The first Domestic Standard Problem (DSP-01) exercise using ATLAS was completed successfully by holding a final workshop in April, 2010. Though the DSP-01 was the first cooperative program for code validation based on an integral effect database, most major domestic organizations, including industries, universities and research institutes, volunteered to contribute to strengthening the technical infrastructure for code validation and to expanding a domestic cooperative network. Technical information sharing and discussion was active between experienced code users. In particular, the water levels of the reactor core and downcomer regions, ECC bypass rate, multidimensional phenomena in the downcomer region, loop seal clearing phenomena, and loop flow characteristics were identified as the crucial phenomena for a close investigation from the viewpoint of code modeling. There were unexperienced code users among the participants and they benefited greatly from this valuable course of exercise. In conclusion, the DSP-01 was a major landmark in the validation of the thermal-hydraulic safety analysis codes. More details can be found in the final comparison report [1] and related paper [2].

A second Domestic Standard Problem (DSP-02) exercise was launched in July, 2010. In the kick-off meeting, the outcome of the DSP-01 was analyzed and discussed by the participants. Noticeable major outcomes were that the DSP-01 provided an opportunity for major domestic nuclear organizations to pursue MARS-KS code validation against qualified IET data and to provide a cooperation network. The know-how and expertise of experienced code users were spread among participants. On the contrary, however, user effects were remarkable owing to great differences in code experience among the code users, and this caused the user effects overshadow the possible code deficiencies. In particular, qualification of the code initialization was highlighted to ensure correct transient calculations. Another lesson from the DSP-01 was that we need to focus on detailed thermal-hydraulic phenomena rather than see only the overall aspects that contribute to a practical code validation. Thus, it was suggested and agreed upon that each participant was responsible for providing an additional analysis on at least one special topic in the DSP-02.

Among the several test scenarios carried out by KAERI, a 6-inch cold leg break SBLOCA test was proposed and accepted as a target test item of the DSP-02. Most participants agreed that the 6-inch cold leg break SBLOCA would be interesting in view of a practical safety analysis. In addition, special assessment topics relevant to code validation were proposed such as 1) 2D behavior in the downcomer, 2) break flow modeling, 3) 3D behavior in the reactor pressure vessel, 4) loop seal clearing, 5) ECC bypass, 6) heat loss modeling, 7) reactor pressure vessel bypass, and so on. In the kick-off meeting, a draft specification of the DSP-02 [3] was distributed along with an updated facility description report (FDR) [4].

A best-estimate safety analysis methodology for small break LOCAs including the DVI line break accidents needs to be developed to identify the uncertainties involved in the safety analysis results. Such a best-estimate safety analysis methodology will contribute to defining a more precise specification of the safety margin and thus will lead to a greater operational flexibility. However, such an effort was lacking because the available integral effect test data were not sufficient.

The current DSP-02 aims at the following:

° Effective utilization of an integral effect database obtained from ATLAS.

° Expanding cooperation network among the domestic nuclear industry, academic institutes, research institute, and regulation organizations.

° Safety analysis technology transfer to inexperienced code users.

° Better understanding of thermal hydraulic phenomena in the upper annulus downcomer region during the DVI injection period of SBLOCAs.

° Investigation of the possible limitation of the existing best-estimate safety analysis codes.

The ATLAS DSP-02 program, like the DSP-01, was organized in collaboration with KINS. KAERI was responsible for a general coordination of the DSP-02, data provisioning, information on the ATLAS facility and DSP-02, code calculation, receipt of submissions, result comparisons, progress meetings, final workshop, and comparison report. As a joint operation agency, KINS was responsible for coordination support, code calculation, progress meetings, and the final workshop.

In the second ATLAS DSP program (DSP-02), thirteen organizations eventually signed an agreement as listed in Table 1. Each signatory organization had an obligation to perform an open calculation within the exercise period using the test results provided by the operating agency, KAERI. Most participants used the best-estimate system code, MARS-KS, and a few participants used the RELAP5 code, and all the participants were also requested to write their analysis results including special assessment topics in an assigned section of the final comparison report [5]. Unfortunately, an organization did not manage to finish its calculation owing to a shortage of manpower, as shown in Table 1.

In the previous DSP-01 program, significantly great user effects were found among the participants. In fact, such great user effects were caused by differences in facility modeling and differences in experiences between skillful and inexperienced code users. Such user effects made it difficult to find the code deficiencies independently. Thus, significant effort has been placed on minimizing the user effects and on assessing the code prediction capability itself based on experimental data.

In the DSP-02, it was agreed that each participant was responsible for performing a sensitivity analysis on at least one special code assessment topic. This activity is expected to help find the code deficiencies and obtain user guidelines that can minimize the “user effects.” As the MARS-KS code was utilized by most participants, this special code assessment activity provided an intensive code assessment environment. Eight code assessment items were defined at the beginning of this program by a discussion among the participants: (1) break flow modeling, (2) loop seal clearing behavior, (3) ECC bypass, (4) RPV bypass, (5) heat loss effects, (6) momentum effects of DVI nozzle, (7) 2-D behavior in the downcomer region, (8) and others, and assigned them to each participant, as shown in the remarks of Table 1.

ATLAS is a thermal-hydraulic integral effect test facility for evolutionary pressurized water reactors of APR1400 and OPR1000. The reference plant of ATLAS is the APR1400, which is an advanced power reactor developed by the Korean industry and has a rated thermal power of 4000 MW and a loop arrangement of 2 hot legs and 4 cold legs for the reactor coolant system [6,7]. ATLAS also incorporates some specific design features of the Korean standard nuclear power plant, the OPR1000, such as a cold-leg injection mode for a safety injection and a low pressure safety injection mode. ATLAS can be used to investigate the multiple responses between the systems for a whole plant or between the subcomponents in a specific system during anticipated transients and postulated accidents.

