Resolution of Generic Safety Issues: Issue 92: Fuel Crumbling During LOCA (Rev. 1) ( NUREG-0933, Main Report with Supplements 1–34 )
Experiments conducted at several test facilities prior to 1984 showed that irradiated fuel can fragment (crumble) into small pieces during a LOCA. Some evaluation of this effect was made for NRC by EG&G.622 Although it was concluded that temperature increases due to relocation of crumbled fuel would be smaller than those due to other processes that are treated conservatively in Appendix K, the question was raised whether this effect should be treated as a long-term generic safety issue.622
During the course of a LOCA, the primary system pressure drops and the fuel rods heat up. As the rods heat up, the zircaloy cladding experiences a series of phase changes. Thus, the strength of the cladding varies, and at a certain interval during the heatup, the internal pressure within a fuel rod will cause the cladding to plastically deform and swell, a phenomenon normally referred to as "ballooning." By this time, the ceramic fuel pellets may be cracked into small pieces. As the cladding swells, the crumbled fuel core drops into and (partially) fills the ballooned region while still producing thermal energy from radioactive decay. Thus, as the fuel settles into a shorter, fatter stack, the local linear power density (kw/ft) increases, even though the total rod power remains constant.
At the time of the initial evaluation of this issue in July 1984, the existing ECCS performance analysis codes did not account for fuel settling into ballooned regions. Thus, the lack of inclusion of this effect was a non-conservatism; however, the EG&G study622 concluded that known conservatisms would more than offset this effect.
The only known solution to this possible problem was to account for the fuel settling and increased kw/ft in the ECCS calculations. This would result in stricter limits on FQ or MAPLHGR, which would make maneuvering more difficult and could result in a plant derate.
This item was an issue only for a mitigated LOCA. Therefore, the frequencies of A and S1 LOCAs in WASH-140016 were added. S2 LOCAs were not included because, if they were mitigated at all, they would be so far from the regulatory pellet cladding temperature (PCT) limit of 2200F that additional heat from fuel crumbling could be accommodated. The sum of the A and S1 frequencies was 4 x 10-4 /RY.
Not all LOCAs are design basis LOCAs. For a design basis LOCA, it is necessary to have the worst break location, the worst break size, the core power distribution at the FQ (or MAPLHGR) limit, and the worst single failure of the ECCS, plus other conservatisms in initial conditions, system capacities and calculational modeling. The conservatisms in traditional LOCA analyses may well be sufficient to compensate for the effect of fuel crumbling, as was discussed in the EG&G study.622 However, the opposite approach was taken and no credit for calculation conservatism was assumed. Instead, the initial conditions in a probabilistic manner were included.
The probability of a LOCA being a design basis LOCA is very small. However, the worst case is often the worst by only a small margin, i.e., the worst break size and location may be closely followed by other break sizes in other locations. Thus, the probability of a near design basis LOCA is significant. Based primarily on judgment, it was assumed that there was at most roughly a 10% probability of a LOCA approaching design basis conditions.
The EG&G study622 discussed a PBF calculation that was somewhat conservative but indicated that peak cladding temperatures rose 46F and peak centerline temperatures rose 790F above the nominal (no fuel relocation) case when circumferential strain was 44%. A companion calculation, assuming 89% cladding strain, yielded a cladding T of 406 and a centerline T of 2290F. However, 89% core-wide cladding strain is not realistic. According to NUREG-0630,634 a uniform cladding strain of 70% corresponds to all pins swollen into a square shape and pressed against each other with no space remaining to accomodate further swelling. The numbers associated with 44% cladding strain, which corresponded to about 78% channel blockage, were used for core-wide calculations, i.e., it was assumed that every fuel pin in the core would swell 44% throughout its length.
In addition, at some point along each fuel rod, the cladding swells to the point of bursting, which relieves the internal pressure and terminates the ballooning process. Thus, it was further assumed that a one-foot section of each fuel rod (i.e., about 10% of the length of every fuel rod in the core) would swell to 89% cladding strain.
The rise in temperature was the main concern. Because the degree of ballooning is a function of how the cladding passes through its various phases, rather than only a function of its peak temperature, the increased temperature was not expected to increase the amount of ballooning. Consequently, fuel crumbling and settling were not expected to result in more flow blockage. Some temperatures of interest were:
3455ºF - Eutectic forms (NUREG/CR-1250,161 Volume 2, Part 2, p. 513)
5148ºF - UO2 melts
During a design basis LOCA, the hottest cladding temperature will be at or slightly below 2200F before the core is quenched by emergency core cooling water. Using the T calculated for 44% strain, the effect of fuel settling into ballooned regions will raise peak cladding temperatures by 46F to about 2250F, and peak centerline temperatures 900F above this to about 3150F. These temperatures were well below those needed to cause loss of coolable geometry, or release copious quantities of non-volatile fission products or actinides from the fuel matrix via melting or eutectic formation. The only effect would be to drive off more noble gases and iodine.
