Title: Jin Jiang
1Jin Jiang Sungwhan Cho Department of
Electrical and Computer EngineeringThe
University of Western Ontario,London, Ontario,
N6A 5B9, Canadajjiang_at_eng.uwo.ca
Analysis and Optimization of Surveillance Test
Interval for k-out-of-n Safety Systems with
Applications to SDS1 in CANDU Reactors
2Outline of the Presentation
3CANDU Plant Overview
4Shutdown System 1(SDS1)
5Surveillance Test Effect on Reliability and
Spurious Trip
- Ten Trip Parameters
- High Regional Neutron Power
- Rate Log High Neutron Power
- Heat Transport System High pressure
- Primary Heat Transport Low Flow
- Reactor Building High Pressure
- Pressurizer Low Level
- Steam Generator Low Level
- Moderator High Temperature
- Heat Transport Low Pressure
- SG Feed line Low Pressure
The required unavailability should be less than
10-3 years/year 8.76 hours/year
63-channel Systems
- Increase Reliability
- Enable On-line Test
7Surveillance Tests
- Benefits of Surveillance Tests
- Increase Overall Reliability
-Recovery
-Detection of Hidden Failure
Test Interval
-Degradation -Wear Out
8Surveillance Tests
- Shortcomings of Surveillance Tests
- Increase Spurious Trip Probability
- Human Resource
- Wear out to System Components
9Methods for Determination of the
Surveillance Test Intervals
- Target Reliability
- - Guides, regulations
- Reliability Analysis Techniques
- - Classical fault tree
- - Dynamic fault tree
- - Markov process models
- Determination of the surveillance test interval
- - Reliability
- - Human resource requirement
- - Cost, wear out
- - Spurious trip probability
10Issues in Determining the Test Intervals
- Test Duration and Recovery Time
- Ignored in classical fault tree models
- Models for Quantifying the Unavailability
- Classical fault tree
- - Describe binary state
- - Cannot model configuration changes
- Dynamic fault tree
- - Configuration changes can be incorporated
- Markov process models
- - Capability of modeling multi-states
- - Mathematical complexity
11State Space Model
- Step 1 Grouping States
- Step 2 Assigning State Transition Rates
- ?f Failure Rate - µTF Test Transition Rate -
µTD Recovery Rate
µTF
?f
µTF
SD Desirable State
SF Undetected Failure State
ST Test State
µTD
12Modeling of states in SDS1
13Probability of normal operation
14Quantification of Risk from Surveillance Test
Performing Surveillance Tests
Decrease Channel Unavailability
Increase Spurious Trip Probability
Quantification of Effect on Core Damage
Probability
15Failure Scenario
Class IV Failure
16CDP with Class IV Power Failure
17Conclusions
- State space models using Markov process
technique are successfully developed for
determination of the optimal test frequency
for redundant emergency shutdown systems. - The states of a system can be decomposed to (a)
the desired, (b) the failed, and (c) the test. - The optimal surveillance test frequency is a
function of the component failure rates and
the test duration. - Progress is underway to study the state
dependence for common cause failures.
18