Despite over 50 years of nuclear plant operation, not all phenomenological processes to which a nucl

1
Despite over 50 years of nuclear plant operation,
not all phenomenological processes to which a
nuclear plant may be exposed are rigorously
understood. Risk-related questions linger with
regard to (1) the likelihood of boric acid
precipitation following a Large Break
loss-of-coolant accident (LOCA), (2) the
likelihood and impact of debris accumulation at
the ECCS sump as at least a function of break
size, (3) steam generator tube flaw growth, (4)
reactor coolant pump seal performance under
degraded conditions, (5) performance of pressure
relief mechanisms following anticipated
transients without scram and (6) issues
associated with pressurized thermal shock.
Nevertheless, a skilled PRA analyst is able to
characterize relative risks of various design
features and operating practices at the
plant. Accurate risk insights (used to advocate
an operating practice like increasing technical
specification allowed outage times) must be based
on a full understanding of the contributors to
the PRA results and the impacts of the
uncertainties. Epistemic uncertainty is the
consequence of adding or removing features from
the model. Modeling simplifications introduce
biases and epistemic uncertainty into PRA
calculations. But, even the most artful modeling
will not eliminate all model uncertainty as
increased model detail often relies on less
precisely understood failure-modes.
Raymond Schneider, WestinghouseSteven Farkas,
Hudson Global Resources
2

• Breaks can occur at any location around the
  typical RCS. While this is an obvious statement,
  many analyses used for setting success criteria
  are based on the limiting break location, i.e.,
  at the bottom of the cold leg.
• Large breaks can occur at
• Hot Leg
• Cold Leg
• Crossover Leg (SG to RCP)
• ECCS injection nozzles
• Surge Line
• Bolted Flanges
• Shutdown Cooling Suction
• RPV Head (Upper and Lower)
• Smaller breaks can also appear at
• RCP Seals
• Pressurizer spray lines
• Charging nozzles
• SG manway seals
• CRD Nozzle Penetrations
• Safety relief valve seats

3
Key Uncertainties

• For the ECCS, the SI flow is directed into a
  common header, high or low pressure header
  (dependent on injection pump under
  consideration). The header accepts the pumped
  inventory and distributes the flow to the various
  RCS cold legs. This consideration is important
  when a valve located in the injection line is
  closed or a flow control valve (normally closed)
  does not open. WEC PWRs do not have flow control
  valves and the likelihood of an unavailable
  injection path is low. The header concept is
  important from several perspectives.
• Failure or unavailability of a flowpath will
  increase the net system hydraulic resistance and
  reduce the SI flow to the vessel. Loss of
  injection flow may compromise event success.
• Distribution of SI to all cold legs will assume
  that, for a cold leg LOCA, there are liquid
  inventory losses of injection water from the
  break. For example, during a large cold leg
  break LOCA in the ECCS headered plant with N cold
  legs, something greater than 1/N fractions of the
  injected inventory will spill directly into the
  containment (lower flow resistance into
  containment).

Evolution of Mean LOCA Frequencies (Events per
Plant Year) (Pipe break Contribution only)
PSAs typically consider 3 or 4 break size
designations, i.e., VSLOCA, SBLOCA, MBLOCA and
LBLOCA. The largest break is associated with the
DBA DEGB. VSLOCAs may be important to
ice-condenser PWRs as a result of low containment
pressure actuation setpoints for the
containment-spray system as well as the need for
high-pressure recirculation to mitigate the event
1 Median value 2 Median value 3 Median
value 4 Includes Japanese plant data 5
Estimated values adjusted for mean (NOT FINAL).
Results based on Calendar year. Lowered breaks
impacted by inclusion of SGTR. Based on BWRs
small piping should account for a 5 x 10-4 per
calendar year LOCA frequency. 6 Estimate
includes SGTRs. WEC estimates that SGTR
frequency to be between 0.007 to 0.019 per
operating year. Variation reflects the number of
SGs per plant. It is estimated that, per
calendar year values decrease from SGTR. (See
WCAP-15955, Steam Generator Tube Rupture PSA
Notebook, December 2002.
4
Key Assumptions

• The simplest PRA would estimate the frequency of
  any type of LOCA lumping the very smallest hole
  in the RCS to the very largest. That is, make a
  frequency estimate of any event that directly or
  indirectly caused water to exit from the RCS
  faster than the pressurizer level control system
  could accommodate.
• SGTRs and ISLOCAs are special classes of LOCA
  because instead of discharging RCS water into
  containment (available for recirculation), RCS
  water escapes into the atmosphere or into the
  auxiliary building respectively.
• The PRA model would then have to determine the
  availability and reliability of the SSCs that can
  allow the operators to restore inventory control
  and establish long-term cooling. At a simple
  level, to achieve success (because the LOCA could
  be of any size), the event tree would have to
  include AFW and all the ECCS SSCs from HPSI to
  LPSI to CS to the safety-injection tanks, not to
  mention the support systems like air, cooling
  water, and electric power.
• PRA analysts with even limited experience in
  estimating core-damage frequency (CDF) can see
  that this model could come up with a valid
  estimate of CDF, but that the model would assign
  nearly equal importance to all of the SSCs
  designed to help the operators restore inventory
  control and establish long-term RCS cooling. In
  fact, this is the situation in deterministic
  modeling that simplifies the PRA problem by
  employing defense-in-depth and guarding against
  the worst single-active-failure.
• Some of the LOCAs are so small that the
  availability and reliability of accumulator tanks
  are not relevant to a large fraction of the LOCAs
  captured by the lumped LOCA frequency. A
  sophisticated PRA model will not penalize the
  results by requiring all SITs to be available and
  reliable for SBLOCAs and SGTRs. In fact,
  including SITs in the SSC success set for small
  LOCAs has the perverse effect of lowering the
  core-damage frequency estimate because the SITs
  are inherently reliable.
• The PRA model can instead set up LOCA initiators
  that occur at a nearly infinite number of
  locations around the RCS. Fortunately for the
  PSA analyst, there relatively few permutations of
  AFW and ECCS equipment that can successfully
  restore RCS inventory and maintain long-term
  cooling. A rigorous model would be built by
  determining the frequency of LOCA break sizes
  that can be accommodated by each permutation of
  SSCs that can successfully restore RCS inventory
  and maintain long-term cooling.
• The LOCA break sizes that can be accommodated by
  a particular set of SSCs depends on assumptions
  surrounding the behavior of the water contained
  in the RCS after the break occurs. There are a
  few key features of a thermo-hydraulic model that
  dictate whether or not an SSC will be helpful in
  restoring inventory control and establishing
  long-term cooling. For instance the size of the
  hole dictates how quickly RCS pressure falls
  below the shutoff head of HPSI pumps. The
  location of the break changes how fast the water
  level in the core barrel drops (e.g., cold-leg
  breaks empty the reactor pressure vessel faster
  than a hot-leg break). More subtle assumptions
  in the thermo-hydraulic model determine the mass
  of water in the lower plenum of the reactor
  vessel at the end of the blowdown phase. That
  amount of water determines how much water the
  ECCS has to put back into the RCS to recover the
  core. Of course, neutronics dictates how long
  the fuel can remain uncovered yet retain its
  structural integrity, and thus the flow rate ECCS
  needs to achieve in order to avoid core-damage.
• Some of the LOCAs are a result of general
  transients that cause the primary safety-relief
  valves to lift. As those types of valves have a
  random chance of sticking open, the general
  transient can induce a LOCA putting demands on
  the same large set of SSCs mentioned above.