ATLAS has the same two-loop features as the APR1400 and is designed according to the well-known scaling method suggested by Ishii and Kataoka [8] to simulate various test scenarios as realistically as possible. It is a half-height, 1/288-volume scaled test facility with respect to the APR1400. The main motive for adopting the reduced-height design is to allow for an integrated annular downcomer where the multidimensional phenomena can be important in some accident conditions with a DVI operation. According to the scaling law, the reduced height scaling has time-reducing results in the model. For a one-half-height facility, the time for the scaled model is

times faster than the prototypical time. The friction factors in the scaled model are maintained the same as those of the prototype. The hydraulic diameter

of the scaled model is maintained the same as that of the prototype to preserve the prototypical conditions for the heat transfer coefficient. Major scaling parameters of ATLAS are summarized in Table 2.

The configuration of ATLAS for a cold leg SBLOCA consists of a primary system, a secondary system, a safety injection system, a break simulating system, a containment simulating system, and auxiliary systems as shown in Fig. 1. The primary system includes a reactor vessel, two hot legs,

four cold legs, a pressurizer, four reactor coolant pumps, and two steam generators. The secondary system of ATLAS is simplified to be of a circulating loop type. The steam generated at two steam generators is condensed in a direct condenser tank, and the condensed feedwater is again injected into the steam generators. Most of the safety injection features of the APR1400 and OPR1000 plants are incorporated into the safety injection system of ATLAS. It consists of four safety injection tanks (SITs), a high pressure safety injection pump (SIP) that can simulate a safety injection and long term cooling, a charging pump for charging an auxiliary spray, and a shut down cooling pump and a shutdown heat exchanger for low pressure safety injection, shutdown cooling operation, and recirculation operation. The break simulation system consists of several break simulating lines such as LBLOCA, a DVI line break LOCA, SBLOCA, SGTR, MSLB, and FLB. Each break simulating line consists of a quick opening valve, a break nozzle, and instruments. It is precisely manufactured to have a scaled break flow through it, as shown in Fig. 2. The containment simulating system of ATLAS has a function of collecting the break flow rate and maintaining a specified back-pressure to simulate a containment atmosphere. In addition, ATLAS has some auxiliary systems such as a makeup system, a component cooling system, a nitrogen/ air/steam supply system, a vacuum system, and a heat tracing system. The detailed design and description of the ATLAS development program can be found in reference 9.

In the present experimental test, SB-CL-09, a small break LOCA was assumed primarily to occur at a cold leg-1A piping located between the outlet of the RCP-1A and the corresponding RV inlet nozzle. In addition, a single-failure of a loss of a diesel generator, resulting in the minimum safety injection flow to the reactor pressure vessel, was assumed to occur in concurrence with the reactor trip. Therefore, the SI water from the SIP is only

injected through the DVI-1 and -3 nozzles and the SI water from the SIT is injected through all of the DVI nozzles.

2.2.1 Actual Test Conditions

The present test conditions were determined by a pretest calculation with a best-estimate thermal hydraulic code, MARS 3.1. First, a transient calculation was performed for the 6-inch SBLOCA of the APR1400 to obtain the reference initial and boundary conditions. A best-estimate safety analysis methodology, which is now commonly accepted in nuclear industries, was applied to the transient calculation of the APR1400. A single failure assumption for a safety injection system was assumed in the MARS calculation; four SITs and two of the SIPs were utilized during the test period. The initial and boundary conditions were obtained by applying the scaling ratios, shown in Table 2 to the MARS calculation results for the APR1400. Table 3 compares the rated steady-state condition between the APR1400 and ATLAS for the SBLOCA test.

There are four bypass valves connected to the downcomer in the ATLAS. Two bypass valves of FCV-RV-37 and FCV-RV-38 are between the downcomer and the upper head, and two bypass valves of FCV-RV-95 and FCVRV-96 are between the downcomer and hot legs. The two bypass valves of FCV-RV-37 and FCV-RV-38 should be opened by 74% and 65%, respectively, to provide the required flow rate of 0.25 kg/s each, and the two bypass valves of FCV-RV-95 and FCV-RV-96 should be opened by 81% and 97%, respectively, to provide the required flow rate of 0.71 kg/s each.

2.2.2 Test Procedure

Prior to a transient test, several actions were taken. They included an instrument calibration with the ATLAS system drained, purging and filling the ATLAS system including leakage tests, an instrument calibration with the water-filled primary system, and an implementation of test specific control logics into the process control computers for sequence control. The sequence control logics

executed the required control actions for the corresponding control devices such as the main core heater, RCP, SIP, and valves.

Reaching a specified initial condition of the whole system for the test, as shown in Table 4, the steady-state conditions of the primary and secondary system were maintained for more than 30 minutes. After this steadystate period, the main test started by an opening of the break simulation valve, OV-BS-06. With the start of the test, the primary system pressure decreased rapidly below 10.7 MPa, which was the set-point of the low-pressurizer pressure (LPP) signal. When the LPP signal occurred, the RCP and pressurizer heater stopped, and the main feed water isolation valves and the SIP were actuated with specified delay times. Further decreasing the primary pressure to below 4.03 MPa resulted in a passive injection of the SIT water. Tables 5 and 6 show the scheduled sequence of events and the actual progress of the events observed in the SB-CL-09.

