanical data from the CFD run are stored at time-steps of interest, and
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
1661
Fig. 6. Input conditions to the CFX-4 simulation of the lower target derived from the RELAP5 system analysis for the beamtrip transient: (a) beam power, (b) inlet temperatures and (c/d) annulus/bypass mass flowrates.
these become input data for a series of static, stress analysis calculations carried out using ABAQUS. 4.1. Beamtrip and restart Although the SINQ facility at PSI is meant to be a continuous spallation source, the proton beam does not operate continuously because ultra-sensitive detectors employed at the injector from the cyclotron, and along the beam lines, can trigger a shutdown of the beam whenever a fault signal is detected. Two types of beam outages may be discerned: a beamtrip and a beaminterrupt. In the former case, the beam is shut off instantly, followed by a 10 s down time, and a ramp up to full power over 20 s. This event occurs quite frequently—perhaps 50 times a day. In contrast, for a beaminterrupt, the beam is shut off instantly, but recovery is delayed for an extended period, from 30 s, to several minutes, or for a few days. Beam-interrupt events occur much less frequently, perhaps two or three times a week, and may be instigated by the operators: for example, during regular maintenance periods. In such circumstances, the target has to be in a hot standby condition, ready for restart. The present study focuses on the response of the lower target components to the beamtrip and restart event. Details of conditions elsewhere in the cooling system have been described at length in Leung and Smith (2005). Details from the RELAP5 simulation are shown in Fig. 6. For any beamtrip event, the loss of beam power is instantaneous, but the recovery to full power is a controlled ramp over a 20 s period. This is an enforced operating practice, to avoid thermal shocks. The inlet temperature histories (Fig. 6b) generally follow that of the power, but are smoothed by the continuing heat transfer processes taking place between the LBE, the still hot structures, and the cooling pins in the THX. Though the pumps remain in operation, the loss of the buoyancy-driven part of the flow circu-
lation is seen in the reduced flowrates to the annulus (Fig. 6c) and bypass (Fig. 6d) inlets. As a consequence of the thermal inertia of the system, full recovery to steady-state operation takes altogether about 60 s, to be compared with the power transient of 30 s. The transient mass flowrates and temperatures given in Fig. 6 are used as boundary conditions for the accompanying CFD simulation of the lower target region. Fig. 7 illustrates the temperature response at the centre of the window and at the point of maximum temperature on the guide tube, obtained from the CFX-4 simulation. Following beam shut-off (at t = 0), the window, which is still being cooled by the bypass LBE flow, loses heat very rapidly. The minimum temperature occurs 10 s into the transient, at the time the beam power starts to be ramped up. Because of the slow thermal response of the target system, about 60 s are required for conditions to return to steady-state. The guide tube, which is the most stressed component in the lower target region, is not subjected to such strong thermal transients due to the heat capacity of the LBE.
Fig. 7. Temperature responses in the lower target region for the beamtrip transient, as determined using CFX-4.
1662
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
Fig. 8. Time histories of (a) the maximum temperature and (b) the stress history in the lower-target structures for beamtrip transient.
Time-dependent thermal data for the lower-target structural components, derived by CFX-4, were passed to the ABAQUS code for the stress analysis. Results are presented in Fig. 8. As expected, minimum temperatures in the structures are seen at the end of the beam-off period (t = 0–10 s), and then gradually increase as the beam is ramped up to full power (t = 10–30 s). Under normal operation, the beam deposits 4.2 kW of heat in the window, which is transferred to the LBE at the inner, cooled surface; the outer surface is assumed adiabatic, because of the poor heat-transfer properties of the helium in the gap between the target and safety hulls. The temperature difference from the outer to inner surface of the window (about 21 ◦ C) sustains the heat flux at the inner surface necessary for this heat transfer process. Following the beamtrip, the heat source is lost, and the temperature becomes uniform through the window thickness, as seen in Fig. 8a at t = 10 s. During the subsequent 20 s ramp to full power, the temperature gradient is re-established monotonically. Fig. 8b shows that no stresses are created in the lower target structures, and conditions return to their equilibrium states without any transient stress crises.