In the WASH-140016 calculations for Release Categories PWR-9 and PWR-8 (mitigated LOCA with and without containment isolation), it was assumed that 3% of the noble gases and 1.7% of the iodine are released to the containment. These figures were scaled up such that all the noble gases are released, i.e., it was assumed that the radiological consequences of a fuel pin segment exceeding 2200F at the cladding are 331/3 times those of these two release categories. Thus, 100% of the noble gases and 57% of the iodine would be released from fuel which exceeds 2200F at the cladding. This was a bounding calculation.
The next question was, given a uniform 46F increase in cladding temperature throughout the core, what fraction of the core will exceed the 2200F licensing limit during the LOCA? Previously, only the hottest point touched 2200F. Generally, a 10F change in PCT corresponds roughly to at most a 0.01 change in FQ for temperatures up to 2200F. This rule of thumb was combined with three-dimensional power distribution information.633 The result was that a uniform 46F rise will result in roughly 0.015 of the core exceeding the 2200F limit. The consequences due to core-wide uniform ballooning and fuel setting were then bounded as follows:
|R||< (containment isolated) (0.015)(33 1/3)(120 man-rem/PWR-9)|
|< 60 man-rem|
|R||< (containment not isolated) (0.015)(33 1/3)(75,000 man-rem/PWR-8)|
|< 37,500 man-rem|
As discussed earlier, it was assumed that a portion of each fuel rod totaling 10% of its length balloons well in excess of 44% strain. If this strain were 89% as in the second PBF calculation, and if the section in question previously just touched the 2200F PCT limit, the revised peak cladding and centerline temperatures would be roughly 2600F and 4640F, respectively. The centerline would not be hot enough to melt and the cladding would not be hot enough to form a eutectic, but, given the roughness of these estimates, the possibility of a copious release of fission products cannot be ruled out. Accordingly, no attempt was made to base a calculation on the degree of ballooning, but instead the radiological consequences were bounded by assuming that these 10% sections of each fuel rod release fission products as if they were molten. (It was assumed that there was no probability of containment overpressure or other mechanistic consequence normally associated with a core-melt.) Because 0.1 of the mass of the core was affected, one-tenth of the radiological consequences of a PWR-7 release (core-melt with no containment failure) and a PWR-5 release (core-melt with failure of the containment to isolate) were used. The increase in consequences due to enhanced release at the rupture points on each fuel rod were then calculated as follows:
|R||< (containment isolated) (0.1)(2300 man-rem/PWR-7)|
|< 230 man-rem|
|R||< (containment not isolated) (0.1)(1,000,000 man-rem/PWR-5)|
|< 100,000 man-rem|
Combining these values, the public risk associated with this issue was estimated to be as follows:
|FR||< (4 x 10-4 LOCAs/RY)(0.10 near design basis LOCAs/LOCA) x [(0.9 containment isolation/DBLOCA)(60 + 230) man-rem + (0.1 isolation failure/DBLOCA)(37,500 + 100,000) man-rem]|
|< 0.56 man-rem/RY|
For a 30-year plant life, the public risk was estimated to be about 20 man-rem/reactor.
Industry Cost: A possible fix was to reduce FQ limits by about 0.05 which would lower the calculated PCT by about 50F. This would make it harder to maneuver the plant; startups and load changes would take longer. If a plant is or becomes LOCA-limited, this would also cause a derate. In addition, changes in the ECCS analysis generally involve considerable administrative expense. As a minimum cost, it was assumed that 5 startups (scram recoveries) a year would be extended by one hour, and that one staff-year (including NRC staff time) and $50,000 of computer expense would be expended in TS changes. Thus, the total cost over a 30-year period would be at least $1M/reactor.
NRC Cost: The NRC cost was negligible compared to the industry cost.
Based on an estimated public risk reduction of 20 man-rem and a cost of $1M for a possible solution, the value/impact score was given by:
The above calculations indicated that this issue should be placed no higher than the low priority category. This meant that there was insufficient risk-based justification for starting a major re-review of existing ECCS Appendix K performance analyses. However, it was noted that there were ongoing efforts to develop and license ECCS performance models that were more realistic (and consequently less conservative) than the models in use at the time of this evaluation in July 1984.
It was not valid to conclude that the effects of fuel crumbling and settling into ballooned regions could necessarily be neglected in any new, more realistic models. Instead, it was expected that these effects (which are real physical phenomena) would be appropriately addressed in the calculations. Moreover, a separate generic issue on fuel crumbling was not necessary; such work was best done within the scope of the review of the new calculational methodology. Thus, the issue was given a low priority (see Appendix C). In NUREG/CR-5382,1563 it was concluded that consideration of a 20-year license renewal period did not change this priority. Further prioritization, using the conversion factor of $2,000/man-rem approved1689 by the Commission in September 1995, resulted in an impact/value ratio (R) of $50,000/man-rem which placed the issue in the DROP category.