ARA Presentation, Tornadoes and High Wind …HW PRA Family of Tornado Hazard Curves 1.E-08 1.E-07 1.E-06 1.E-05 1.E-04 1.E-03 1.E-02 0 50 100 150 200 250 300 350 Exceedance Probability

Lawrence A. Twisdale, Jr., P.J. Vickery, J.C. Sciaudone

Sudhan Banik, David MIzzen

Tornadoes and High Wind Hazard/Fragility Analyses for Nuclear Power Plants

Technical Discussions NRC-ARA

May 28, 2015

Applied Research Associates, Inc. 8537 Six Forks Rd, Suite 600 Raleigh, NC 27615

Copyright ARA 2015. Do not reproduce or distribute without ARA written consent.

2 Copyright ARA 2015. Do not reproduce or distribute without ARA consent.

Prelude

We will cover some of these hazards/effects.


Outline 1. Hazard Curve Estimation and Uncertainties

a. Straight Wind/Hurricanes: Vickery b. Tornadoes: Twisdale

2. Plant Walkdowns: Sciaudone 3. Fragility Analyses

a. Code-Based Wind Pressure Fragility: Twisdale b. Detailed Progressive Failure Modeling: Vickery c. Missiles: Twisdale/Sciaudone

4. Conditional Loop for High Winds: Twisdale 5. PRA Integration: See PSA 2015 paper

Many topics. Overview Discussion. ARA is very interested in developing research projects to address many of the

industry needs/issues in the High Wind area.

4

1a. Straight Winds/ Hurricanes

5

Straight Winds/Hurricanes

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

1.0E-03

1.0E-02

1.0E-01

1.0E+00

0 50 100 150 200 250 300

Annu

al E

xcee

danc

e Pr

obab

ility

Peak Gust Wind Speed at 10 m in Open Terrain (mph)

NominalMeanMedian5th and 95th

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

1.0E-03

1.0E-02

1.0E-01

1.0E+00

0 50 100 150 200 250 300An

nual

Exc

eeda

nce

Prob

abili

ty


NominalMeanMedian5th and 95

Extratropical Cyclones

Thunderstorms

For this site, 1E-03 winds are about 85 vs 115 mph

1. State of Practice: analyze separately (see ASCE 7)

2. Major benefits: i. Reduces uncertainties

~900 vs 30 events ii. More accurate iii. Low Cost to implement

3. Thunderstorms typically dominate in Eastern US

4. Two loop Monte Carlo code developed

5. Modeling Uncertainties include: anem. ht. cor, terrain, effective gust duration, and extreme value parameters

6. Nominal means well below derived means

6

Time Series of Thunderstorm and Non-Thunderstorm Wind Speeds

05

101520253035404550

J-75 J-76 J-77 J-78 J-79 J-80 J-81 J-82 J-83 J-84 J-85 J-86 J-87 J-88 J-89 J-90 J-91 J-92 J-93 J-94 J-95 J-96 J-97 J-98 J-99 J-00 J-01 J-02 J-03 J-04 J-05 J-06 J-07 J-08 J-09 J-10 J-11 J-12

Peak

Gus

t Win

d Sp

eed

(m/s

ec)

Month and Year

0

5

10

15

20

25

30

35

J-77 J-78 J-79 J-80 J-81 J-82 J-83 J-84 J-85 J-86 J-87 J-88 J-89 J-90 J-91 J-92 J-93 J-94 J-95 J-96 J-97 J-98 J-99 J-00 J-01 J-02 J-03 J-04 J-05 J-06 J-07 J-08 J-09 J-10 J-11 J-12

Peak

Gus

t Win

d Sp

eed

(m/s

ec)

Month and Year

7

Wind Speed Adjustments All wind measurements adjusted for anemometer height,

averaging time and terrain Anemometers often located in less than ideal conditions for

measuring open terrain winds, requiring terrain corrections Historical gust wind speeds in

the US have had differing measurement systems

8

Stochastic Models Used to Develop Hazard Curves NOT Annual Extremes

Thunderday peak gusts are fit to an extreme value distribution

Annual thunderstorm hazard curve

developed using a Poisson arrival rate assumption

λ is the average number of thunderdays/year

𝑃𝑃𝑎𝑎(𝑣𝑣 > 𝑉𝑉) = 1− 𝑒𝑒𝑒𝑒𝑒𝑒[−𝜆𝜆𝑃𝑃(𝑣𝑣 > 𝑉𝑉|𝑇𝑇])

9

Stochastic Models Used to Develop Hazard Curves NOT Annual Extremes

Extratropical storm model uses method of independent storms using exceedances of a threshold wind speed separated by at least two days. Data fit to Type I as for thunderstorms.