LBE container during this transient, and to evaluate the potential for target operation by natural circulation only, perhaps at reduced power. The initial, steady-state condition is that pertaining to normal operation of the target. The pump failure incident begins at t = 0. According to the system calculation (Leung and Smith, 2005), the mass flowrate to the annulus falls sharply from the initial 37.5 kg/s, and then recovers partially to about 23 kg/s, driven by the buoyancy of the heated LBE in the riser (Fig. 9a). The bypass flow is hardly affected, under the assumption that the bypass pump continues to function normally (the double-pump failure scenario is of lower probability, and will be analysed at a later date, if requested by the safety authorities). The temperature at the top of the annulus falls, due to the enhanced cooling of the LBE by the THX resulting from the reduced flow in the primary circuit (Fig. 9b), while full flow is maintained in the intermediate cooling loop. The temperature drop triggers operation of the ICL three-way valve (Fig. 2) at t = 10 s, isolating the THX. This leads to temperature recovery in the primary LBE circuit (Fig. 9b). Further operation of the ICL control system leads ultimately to new steady-state operating conditions. The monotonically-increasing nozzle exit temperature (which in the RELAP5 calculation includes the effect of heat transfer to the bypass duct from the LBE in the annulus) reflects the establishment of generally hotter operating conditions overall in the target due the reduced LBE circulation rate. New equilibrium conditions (at the reduced flowrate) are established after about 150 s. Using these data, a transient, conjugate-heat-transfer, CFD simulation was performed to determine the flow and thermal
4.2. Failure of main pump In the event of failure of the main EMP, the flowrate in the primary circuit reduces very quickly, because of the low-inertia and high-resistance characteristics of the loop, but partial flow is maintained as a consequence of the buoyancy induced by the beam heating. For the lower target region, where the heating takes place, it is necessary to ensure that there is no risk to the
Fig. 9. Input conditions to the CFX-4 simulation for the main-pump-failure transient.
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
1663
Fig. 10. Velocities and temperature contours 5 s into the main-pump-failure transient.
conditions in the lower target region. At regular intervals (and other times of particular interest), thermal and mechanical load data were stored for the subsequent structural analysis. The simulation was run for 100 s, which is close to the new steady-state operating point. Fig. 10 shows conditions 5 s into the pump-failure transient. From the velocity field, one notices that the bypass jet has penetrated into the opposite annulus. This is because the main flow has been reduced, and flow conditions are no longer optimal over the window. However, the window itself remains adequately cooled, since the bypass flow has not been affected by the loss
of the main EMP. That is, the velocity of 1.43 m/s at the bypass nozzle exit remains as before, the wall jet being the controlling mechanism for regulating the window temperature. However, the non-optimized flow field has an adverse effect on the cooling of the target hull. The bypass flow picks up heat from the window, and directly from the beam, and transports it to the opposite annular gap (instead of to the riser region), producing a hot spot on the sidewall. This is expected to result in unwelcome local stresses in the material, as later confirmed by the structural analysis. Already, the maximum temperature in the LBE has increased by 70 ◦ C, and the proximity of the maximum to the
Fig. 11. Velocities and temperature contours 30 s into the main-pump-failure transient.
1664
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
Fig. 12. Maximum temperature histories in the LBE and on structural components for the main-pump-failure transient.
