Resolution of Generic Safety Issues: Issue 79: Unanalyzed Reactor Vessel Thermal Stress During Natural Convection Cooldown (Rev. 3) ( NUREG-0933, Main Report with Supplements 1–34 )
On March 18, 1983, B&W expressed530 concern for unanalyzed reactor vessel thermal stress that could occur during natural convection cooldown of PWRs. The concern emerged from a preliminary B&W evaluation of a voiding event that occurred in the upper head of the St. Lucie reactor on June 11, 1980. Based upon several conservative assumptions, B&W tentatively concluded that, during natural convection cooling, axial temperature gradients between 150ºF and 200ºF could develop in the vessel flange area which could produce thermal stresses in the flange area or in the studs that might exceed Code allowables, when added to the stresses already considered (bolt-up loads, pressure loads, etc.). B&W acknowledged the preliminary nature of their analysis and noted that, if their conservatively calculated cooling rate of the stagnant coolant in the vessel head (2ºF/hour) were in the range of 20ºF/hour, then the estimated vessel stresses would not be excessive.
B&W requested assistance from the NRC in obtaining any data or information which might help in the technical evaluation of this problem, but noted that this phenomenon was not likely to be a serious near-term safety concern. However, it was their view that it did represent an unanalyzed situation with the potential for margin reduction over the life of a plant. Preliminary evaluations531,533 by the NRC staff also led to the conclusion that the problem was not a serious near-term safety concern, although a Board Notification534 was issued.
In an RES memorandum668 to NRR, it was pointed out that the closure flange regions of reactor vessels were of concern in drafting revisions to Appendices G and H of 10 CFR 50; these regions control the pressure-temperature limits at low pressures for vessels that have very little radiation damage in the beltline region. Critical flaw sizes are of the order of 0.25 inch deep. The existence of a possible new source of stress in the flange region could have a bearing on the revisions to 10 CFR 50. An additional analysis of the concern was made by the B&W Owners' Group.669
When unanalyzed reactor vessel thermal stress is added to existing stresses, the stresses in the flange area or studs could exceed the allowable stress. Moreover, the cycling of these temperature gradients could cause a reduction in the fatigue margin or usage factor of the vessel over the life of a plant. In addition, depending upon the vessel temperature distribution, there is a possibility of vessel fracture under these circumstances. These factors could cause vessel cracking that could lead to unacceptable vessel failure during the life of the vessel. However, it was assumed in this analysis that sufficient water is available to prevent dry-out of the steam generators; otherwise, the consequences could be more serious than estimated.
If unanalyzed thermal stresses cause a reduction of fatigue life or lead to reactor vessel stresses that exceed the allowable stresses for the vessel, the solution was assumed to be a slower cooldown rate than the existing allowable rate of 100ºF/hour.
In an effort to establish the frequency of the occurrence of unanalyzed stress, it was noted in the B&W letter530 that "[t]hese non-uniform effects in the reactor coolant may occur once the reactor coolant pumps are secured and the decay heat removal system has been actuated in the normal cooldown mode or during natural circulation." This B&W statement was somewhat of an oversimplification, however, and may have been misleading since the effects resulting from normal cooldown and natural circulation cooldown could be substantially different with regard to the non-uniform effects in question. (In the normal cooldown mode, the RCPs continue to operate until the reactor and temperature are reduced to the cut-on levels for the decay heat removal system pumps which are well below the values at which natural circulation would be initiated, if required.) Therefore, with the RCPs running during the early phase of a normal cooldown, it would be expected that significant mixing of the fluid in the reactor head would occur in this period of the cooldown so that the non-uniform effects are likely to be minimized, if not precluded. On the other hand, the natural convection mode of cooldown would tend to have larger non-uniform effects because it starts with the highest fluid and material temperatures and the thermal mixing is assumed to be substantially lower than with the RCPs in operation. Therefore, the frequency corresponding to this cooldown mode was used in this analysis.
A dominant factor that could affect primary coolant pump availability leading to the need for natural convection cooling is the loss of offsite power, because the large power demand of these pumps usually exceeds the on-site power supply capability. It was noted, however, that Regulatory Guide 1.93532 provides guidance that permits continued operation of a plant for up to 24 hours after loss of offsite power followed by 6 hours of hot standby, if this implements the safest operating mode whenever the offsite power sources are less than the limiting conditions of operation. For this time period (30 hours), it was estimated from the information provided in NUREG/CR-3226526 that the probability of not recovering offsite power was approximately 0.03. If offsite power did not return in this time, a plant is brought to cold shutdown and natural convection cooling would follow. However, the licensee need not keep the plant up for 30 hours and can elect to immediately bring the plant to the cold shutdown state if it is deemed the safest mode of operation following the loss of offsite power. In this case, natural convection cooling of the reactor would begin promptly and it was estimated from NSSS vendor studies535, 536 that the maximum thermal gradients in the reactor vessel would occur in approximately 2 to 3 hours. From the information in NUREG/CR-3226,526 it was estimated that the probability of not restoring offsite electric power within two hours was approximately 0.35. Inasmuch as this latter case represented the more stringent conditions for this issue, this analysis was based on immediate shutdown to the cold shutdown condition as a conservative scenario.