2.2.3 Sequence of Events

After an opening of the break simulation valve, OVBS- 06, the test sequence was controlled by the corresponding control logic, which defined the set-point and related time delay, as shown in Table 5. When the pressurizer pressure measured by PT-PZR-01 decreased below 10.7 MPa, an LPP signal was issued. After the LPP signal, RCPs and pressurizer heater were tripped with no time delay, and the reactor was tripped with a 0.354 s delay. The closing of the main feed water isolation valves and the actuation of the SIPs was scheduled to occur with 7.08 s and 28.29 s delays from the LPP signal, respectively. The SIT injection was initiated by the low upper downcomer pressure (LUDP). The actual sequence of events during the SB-CL-09 was presented in Table 6.

2.2.4 Core Power and Heat Balance

The initial steady-state heat balance was checked by comparing the applied core power with the heat removal rate through two SGs calculated by steam flow rates and temperature differences between the feed water and the main steam. Fig. 3 shows a heat balance at a steady state condition of the SB-CL-09. The total applied power was about 1,633 kW and the total removal heat energy through the two SGs was up to about 1,471 kW. There is a 162 kW difference between the applied and removed power owing to the heat loss of the system, including the primary and secondary systems.

The heat loss from the primary system into the environment cannot be completely prevented even though thick insulation materials envelop the reactor pressure vessel and the primary pipelines. The heat loss was estimated by the following simplified empirical correlation [4];

where, T_W is the outer wall surface temperature measured at the middle of the reactor pressure vessel, and T_atm is atmospheric temperature measured on the test day.

The heat loss from each steam generator into the environment was also estimated according to the empirical correlation [4].

where, T_W is averaged values of the temperatures measured at the outer wall surface of the steam generators, and T_atm is the atmospheric temperature.

The estimated heat losses through the primary and secondary systems are about 87 kW and 62 kW, respectively. The estimated heat loss through the system can be seen in Fig. 4.

2.2.5 Identified Thermal Hydraulic Phases during the SB-CL-09

Compared with a large break LOCA, the phases of

the small break LOCA prior to core recovery occurred over a long period. To identify various phenomena, a small break LOCA can be divided into five phases: a blowdown, natural circulation, loop seal clearance, boiloff, and core recovery [10]. The duration of each phase depends on the break size and performance of the ECCS. In the SB-CL-09, the five characteristic phases were clearly identified, and they were compared with the characteristic parameters such as the primary pressure (PT-PZR-01), secondary pressure (PT-SGSD1-01), and collapsed water level of the downcomer (LT-RPV-04A) in Fig. 5. Detailed

descriptions on the phase separation during the cold leg break SBLOCA case can be found in the literature [10]. In this report, these five characteristic phases of SBLOCA will be described with the experimental results.

The blowdown phase started with the opening of the break valve, OV-BS-06. Upon initiation of the break at the bottom of the cold-leg, the RCS primary side was rapidly depressurized until a flashing of the hot coolant into steam began. A reactor trip and an RCP trip were initiated by the low pressurizer pressure (LPP) setpoint of 10.72 MPa. Closure of the condenser steam dump valves isolated the SG secondary side. As a result, the SG secondary side pressure increased up to the main steam safety valve (MSSV) set point of 8.1 MPa, and steam was released through the MSSV. The ECCS actuation signal was also generated at the LPP signal with a time delay. In this phase, coolant in the RCS remained in the liquid phase. The rapid depressurization ended when the pressure fell to just above the saturation pressure of the SG secondary side. The break flow in the RCS was single-phase liquid throughout the blowdown period.

When the blowdown phase ended, a two-phase natural circulation phase was established in the RCS loops with the decay heat removed by a heat transfer (condensation and convection) to the SG secondary side. As more coolant was lost from the RCS through the break, steam accumulated in the downhill side of the SG U-tubes and the crossover leg. The natural circulation phase might continue until there is insufficient driving head on the cold leg side of the loops, owing to the accumulation of steam in loops between the top of the steam generator tubes and the loop seals. With the loop seals present, the break remained covered with water. The RCS water inventory continued to decrease and the steam volume in the RCS increased. The relative pressure in the core increased, which, together with the loss of coolant inventory through the break, caused liquid levels in the core and the SG to continue to decrease. The behavior of the collapsed water levels of the downcomer and the core region can be observed in Fig. 6, which shows that the core mixture level dropped below the top of the core. However, the cladding experienced no temperature excursion. A differential pressure between the downcomer and the upper head region (DP-DCUH1-01) reached its maximum value during this instant, as presented in Fig. 7.

When the liquid level of the downhill side of the SG was depressed to the elevation of the loop seals, the seals were cleared and steam in the RCS was vented to the cold legs. The break flow changed from a low quality mixture into primarily steam. This relieved the back-pressure in the core, and the core liquid level was restored to the cold leg elevation by a flow from the downcomer.

After the loop seal clearance, the RCS primary side pressure started to decrease below that of the secondary side, as can be seen in Fig. 5. This is mainly due to the increase of the break flow quality, resulting in a lower mass flow rate but a higher volumetric flow through the break. The collapsed water (or mixture) levels of the downcomer and the core region decreased a little bit as a result of the core boiling in this phasedue to the fact that the RCS pressure is too high for the safety injection from the SIT to make up for the boil-off rate.

The core recovery phase started with the instant of SIT injection. The vessel mass inventory increased and the core recovery was established, as can be observed in Fig. 6. In a small break LOCA case, the SIT injection into the RPV starts before the reactor coolant is completely

discharged into the containment vessel, and the RCS pressure is still above the containment pressure.