lower end of the heater rod will mean increased temperatures in the material, stronger temperature gradients, and increased thermal stresses. The situation 30 s into the transient is shown in Fig. 11. In the plane of the bypass jet, the upflow region in the annulus has reached a height of more than 27 cm, with accompanying increase in temperature. The region of strong beam heating is much narrower (cf. Fig. 3), and the maximum LBE temperature has now increased from the estimated 387 ◦ C under normal operating conditions to 600 ◦ C. Maximum temperatures in the window, guide tube, heater rod and LBE are given in Fig. 12 for the first 80 s of the transient. There is a strong heating up of the lower part of the target (except the window) over the first 15 s, followed by a more gradual return to steady-state operating conditions, but one in which the main flow is now buoyancy driven. The window remains cooled by the bypass flow, with the mean of the maximum temperature staying constant at about 375 ◦ C. However, the maximum is seen to fluctuate, with many short-lived peaks. It should be remembered that maximum window temperatures are plotted here rather than the
temperature evolution at a given point, and the erratic behaviour may not reflect any local physical development. The frequency of the fluctuations is in any case too high to be of consequence to the stress history in the window material. The maximum LBE temperature increases by about 180 ◦ C, and is also oscillatory after 35 s. Maximum heater rod and guide tube temperatures increase by about 100 ◦ C and 125 ◦ C, respectively, the latter displaying the same oscillatory behaviour as the LBE. In contrast to the seemingly random fluctuations over the window, however, the oscillations in the LBE and on the guide tube are more regular, and are most likely caused by instabilities in the downflow region on the narrow-gap side of the guide tube into which the bypass jet penetrates, though more detailed analysis would be needed to confirm this. The temperatures and stresses over the outer surface of the window are given in Fig. 13 at different times in the transient. The maximum window temperature is slightly displaced from the centre as a consequence of the cross-flow induced by the bypass jet (indicated by the arrow), which, by assumption, still functions normally. The von Mises stress pattern changes with time, but there are no severe stress concentrations. Maximum stresses for the lower-target structures at various times are summarized in Table 2. Stresses remain more-or-less constant over the window, as previously noted, but increase markedly for the guide tube (from 49 MPa at t = 0 to 181 MPa at t = 30 s) and for the cylindrical part of the target vessel (from 45 MPa at t = 0 to 90 MPa t = 20 s). Though large, these stresses do not pose threats to the structural integrity of the components: the yield stress for T91 steel (window and target hull) at 400 ◦ C is 460 MPa, and that for SS316 (guide tube) at the same temperature is 190 MPa. 4.3. Failure of bypass pump The bypass pump has a particular function in providing a continuous jet stream of LBE across the window, and stabilizing the flow in this region. Sudden failure of the pump is not expected to represent a catastrophic event, because the slanted design of
Fig. 13. Temperatures (top) and von Mises stresses (bottom) over the window outer surface at: (a) t = 0, (b) t = 5 s and (c) t = 30 s into the main-pump-failure transient; the arrow indicates the bypass flow direction.
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
1665
Table 2 Maximum von Mises stresses at various times during the pumptrip transient
Table 3 Changes in input conditions for bypass trip transient
Time (s)
Inlet to
0 1 2 3 4 5 10 15 20 30 40 60 80
Maximum von Mises stress (MPa) Guide tube
Window
Target vessel
49.1 57.5 54.2 59.0 65.0 70.8 118.8 149.1 163.0 180.9 140.0 155.1 175.5
44.6 47.5 43.9 43.2 43.0 47.1 41.6 44.8 45.1 46.4 44.3 45.3 47.8
44.6 47.5 43.9 43.2 43.0 47.1 41.6 61.7 89.7 55.4 53.0 45.3 60.2
the guide tube would still maintain a cross-flow over the hottest part of the window. However, the transient still requires analysis to give input to the decision whether to continue operation of the target. According to the RELAP5 simulation (Leung, 2004), in the event of failure of the bypass EMP, with the main EMP operating normally, the flowrate in the bypass duct decreases very quickly (low inertia), but maintains a bleed flow at the nozzle of about 0.9 kg/s. Due to the heating effect of the pump, LBE enters the top of the duct at around 260 ◦ C, cools as a result of heat exchange with the LBE in the annulus at 230 ◦ C, but then heats up again because heat transfer from the riser to annulus region increases the temperature of the LBE in the annulus gap with respect to that in the duct. According to the RELAP5
Annulus Bypass
Temperature (◦ C)
Mass flowrate (kg/s)
t<0
t>0
t<0
t>0
230 242
230 254
37.5 2.5
39.9 0.9
simulation, this complex interaction results in the exit temperature at the nozzle being raised to 254 ◦ C, and this value is used for the subsequent CFD simulation. Since the temperatures and flowrates for the main and bypass circuits change so abruptly, simple step-changes in the input conditions for the CFD calculation suffice to define a worst-case scenario. Details are given in Table 3. The CFD simulation has been run for 100 s, which is close to the new operating point. Fig. 14 shows the temperature field in the target 5 s after the start of the transient; the temperature scale is kept the same as that in Fig. 3 in order that comparisons may be made. The cross-flow over the window is still evident in Fig. 14a, which is the plot taken in the plane of the bypass flow; the flow remains more-or-less symmetric in the plane perpendicular to this (Fig. 14b). The higher main flow reduces the region of maximum LBE temperature almost to a point, just above the spallation region. In fact, the highest temperature in the target is now at the window, which peaks at this time, 140 ◦ C above normal, steady-state conditions. Time histories of the maximum temperatures in the LBE, and on the window and guide tube surfaces, are shown in Fig. 15. The maximum window temperature peaks at t = 5 s, then stabilizes after t = 10 s. The final temperature is 115 ◦ C above normal. The
Fig. 14. Temperature contours 5 s into the bypass-pump-failure transient (a) in the plane of the bypass jet and (b) perpendicular to it.