The frequency of the loss of offsite power was established as 0.118/RY526 and the probability of not restoring offsite power within 2 to 8 hours was about 0.35. The frequency of attaining the maximum thermal gradients within the reactor, as a result of natural convection cooling, was estimated to be (0.118)(0.35)/year or 0.04/year.
Following the loss of the primary pumps, the reactor will be scrammed, natural convection cooling will commence, and the core and reactor vessel will begin to cool down. It was assumed that on-site AC power is available, otherwise the issue would become station blackout (which was addressed in Issue A-44).526 Based on the cooling rate of approximately 100ºF/hour, it was expected that the large temperature gradients described by B&W530 would begin to develop after approximately 2 to 3 hours, because the vessel head cools down very slowly (~2ºF/hour). Since the thermal stress is likely to be greatest in the region from just below the vessel flange530 to the "knuckle" just above the vessel head flange,535 a crack was postulated to occur in the upper part of the vessel or vessel head, if the allowable stress were exceeded following the addition of the thermal stress. As a result, the postulated crack was located above the core region.
Upon development of the crack in the upper part of the reactor vessel, the reactor system will blow down as in a LOCA and the core is likely to become partially or totally uncovered. At this time, 2 to 3 hours after reactor shutdown when the large thermal gradients develop, the decay heat is in the order of 1.5% of full power. The time required for the core to slump under these conditions was estimated to be slightly over 1 hour for a 3500 MWt reactor with a core of approximately 1100 ft3 and a total core heat capacity of 54 BTU/ft3/ºF. Within this time period, it is expected that the operator can properly assess the situation and will take appropriate action to prevent core slumping by releasing the water in the accumulators and submerging the core. The integrity of the reactor vessel is expected to be maintained to a level above the height of the core because the high stress regions were limited to the vicinity of the vessel flanges. Boil-off will continue, but it was assumed that water replacement from the reactor plant systems will be available to continue to cool the core. Some gap activity release is possible during the period of postulated core uncovery.
The above scenario is comparable to PWR Release Category 8 or 916; in these categories, the core does not melt and only some of the activity in the gaps of the fuel rods is released. However, in PWR Release Category 8, it was postulated that the containment fails to isolate properly and the public dose was estimated to be 75,000 man-rem. In Release Category 9, the containment isolates properly and the public dose is only 120 man-rem, which can be neglected. Inasmuch as there is no assurance that the containment will isolate properly for this issue, despite the fact that the accident is estimated to occur 2 to 3 hours after plant shutdown, it was assumed that Release Category 8 represented the scenario of this problem. A public dose of 75,000 man-rem was therefore postulated.
From NUREG/CR-2800,64 the probability that the containment fails to isolate is 0.0073. Therefore, the potential risk reduction was estimated to be (0.04)(0.0073)(75,000) man-rem/RY or 22.6 man-rem/RY. For an average plant lifetime of 28 years, the total risk reduction was estimated to be (22.6)(28) man-rem/reactor or 633 man-rem/reactor.
The possible solution to this issue required that natural convection cooling proceed at a slower rate than the allowable rate of about 100ºF/hour. It was assumed that the thermal stress condition would be ameliorated if the cooldown time were increased from approximately 80 to 160 hours. Moreover, it was assumed that sufficient water is available to prevent dry-out of the steam generators for these time periods. The increase in cooling time represented down-time for the reactor of an additional 80 hours or 3 1/3 days. At a cost of $300,000/day, the additional licensee cost for each natural convection cooldown was approximately $1M. It was estimated that 2 to 3 natural convection cooldowns would be required over the remaining life of the affected plants, based on the expected frequency of 0.118/RY for the loss of offsite power. The estimated licensee cost associated with three natural convection cooldown events was ($1M)(3)/reactor or $3M/reactor.
Additional NRC and licensee costs associated with technical studies to be performed for the determination of required cooldown rates, possible TS changes, etc., were estimated to be approximately $100,000/plant. Thus, the total cost associated with the possible solution to this issue was estimated to be $3.1M/reactor.
Based on the potential risk reduction of 633 man-rem/reactor and a cost of $3.1M/reactor for the possible solution, the value/impact score was given by:
Based on the value/impact score and the potential risk reduction, this issue was given a medium priority ranking (See Appendix C). However, these results were based on the assumption that sufficient cooling water is available to prevent dry-out of the steam generators without offsite power. Without this assumption, the consequences would be considerably more serious than estimated above.
In resolving the issue, a detailed stress evaluation of the closure region of the 177 Fuel Assembly reactor vessel was performed by B&W for the natural circulation cooldown condition and submitted to the NRC; this fuel assembly design was utilized on all operating reactors. An independent confirmatory stress analysis of the closure was performed by BNL. Based on the results of the BNL review and analyses and an RES staff evaluation of the adequacy of the RV closure region fracture toughness for the natural circulation cooldown condition, RES concluded that the B&W 177 closure region met all existing applicable regulatory design criteria. RES additionally concluded that, except for the issuance of Generic Letter 92-021431 for information, no immediate generic or plant-specific actions were necessary. Thus, this issue was RESOLVED and no new requirements were established.1432 In an RES evaluation,1564 it was concluded that consideration of a 20-year license renewal period did not affect the resolution.