The flow direction in the four cold legs during the present experiment can be observed in Fig. 8. With the opening of the break valve at 204 s, a reversed flow direction in cold leg-1A was observed. On the other hand, in the other cold legs, the flow rate was largely increased. This flow reversal in the broken cold leg and the flow rate increasing phenomena is typical for the cold leg break SBLOCA case. As can be seen in Fig. 8 (b), the flow rate in the broken cold leg (CL-1A) showed a very fluctuating nature. Moreover, from a more detailed observation, it can be observed that the flow direction in cold leg-1A was highly changeable with time.

The submitted transition calculation results were qualitatively compared with the measured data. All compared figures are included in Appendix A of reference 5. A qualitative prediction accuracy of the submitted calculations was discussed focusing on the important thermal-hydraulic parameters which have high relevance to safety, e.g. pressure and differential pressure, loop fluid temperature, downcomer fluid temperature, downcomer water level, core water level, loop seal clearing, etc., as described in section 5.3 of reference 5.

It is noteworthy that a CPU time comparison is discussed in this section. The CPU time and the time step were requested to examine the calculation performance of the submitted calculations. The actual CPU time until 1,600 s and time step at 1,600 s are tabulated in Table 7. A great difference in the CPU time was found among the calculations. Participants B1 and B4 showed the fastest calculation time among the participants, while participants B3, C1, and C3 showed the latest calculation time. The time step during the calculation was also requested to investigate the calculation performance. The time step did not show a significant variation during the whole calculation time. Participant B3 had a minimum time step of 0.001 s. Participants A1 and A3 had a maximum time step of 0.01 second. The CPU time was the shortest for participants A2, B1, B2, B4 and C2 when the time step varied around 0.034 seconds.

The break flow rate is one of the most important boundary conditions for the test simulation. It is considered that choking occurs during the discharge of the singlephase liquid, two-phase mixture, and the single-phase steam throughout the test. In the ATLAS facility, the total break flow rate were was calculated using the experimental data of QV-CS-03, LC-CS-01, and LC-CS-02, and as a complementary method to the load cell-based break flow measuring method, a RCS inventory-based break flow estimation method has also been applied. From all the participants’ predictions, the break flow is high and fluctuating during the earlier stage at up to 174 s, decreases rapidly, and maintains a low and steadier flow thereafter during the DSP-02 test.

As a ratio of the downstream to the upstream pressures of the break spool piece decreased up to 0.2, it was presumed that a choking flow was maintained throughout the test. A subcooled liquid choking flow was observed in the earlier period of the test, and then rapidly converted into a saturated two-phase choking flow. After it maintained a short duration, the two-phase choking was changed into a single-phase steam choking flow after around 174 s, and maintained a single-phase choking condition during the remaining test period of the DSP-02 test.

Generally the break flow rate decreased rapidly after the break, reached a plateau with a short duration, and then decreased again in every simulation. Most of the simulation results showed the trends of subcooled liquid choking, saturated two-phase choking, and single-phase steam choking. However, there are some differences case by case. A1, A2, and A3 of Group A, B2 and B4 of Group B, and C1, C2, C3, and C4 of Group C predicted the experimental break flow rates well, but B1 and B4 of Group B under-predicted the experimental data during the whole period.

Table 8 shows a brief summary of the break flow rate and accumulated mass. The peak flow rate, mean flow rate at the plateau, transition time from the two-phase to single-phase flow, and the accumulated masses at the transition time and at 1,600 s were compared with the experimental data. The MARS and RELAP5 code has modified Henry-Faukse critical flow models. The reported critical flow models are listed in Table 8, including the discharge coefficient used to fit their simulation results to the experimental data. The reported discharge coefficients were between 0.80 and 0.93.

For a review of the downcomer fluid temperature, there are six subsections for each elevations from elevation 1 to elevation 6 azimuthally and in one section, for elevation 7. Please refer to the definitions of the elevations and azimuthal subsections in Figs. 9 and 10, respectively.

Most of the predictions showed that there were some multi-dimensional effects along the vertical elevation and

the azimuthal direction. In this section, the multi-dimensional effects were considered according to the azimuthal direction for each level in the downcomer and vertical direction for specific subsections. There was just one piece of data for elevation 7, i.e., lumped data, and thus was not considered for the azimuthal effect.

The test data and participant predictions show similar trends with acceptable deviations for lower elevations, e.g., from elevation 1 to elevation 4. For higher elevations, e.g., from elevation 5 to elevation 7, there are quite large deviations between the test data and participants predictions. From the RPV downcomer level data, e.g., Figure A.119 of reference 5, most of the data remains above the top of the active core, e.g. ~2.7m, during the transient except that some predictions go to just below the top of the active core during 300~500 seconds and then return to higher elevations. For lower elevations, e.g., from elevation 1 to elevation 4, the downcomer temperature does not have much effect from the incoming ECC water due to pre-existing

relatively hot water. But at higher elevations, e.g., from elevation 5 to elevation 7, the downcomer temperatures are mainly affected by the incoming ECC water because there were just steam regions for higher elevations. The top of the active core is included in elevation 4 of the downcomer, which is why most of the temperature data shows different trends below and above elevation 4.

As in a typical SBLOCA scenario, the PZR pressure of the test followed a general trend. As shown in Fig. 11, the general trend of the PZR pressure of ATLAS and all participants’ predictions can be split into 3 regions, e.g. initial blowdown, pressure plateau, and boil-off. In general, a boil-off will be triggered by the occurrence of the loop seal clearing(s), as shown in the figure. From the overall review of the test data and all predictions, certain common trends of the fluid temperature variations in the downcomer were found according to the 3 regions.