1666
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
Fig. 15. Time-plots of maximum temperatures for bypass-pump-failure transient.
maximum LBE temperature at first falls sharply, then also peaks at t = 5 s, stabilizing at t = 10 s. The final (maximum) temperature is the same as under normal operating conditions. The guide tube maximum temperature also falls initially, but then recovers, and thereafter remains constant, about 30 ◦ C above normal. There is evidence of oscillatory behaviour in the lower target region, with period of about 15 s, beginning 30 s after transient initiation. Careful inspection of the flow and temperature field over the window has revealed that the instability is most likely buoyancy-driven, due to overheating at the centre of the window
and periodic lifting off of the attached boundary layer, which the bypass flow is no longer able to stabilize. The principal results from the subsequent ABAQUS calculation are given in Fig. 16. During the first 5 s, the window temperature increases sharply by 180 ◦ C, due to the loss of the direct cooling effect of the bypass jet, and then stabilizes at about 140 ◦ C above nominal operating conditions. The maximum guide tube temperature does not peak, but is nonetheless elevated to 400 ◦ C, which is 60 ◦ C higher than normal. During the phase of rapid temperature increase in the first 5 s, the maximum von Mises stress on the window peaks at 118 MPa on the outer surface (in steady-state, the maximum stress was 44 MPa on the inner window surface), but then stabilizes at 90 MPa. The maximum stress on the inner surface exhibits similar behaviour, with a peak at 98 MPa, also stabilizing to 90 MPa. These stresses are well below the yield stress of T91 at 500 ◦ C of 400 MPa, so the structural integrity of the window is not threatened during this transient. The most dramatic change is the maximum stress in the guide tube, which increases from 49 MPa to 167 MPa, which is approaching the yield stress for SS316 of 190 MPa at 400 ◦ C. However, the stressed region is highly localized, and the overall integrity of the guide tube remains secure. 5. Conclusions CFD analyses for the lower part of the MEGAPIE spallation source target have been carried out using CFX-4 to simulate conditions resulting from one operational transient (beamtrip) and two abnormal transients (failure of main or bypass pump). In all cases, transient inlet conditions have been employed, these having been obtained from system analyses of the different incidents using a version of the RELAP5 code specially modified for use with liquid metals. The CFD model adopted is that used previously for steady-state analysis of the target, here used in transient mode. At regular time intervals, and at other times at which quantities were changing rapidly, structural thermal data were transferred to the ABAQUS code, and stress analysis of structural components was carried out. Main observations from the analyses are listed below:
Fig. 16. Time-plots of maximum temperatures and stresses in structures for bypass-pump-failure transient.
• The beamtrip and power-up transient does not lead to either excessive temperatures or excessive stresses in the lower target structures. Though the beam heating is lost instantaneously, there is a more moderate cooling of the structures, due to their heat capacities. In this scenario, the beam stays off for 10 s, and then full power is restored gradually over a 20 s period. No transient peaks occur during this process, and temperatures and stresses monotonically return to their equilibrium values. • For the main-pump-failure incident, mean temperatures over the window, through which the proton beam passes, remain constant, controlled by the bypass flow, which is assumed to be still functional. Elsewhere, temperatures rise rapidly: by more than 100 ◦ C for the heater rod and guide tube, and by nearly 200 ◦ C in the beam-heating region of the LBE. The flowrate in the annulus drops from the nominal value
B.L. Smith et al. / Nuclear Engineering and Design 237 (2007) 1656–1667
of 37.5 kg/s to that driven by buoyancy alone: around 23 kg/s. The flowrates are no longer optimal for the target geometry: the bypass jet overshoots the window, penetrating into the annulus on the opposite side, and resulting in locally higher sidewall temperatures, and most likely initiating flow instabilities. The incident does not represent a threat to the structural integrity of the target, but further operation at full power warrants careful consideration, given the changes in flow conditions and the temperature increases seen in these calculations. • In the case of failure of the bypass pump, the RELAP5 calculation predicts that the velocity of the LBE jet at the nozzle would fall from the nominal 2.5 kg/s to a bleed flow of 0.9 kg/s. According to the CFD simulation, this is too weak to traverse the window and maintain adequate cooling, and the window temperature increases, finally stabilizing 115 ◦ C above normal. The maximum temperature in the LBE is little changed from that under normal operating conditions, though does peak during the transient. The detailed ABAQUS calculation reveals that though the maximum temperature on the guide tube increases by only 30 ◦ C, it occurs at the narrowgap end of the tube, and results in an undesirable increase in the maximum stress on the structure, from about 50 MPa to 170 MPa. There is also evidence of unstable flow behavior. Thus, although the slant on the guide tube ensures a good cross-flow over the window, even in the absence of the bypass flow, overall conditions are far from optimal, and any further irradiation of the target, even at reduced power, would not be advisable. References ABAQUS/Standard User’s Manual I, II, III, 2001. Hibbit, Karlsson and Sorenson Inc., Providence, Rhode Island, USA (www.abaqus.com). Bauer, G.S., Salvatores, M., Heusener, G., 2001. MEGAPIE, a 1 MW pilot experiment for a liquid metal spallation target. J. Nucl. Mater. 296 (1–3), 17–33.
