QUANTIFYING UNCERTAINTIES IN LEVEL 2 PRA

1
QUANTIFYING UNCERTAINTIES IN LEVEL 2 PRA

• Tunc Aldemir
• The Ohio State University

2
Background

• Existing work indicates that uncertainties in the
  timing of events can significantly affect their
  predicted consequences in Level 2 PRAs
• Real plant Level 2 PRAs consist of hundreds of
  simulations to determine the consequences of
  possible scenarios
• These simulations are currently organized and
  analyzed manually
• Approximate processing time is about one man year
• Uncertainty assessment is expected to play a more
  important role in the Level 2 PRA of Generation
  IV reactors

3
A Station Blackout Example for a PWR
No uncertainty, R 1 Surge line fails
Range where SG tubes may fail if uncertainty
introduced
4
BENEFITS OF DYNAMIC EVENT TREES IN LEVEL 2 PRA

• Timing and order of events determined by a
  systems code
• When an event occurs that could threaten
  containment, a mechanistic analysis of the event
  is performed by the systems code rather than the
  non-mechanistic combining of approximations that
  occurs in the quantification of static event tree

5
APPROACH TO DYNAMIC EVENT TREE GENERATION FOR
LEVEL 2 PRA

• A systems code (e.g. MELCOR, RELAP, MAAP)
  determines the pathway of the scenario within a
  probabilistic context
• New scenarios established and followed with
  parallel processing when branching conditions are
  achieved, such as
• pipe ruptures or not based on user supplied
  criteria
• hydrogen burn occurs or not at given conditions
• Dynamic event tree generation process managed by
  a master controller (driver)
• Determines when branching is to occur
• Initiates multiple restarts of system code
  analyses
• Determines the probabilities of scenarios
• Determines when a scenario can be terminated
• Potential to combine similar scenarios to reduce
  the scope of the analysis

6
PROBABILISTIC MODELING OF SEVERE ACCIDENT
PHENOMENA

• Creep Rupture of reactor coolant system
  components
• Surge line
• Hot leg
• Steam generator tubes
• Hydrogen combustion
• Power recovery
• Pressurizer /Safety Relief Valve stuck open
• Reactor coolant system seal leakage
• Containment failure

7
BRANCHING POINTS FOR CREEP RUPTURE MODEL

• Cumulative distribution function of creep
  rupture parameter R is used as the fragility
  curve for surge line, hot leg, and steam
  generator tubes
• The fragility curve is partitioned into specified
  intervals
• The R values corresponding to the bounds of the
  intervals are chosen as branching points
• Two outcomes are assigned to each branching point
  (rupture or no rupture)

8
OTHER MODELS

• Containment failure
• A fragility curve developed as a containment
  failure probability versus containment pressure,
  then discretized for branching points
• Power recovery (in case of station blackout
  accident)
• Various discrete time points of power recovery
  chosen as branching points
• Two outcomes at each branching at each point
  (power recovered and not recovered)
• Hydrogen combustion
• A cumulative distribution function is developed
  for probability of combustion versus
  concentration, then discretized for branching
  points
• Pressurizer/Safety Relief Valve stuck open
• Upon every demand for the valve to close after it
  opens, a branching occurs with
• valve closes on demand, and,
• fails to close (stuck open)

9
SCHEMATIC OVERVIEW OF COMPUTATIONAL INFRASTRUCTURE
10
GRAPHICAL USER INTERFACE

• Two modules
• Static Module
• Display and organize data for existing Level 2
  PRAs
• Tool for concise analysis of simulations that
  have already been run
• Dynamic Module
• Display, control, and organization of dynamically
  created event trees

11
STATIC MODULE
Processed input file stored in database
User Interface displays interactive tree
User Interface processes Input File
User clicks on branch to retrieve additional
information (Plot file path information stored
indefinitely)
12
Static Tree Display Zoom 100
13
Locate Scenarios (Zoom 40)
14
Specify File Path and Choose Variable to Plot
15
Plot
16
DYNAMIC MODULE
Process repeats
Driver executes simulation on Processor 1
Branch point reached, simulation execution stops
Driver dumps info to SQL Database
GUI updates when new database info becomes
available
Driver executes simulation on Processor 2
Simulation time0
Simulation timet1
Process repeats
17
Dynamic Tree for Creep Rupture with Power
Recovery (10 Zoom)
18
CONCLUSION

• Existing work indicates that uncertainties in the
  timing of events can significantly affect their
  predicted consequences in Level 2 PRAs.
• Uncertainty assessment is expected to play a more
  important role in the Level 2 PRA of Generation
  IV reactors.
• A code-independent computational tool is being
  developed for the mechanized and computationally
  efficient generation of dynamic event trees for
  Level 2 PRA.
• Tests to date indicate that the tool can handle
  up to 200 simultaneous scenarios using MELCOR as
  the system simulator.
• The graphical output interface provides
  information regarding the status of the runs,
  relative duration of the completed runs, event
  hierarchies and likelihood of scenarios, as well
  as plots of parameters of interest.
• The static version of the graphical output
  interface have been tested with 20 GByte output
  of an existing Level 2 PRA .
• Future work will include epistemic uncertainty
  quantification using an outer loop with Monte
  Carlo sampling.

19
?
QUESTIONS
?
?
?
?
?
Thank you !!!