In review of the azimuthal effect, most of the predictions showed that there were two stages that represented the occurrence of a multi-dimensional effect, e.g., the initial stage and after SIT injection. For example, Fig. 12 showed these kinds of aspects in the SNU’s calculation (e.g. elevation 4).

(In the DSP-02 exercise, SNU performed a sensitivity calculation for the 2-dimensional behavior in the downcomer as remarked in Table 1.) As shown in the figure, most of the data for the initial stage looked to be affected by the actuation of the safety injection pump to a certain extent. After the loop seal clearing until SIT actuation, there are few azimuthal effects, which means a governing of the boil-off. But for the SIT injection stage, most of data predictions showed less relation to the actuation of the safety injection tanks. For a higher level, e.g., elevation 5, most of the data predictions were different from those of elevation 4 as shown in Fig. 13. For a comparison between the test data and predictions, test data for elevations 4 and 5 are shown in Fig. 14and Fig. 15, respectively. From both figures, the initial stage shows quite a larger azimuthal effect than the predictions, but for the SIT injection, less azimuthal effects than the predictions. The test data of elevation 6 shows quite a large effect on the azimuthal direction for after the initial stage due to SIP and SIT injections as shown Fig. 16.

In the review of the vertical direction, the first 60 seconds from the break corresponded to the ‘blowdown’ region. In

this region, most of the fluid temperatures in the downcomer decreased sharply just after the break, and then recovered to a certain value at the end of the blowdown region, as shown in Figs. 17 and 18. The ‘pressure plateau’ region remained from 60 seconds to 200 seconds.

As shown in Figs. 17 and 18, when the system pressure reached at the ‘plateau’ region, the fluid temperatures varied with respect to the vertical downcomer elevation. For example, in lower elevations 1 and 2, the fluid temperatures decreased slowly until the middle of the plateau and then increased to a certain value at the end of the plateau. But at the higher elevations 5 and 6, the fluid temperatures showed quite large oscillations, especially to lower values during the period of the plateau. Also, at middle elevations 3 and 4, the fluid temperatures showed intermediate trends, but close to those of the lower elevations. The oscillations of the fluid temperatures can be due to the actuation of the safety injection pump at 55 seconds. This was just before the entrance of the plateau. It can be concluded that the fluid temperatures in the higher downcomer elevations were affected by the cold ECC water from the safety injection pump. The loop seal clearings occurred at the

intermediate leg 1A and 2B at 196 seconds, e.g., just before the end of the plateau. The test data showed that the oscillations became smaller at the loop seal clearings. When the system reached at the end of the plateau, most of the fluid temperatures came to concentrate at certain values.

It is noteworthy that the fluid temperature of the elevation 7 maintained the highest values for the blowdown and plateau regions. From the collapsed water level data, it was quickly depleted during the blowdown period at elevation 7, and since then no water level was detected. Thus, the steam at elevation 7 would have some effect from the hot vessel wall, e.g., wall heat transfer effect. This seems to be the reason why its temperature was always higher than the others’. These trends can be seen in Figs. 17 and 18.

When the system pressure reached the entrance of the boil-off region, e.g., at 200 seconds, most of the fluid temperatures showed nearly the same values, and then decreased according to the system pressure changes until 446 seconds, as shown in Figs. 19 and 20. After that, the temperature of elevation 7 showed quite large oscillations, especially to higher values until the end of the test. As mentioned before, there was no water at elevation 7, and

thus its temperature would be influenced by the wall heat transfer and showed the highest temperature. For the other elevations, the trends of the fluid temperatures between 2 subsections became different. In subsection 2 (refer to Fig. 19), all fluid temperatures in elevations 1 through 3 showed similar trends and values like a single curve. In subsection 4 (refer to Fig. 20), most of the fluid temperatures at elevations 1 through 3 showed similar trends and values like a single curve, as in subsection 2. But the temperatures in elevations 4 through 6 showed some oscillations, especially to lower values. This is because of the effect of the ECC water from the safety injection tanks, which actuated at 446 seconds. But in subsection 2, there was no evidence of the effect of the ECC water from the safety injection tanks.

Unlike the LBLOCA, the loop seal clearing phenomena during the SBLOCA, especially for the cold-leg break situation, entirely governs the whole thermal-hydraulic behavior of the primary system. For the APR1400, four intermediate legs exist between the two SGs and four RCPs. In the SB-CL-09, a 6 inch cold-leg break was simulated. Generally, for the cold-leg SBLOCA case, the pressure of the upper-head region would be increased mainly due to the accumulated steam and water inventory in the upperplenum. This built-up pressure acts as a suppression force to the core water level and resultantly the core water level will decrease possibly up to and/or below the top of the active core region. Simultaneously, the downcomer water level would increase due to the evacuated water inventory from the lower part of the core region.

This unbalanced hydro-static pressure between the core and the downcomer region acts as a potential pushing force to the SG side intermediate leg. The potential pushing force will be increased with time to overcome the hydrostatic head in the intermediate leg. The unbalanced hydrostatic pressure finally can be dissolved with the occurrence of the loop seal clearing. Therefore, the loop seal clearing phenomena is very important with respect to the PCT, one of the most important parameters to insure the safety of the reactor system.

The loop seal clearing behavior is closely related to the break location and the break size. Usually, a loop seal of the break loop is cleared first, and the number of cleared loop seals is dependent on the break size. The larger the break size is, the more the loop seal is cleared, as can be observed in Figs. 21 and 22, which show a collapsed water level behavior during the SB-CL-01 (4 inch cold leg SBLOCA test) and SB-CL-03 (8 inch cold leg SBLOCA test), respectively. This different loop seal clearing behavior with respect to the break size is mainly dependent on the unbalanced hydro-static pressure between the upper head and the downcomer region, as shown in Figs. 23 and 24 for SB-CL-01 and SB-CL-03, respectively. On the other hand, in the SB-CL-02, intermediate leg-1A and -2B were cleared.