1667
Bolderman, J.W., 1997. Accelerator-driven nuclear energy systems. Australian academy of technical sciences and engineering. In: Academic Symposium: Energy for Ever: Technical Challenges of Sustainable Growth, November 1997. CFX-4.3, 2000. AEA Technology, Harwell, Oxfordshire, OX11 0RA, UK (www.ansys.com). Dury, T.V., 2004. CFD design support at PSI for the international MEGAPIE liquid-metal spallation target. J. Nucl. Sci. Technol. 41 (3), 285–295. Dury, T.V., Smith, B.L., Bauer, G.S., 1999. Design of the European spallation source liquid-metal target using computational fluid dynamics. Nucl. Technol. 127, 218–232. Foucher, Y., 2003. Etude et D´eveloppement d’une Cible de Spallation. Ph.D. Thesis. University of Nantes, France. Kader, K.A., 1981. Temperature and concentration profiles in fully turbulent boundary layers. Int. J. Heat Mass Transfer 24 (9), 1541–1544. Launder, B.E., Spalding, D.B., 1974. The numerical computation of turbulent flows. Comp. Meth. Appl. Mech. Eng. 3, 269–289. Leung, W.H., 2004. On the MEGAPIE target thermal hydraulics—a RELAP5 analysis. In: Proceedings of the 12th International Conference On Nuclear Energy (ICONE 12), Arlington, Virginia USA, April 25–29, 2004. Leung, W.H., Smith, B.L., 2005. The MEGAPIE spallation source target: analysis of operational and accident transients using RELAP5. In: Proceedings of the 11th International Conference on Nuclear Reactor Thermal Hydraulics (NURETH 11), Paper 509, Avignon, France, October 2–6, 2005. Leung, W.H., Petrazzini, M., Alemberti, A., 2003. RELAP5 analysis of the MEGAPIE target. In: Proceedings of the 4th MEGAPIE Technical Review Meeting, Paris, March 18–19, 2003 (FZK Report FZKA 6876, December 2003). Petrazzini, M., Alemberti, A., 2002. Input and RELAP5 Model Description. Ansaldo Report MPIE 1-TRIX 200, Genova, Italy, September 2002. RELAP5/MOD3 Code Manual, 1995. Code Structure, System Models, and Solution Methods, vol. 1. NUREG/CR-5535. Rubbia, C., et al., 1995. Conceptual Design of a Fast-Neutron Operated HighPower Energy Amplifier. CERN/AT/95-44(ET). Smith, B.L., Dury, T.V., Ni, L., Zucchini, A., 2004. A pragmatic coupling strategy between commercial CFD and structural analysis codes. In: ASME/JSME/PVP 5th International Symposium on Computational Technologies, San Diego, USA, July 24–30, 2004. Zucchini, A., 2002. MEGAPIE: Stress Analysis of Lower Target; Preliminary Design. ENEA Report No. RTI/FIS/MET/2002/1. Zucchini, A., Smith, B.L., 2003. Stress analysis of the LMC under accident conditions. In: Proceedings of the 4th MEGAPIE Technical Review Meeting, Paris, March 18–19, 2003 (FZK Report FZKA, 6876, December 2003).