While the prediction of the primary system pressure was quite good, the secondary system pressures were not properly estimated in many calculations. In particular, the secondary system pressures subsequent to the MSSV opening were over-predicted. Only the calculations performed by participants B2 and A3 showed reasonable prediction performance of the secondary system pressures during the late phase calculation times after the MSSV opening. The reasons for the over-prediction of the secondary system pressure are complex. The secondary system pressure can be affected by the loop seal clearing behavior, the heat loss, and the fluid condition inside the U-tubes. In the calculations where the loop seal clearing behavior was not correctly predicted, the secondary system pressures show a tendency to be over-estimated. After the loop seal clearing, the U-tubes of the steam generators were emptied. However, if the loop seal was formed and not cleared in the intermediate legs, the reverse heat transfer in the Utubes resulted in a heat up of the fluid inside the U-tubes, consequently causing an increase in the secondary system pressure.

In the SB-CL-09 test, the primary heat loss at the initial condition was about 66 kW and the secondary heat loss of each steam generator was about 28.5 kW. During the test, the primary heat loss was compensated by adding it to the core decay power table with respect to time. The secondary heat loss, however, was not compensated in the test. Table 9 summarizes the detailed heat loss modeling and the prediction performance of the secondary system pressure in each calculation. As for the primary heat loss, most calculations considered the primary heat loss by subtracting the constant heat loss from the initial core power. The modeling of the secondary heat loss, however, was differently applied to the participant’s calculations as shown in Table 9. At this stage, a definite conclusion on the relation between the heat loss modeling and the secondary system pressure prediction cannot be made owing to a lack of detailed information on the code input modeling and the use of the different codes in each calculation. However, it can be rationally expected that modeling of the heat loss at the steam generators affects the secondary system pressure behavior. Participant C4 performed a sensitivity calculation for the effect of secondary heat loss on the secondary system pressure behavior. In the calculation where the secondary heat loss was considered, the secondary system

pressure presented a similar trend with the SB-CL-09 test data. Despite the difference from a quantitative point of view, it can be found in the participant C4’s calculation result that the heat loss at the steam generators needs to be properly modeled for the accurate estimation of the secondary system pressures.

Another possible source of over-prediction of the secondary pressure is the quality of the steam entering the U-tubes after the loop seal clearing. If the steam quality at the U-tube inlet is lower in the calculation than in the test, it would result in less reverse heat transfer from the secondary to the primary system and consequently higher secondary pressure than the data is obtained in the calculation and vice versa. Unfortunately, however, the present SBCL- 09 test does not provide such detailed local information on the steam quality to support this reasoning.

In the DSP-exercise, most of the participants adopted the MARS-KS code for open calculation, which is why special code assessments could be performed by the participants. In this section, the major findings from those activities are summarized as follows:

(1) Break flow modeling;

It was found that detailed upstream modeling from the break nozzle to the break valve resulted in better agreement with the data. In such a detailed modeling, the discharge coefficient of Cd=1.0 is recommended as one of the user guidelines.

(2) Loop seal clearing;

Most participants were not successful in reproducing asymmetric loop seal clearing behavior. The loop seal clearing was greatly affected by a small model change. From the viewpoint of safety, where the loop seal clearing occurred and how many loops were cleared seem to be unimportant. However, the occurrence timing is very important because it governs the event progress since then.

(3) ECC bypass rate;

This was estimated by injecting boron during the code calculation by one participant, who tried a creative calculation method. Around a 30%-45% ECC bypass rate was obtained. However, this bypass rate needs to be confirmed by experimental data. Unfortunately, experimental evidence to confirm the estimated ECC bypass rate is not available. This finding can be feedback to experimentalists to improve the measurement methodology.

(4) Heat loss effects;

A detailed sensitivity study was done to answer the question why the secondary pressure was over-predicted by almost all calculations. It was found that such overpredictions of the secondary pressure were due to a lack of heat loss modeling in the secondary system. The effects of heat loss were included in the model, and a very nice agreement with the data was obtained.

(5) Momentum effects of ECC water;

This was not a dominant factor affecting the transient calculations.

(6) 2D behavior;

Practical 2-D behavior was observed in the ATLAS experiment. In particular, fluid mixing was not properly predicted by most calculations. A cross junction k-factor was not helpful to resolve insufficient mixing. The use of a turbulent mixing model of the MARS-3D code is recommended for better prediction.

(7) ACC component;

The accumulator component needs to be improved to remove the initial peak and minimize the flow oscillation. When a simple PIPE component is used to model the SITs instead of the ACC component, the initial peak and flow oscillation can be avoided.

(8) Condensation;

Injection of cold water into the downcomer results in excessive condensation, causing an increase in the downcomer water level and a decrease in the core water level. It was found that utilizing the ECC mixer model mitigated the condensation. The condensation model needs improvement.

(9) RPV upper head temperature;

In most calculations, it was close to the hot leg temperature owing to a reverse downcomer-upper head bypass flow path. This caused early flashing and depression in the downcomer water level.

(10) High core water level before the loop seal clearing;

This was due to a high reverse flow from the steam generator to the RPV upper head. Appling the CCFL option to the RPV fuel assembly plate can mitigate this disagreement.

Compared to the DSP-01 exercise, the host organization adopted the BEMUSE phase 2 project methodology for accuracy quantification. The methodology of the accuracy quantification consists of (1) a global acceptability factor for the nodalization development, Q_A; (2) global acceptability factor for the nodalization qualification at the steady state level, Q_B; and (3) global accuracy quantification for the deviations between calculations and measurements quantified by FFTBM, AA_tot. In the quantification of the DSP-01, only this one was discussed in reference 2.

As the quality of transient code calculations is greatly dependent on how well the code model is initialized at a steady state condition, a steady state qualification based on measured data was performed, following the similar methodology as that proposed by the BEMUSE phase-2 program [11] in the present DSP-02. The steady state qualification includes two different steps: one is related to the evaluation of the geometrical data and of numerical values implemented in the nodalizations; the other is related to the quality of the steady state calculation results. Nine parameters have been selected for nodalization qualification as shown in Table 10.

As the first step, the acceptable errors (AE) for the quantification process were determined. Different AEs from 1% to 3% were used depending on the parameters shown in Table 10. The percentile error, E, was defined as the ratio

The percent error, E becomes zero if the calculated value is between the experimental lower and upper values, taking into account the experimental error.

Secondly, weighing factors taking into consideration the importance of the parameters with respect to the present SBLOCA transient were determined. Taking into account the phenomena of the SBLOCA, the relative importance of the secondary inventory was assumed to be half that of the primary inventory. Thus, weighting factors, W_i, of 1.0 and 0.5 were used for the primary and secondary inventory, respectively. The single acceptability factor, Q_Ai was then obtained by the following formula:

where, normalized weighting factors were used. Finally, the global acceptability factor, Q_A, can be obtained by summing the whole single acceptability factors

The effects of selected parameters on Q_Ai are shown in Fig. 25. All calculated Q_Ai are plotted together. The secondary circuit volume (parameter 2) and maximum axial power distribution for the average rod (parameter 9) showed the greatest values in Q_Ai among the others. Final nodalization quantification results are shown in Table 11.

In the literature [11], Q_A<1.0 is required as an acceptable criterion. Around 50% of the calculations fulfilled the global acceptability in the present exercise. In fact, Q_A is affected by two factors, AE and W_i, and a proper choice of those factors are important for a meaningful quantification process. In the present quantification, two factors were determined by considering the relative importance of each inventory during the typical SBLOCA scenario, though they look more or less subjective. From Table 11, participant B4 showed the best result of 0.585, andnt C4 showed the greatest value of 2.14 among thents.

Steady state results can be quantified by using the similar methodology used in the quantification of QA in the previous section. For the quality of the steady state calculation results, 21 parameters have been selected for a steady state qualification, as shown in Table 12.

At first, the acceptable errors (AE) for the quantification process were determined. Taking into account the measurement uncertainties, different AEs from 0.25% to 30% were used depending on the parameters shown in Table 12. The percentile error, E, was defined as Eq. (3).

Secondly, weighing factors taking into consideration the importance of the parameters with respect to the present SBLOCA transient were determined, as shown in Table 12. Also, the single acceptability factor, Q_Bi, was obtained by Eq. (4). Finally, the global acceptability factor, Q_B, can be obtained by summing the whole single acceptability factors as in Eq. (5). Calculated results for the submitted calculations are summarized in Table 13.

In the literature [11], Q_A<1.0 is required as an acceptable criterion. Only 2 calculations fulfilled the global acceptability in the present exercise. From Table 13, participant A2 showed the best result of 0.159, andnt C2 showed the greatest value of 3.058 among thents.

Application of FFTBM to the present DSP-02 calculation was performed to evaluate the accuracy of the submitted calculation results. As in the DSP-01 exercise, 22 parameters have been used to characterize all the relevant phenomena that were measured during the test, as shown in Table 7 of reference 2. Also, the weighting factors were used to consider the different importance from the viewpoint of a safety analysis and to calculate the overall accuracy of the calculation, i.e., the total average amplitude (AA_tot). In the present analysis, the weighting factors used in Table 7 of reference 2 were adopted.

For the quantification of each participant’s calculation, three cases with different time frames were evaluated as shown in Table 14. In the first time frame between 0 s to

24 s, 19 parameters out of the selected 22 parameters were used because the SIP and SIT were not available during this period. The parameters relevant to the SIP and SIT were excluded in the FFTBM calculation. In the second time frame between 0 s to 440 s, the SIT was not activated in the test, and thus 2 parameters relevant to the SIT were excluded in the FFTBM calculation. In the whole time frame calculation, the 22 selected parameters were used to get the final AA_tot. For all cases, the cut-off frequency was set to 1.0 Hz.

A summary of the results of FFTBM application to the DSP-02 can be seen in Table 15. Overall, most calculations showed good prediction results except for participant B1’s

calculation. (In this report, the authors suggest a criteria for the accuracy of a given calculation, e.g., a ‘good prediction’ if AA_tot < 0.5; and a ‘poor prediction’ if AA_tott > 0.5. The original categorization of the total AA_tot can be seen in reference 12.) This FFTBM was based on several assumptions; selection of variables, selection of the time frame of interest, weighting factors, and cut-off frequency. These assumptions were more subjective than objective so that the ranking among the calculations may be changed if different assumptions were used. Therefore, the ranking in Table 15 does not mean definite superiority of one calculation over the others. However, when the present assumptions were used, participant A2’s calculation showed the best prediction results among all participants. By comparing with the quantifications in the previous sections, the current FFTBM application results showed quite good consistency with them.

The results of the accuracy evaluation for the time interval of 0-400 s are summarized in Table 16. Here, AA_norm in the table means the normalization to the results of the AA_tot calculation using the criteria of good prediction, e.g., 0.5, as mentioned in the previous section. The time interval for AA_tot or AA_norm was selected as 0-440 s as defined in Table 14, because most of the major phenomena, e.g., reactor trip, SIP injection, and loop seal clearing, occurred within the interval. The results of the accuracy evaluation are shown in Fig. 26.

As mentioned in the previous sections, the global acceptability factors, e.g. Q_A and Q_B, are identified as acceptable if they are less than 1.0. The normalized global accuracy, e.g., AA_norm, means a good prediction if it is less than 1.0, too. Thus, a combination of 3 global quantities can be meaningful for a quantitative comparison with the introduction of reasonable weighting relations between them. It is noteworthy here that an overall quantity called Q_all is introduced by the authors for a quantitative comparison between participants, as

The weighting factors f_A, f_B, and f_AA assumed in Table 16 were 0.2, 0.2, and 0.6, respectively. The results of the normalization are shown in Fig. 27.

The second domestic standard problem (DSP-02) was successfully completed using a 6-inch cold leg break loss of coolant accident data, where a total of 11 organizations including the research institute, industry, safety authority, and university contributed to in-depth safety analyses. Almost all participants used the MARS-KS code. Besides performing a comparison with the experimental data, each participant performed a sensitivity analysis on at least one special code assessment topic as agreed upon the organizing committee. This activity is expected to help find out the code deficiencies and obtain user guidelines that can minimize the “user effects.”

In the DSP-02 exercise, the overall figure of merit for a code accuracy quantification was fully applied. Three code accuracy quantification indices including (1) nodalization accuracy, (2) steady state accuracy, and (3) transient accuracy were defined and integrated to produce a single quantification index. The obtained outcomes will be used to improve the MARS-KS code model and help code users perform a safety analysis.

A1~A3 Participants categorized into Group A

AFWP Auxiliary feedwater pump

APR1400 Advanced power reactor 1400 MWe

ATLAS Advanced Thermal-hydraulic test Loop for Accident Simulation

B1~B4 Participants categorized into Group B

BEMUSE Best Estimate Methods Uncertainty and Sensitivity Evaluation

BS Break simulator

C1~C4 Participants categorized into Group C

CL Cold leg

CO Core

CPU Central process unit

CS Containment simulator

DC Downcomer

DP Pressure difference

DCUH Downcomer upper head

DSP Domestic standard problem

DVI Direct vessel injection

ECC Emergency core cooling

EN2T Environment & Energy Technology Co., Ltd.

FCV Flow control valve

FDR Facility Description Report

FFTBM Fast Fourier transform based method

FLB Feed line break

FNC Future and Challenge Technology Co., Ltd.

HL Hot leg

HPS High pressure safety injection

ID Identification or inner diameter

IET Integral Effect Test

IL Intermediate leg or cross-over leg (COL)

KAERI Korea Atomic Energy Research Institute

KAIST Korea Advanced Institute of Science and Technology

KEPCO Korea Electric Power Corporation

KEPR Korea Electric Power Research Institute

KINS Korea Institute of Nuclear Safety

KEPCO NF KEPCO Nuclear Fuel, Ltd.

KEPCO E&C KEPCO Engineering and Construction Company, Inc.

LBLOCA Large break loss of coolant accident

LC Load cell

LOCA Loss of coolant accident

LP Lower plenum

LPP Low pressurizer pressure trip

LSC Loop seal clearing

LT Level transmitter

LUDP Low upper downcomer pressure trip

MARS-KS Multi-dimensional analysis of reactor safety (version KS)

MF Main feedwater

MS Main steam

MSIV Main steam isolation valve

MSLB Main steam line break

MSSV Main steam safety valve

NETEC Nuclear Engineering & Technology Institute

OA Operating agency, KAERI

OPR1000 Optimized power reactor 1000 MWe

OV On-off valve

PCT Peak cladding temperature

PT Pressure transmitter

PWR Pressurized water reactor

PZR Pressurizer

QV Volume flowmeter

RCP Reactor coolant pump

RCS Reactor coolant system

RELAP5 Reactor excursion and leak analysis program (version 5)

RETECH Responsible Engineering Technology Co., Ltd.

RPV Reactor pressure vessel

RV Reactor vessel

RWT Refueling water storage tank

SBLOCA Small break loss of coolant accident

SD Steam dome or system design

SDD System Design & Development Co., Ltd.

SENTECH System Engineering Technology Co., Ltd.

SG Steam generator

SGSD Steam generator steam dome

SGSDDC Steam generator (between) steam dome (and) downcomer

SGTR Steam generator tube rupture

SIP Safety injection pump

SIT Safety injection tank

SNU Seoul National University

TC Thermocouple

TF Fluid temperature

Symbols

AA Average amplitude (FFTBM) for a single parameter

AA_norm Normalized AA_tot by 0.5

AA_tot Global average amplitude (FFTBM) for the global calculation

AE Architecture engineer or Acceptable error

Cd Discharge coefficient

d Diameter

E Percentage error

f Frequency or weighting factor

l Axial length

Q Heat loss

Q_A Global acceptability factor for nodalization

Q_B Global acceptable factor for steady state level

Q_i Single acceptable factor

T Temperature

DT Temperature difference

WF Weighted frequency

W_i Weighting factor

Y_E Experimental value of generic quantity

Subscripts

0R model-to-prototype quantity ratio

1 primary system

2 secondary system

A Q_A-related

AA AA_norm-related

atm atmospheric condition

B Q_B-related

max maximum

norm normalized

s sample

tot total

w wall