Annual extratropical hazard curve developed assuming independence of storms so that

r is the average number of storms/year

10

Uncertainties

11

Uncertainties All uncertainties propagated through a simulation code to

generate a family of hazard curves

12

Directional Analysis of Str. Wind Hazard 1. Straight winds are typically

analyzed on an “all direction basis”.

2. The all-direction approach is conservative.

3. We have used directional wind analysis for vulnerable structures to reduce the conservative margin.

4. Directional analysis has been used in the wind tunnel testing area for decades for high rise structural design.

5. Directional wind hazard analysis requires direction dependent fragilities. Should be used to reduce conservative margin for straight winds, as necessary.

13

Rain Correlation with High Winds

𝐹𝐹(𝑉𝑉) = 1

1 + 𝑒𝑒−(𝛽𝛽0+𝛽𝛽1𝑉𝑉)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0 10 20 30 40 50 60 70 80 90 100

Prob

abili

ty o

f Rai

n

Peak Gust Wind Speed (mph)

Thunderstorms

Rain Prob. Day1 After Damaging TS WindsRain Prob. Day2 After Damaging TS WindsRain Prob. Day3 After Damaging TS Winds

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1 (Day of HW) 2 3 4

Cum

ulat

ive

Prob

abili

ty o

f Rai

n

Day

Cumulative Probability of Rain Given High Winds

Maximum CumulativeProb of Rain - LogisticModel

72 Hour Probabilty

1. High winds are often accompanied by rain.

2. A conservative approach is to assume rain accompanies high wind and that vulnerable electrical equipment fails as soon as the building envelope begins to fail.

3. One way to reduce this conservatism is to perform a site-specific wind- rain correlation analysis. This approach is conservative. The highest winds do not always come from the worst loading direction on a structure.

4. We used 50 mph as the threshold to develop wind-rain correlations for both thunderstorms and extratropical cyclones.

5. Logistic regression models were fit to the data.

14

Hurricane Simulation Approach

15

Hurricane Hazard Modeling

1. Can be screened based on distance from the nearest coast

2. Mathematical simulations are the widely accepted approach

3. NUREG and ASCE 7 are based on Vickery et al model

4. Results from NUREG can be used with ASCE 7 information to develop a good nominal mean hazard curve for a US site.

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

1.0E-03

1.0E-02

1.0E-01

1.0E+00

0 50 100 150 200 250 300

Annu

al E

xcee

danc

e Pr

obab

ility


MeanMedian5th and 95th

16

Stochastic Storm Surge Modeling

• Use coarse grid model (ADCIRC or SLOSH) model 1,000,000 or so years of simulated storms

• Down select a few thousand or so large storm surge producing events and run finer grid coupled storm surge-wave-tide models to develop flood hazard curves

17

Research Needs Need downburst quantitative regional down climatology's to

model the wind hazard curve transition between the thunderstorm hazard and the tornado hazard

This affects straight wind hazard risks, which are dominant contributors to HW PRAs

Probabilistic storm surge analyses need to be performed instead of using the maximum credible hurricane approach Validated hurricane model must be used

A detailed combined wind-storm surge analysis should be done for a coastal plant

Timing of effects will influence PRA and HRA

18

1b. Tornado Hazard Modeling

19

Overview 1. New work is needed to update the industry’s

guidance on tornado wind risk . (NUREG CR/4461 Rev 2 appears to be unconservative).

2. There is a significant amount of new data available over the past decade.

3. NIST is funding ARA to develop tornado risk maps for the US. Expected to be part of ASCE 7. All areas that affect tornado wind hazard risk

are on the table for investigation. 4. Significant new amounts of information are being

considered in the research. Funding from NRC could be instrumental in supporting effort and developing the next generation of tornado risk analysis methodology and windspeed risk.

5. The work will be published in journals and conference proceedings and provides for inputs/review comments from Stakeholders.

NIST Project: Tornado Risk Maps for Building Design Year 1: Framework

and Research Year 2:Research

and Preliminary Results

Year 3: Research and Tornado Windspeed Hazard Risk/Maps (2017)

20

Size of Building/Plant on Tornado Strike Risk

Tornado Risk Depends on Target Areas—Important to Nuclear Plants

21

Tornado Hazard Comparison

1.00E-08

1.00E-07

1.00E-06

1.00E-05

1.00E-04

1.00E-03

0 50 100 150 200 250

Exce

edan

ce P

roba

bilit

y (p

er y

ear)

Windspeed (mph)

Nominal Probabilistic TORRISK Calculations vs NUREG

Union (Plant) EF

Nominal Point EF

NUREG EF

Intersection EF

1. TORRISK vs. NUREG v 2 (2007) 2. Plant results are based on TORRISK

model for 1950-2013 SPC data. Does NOT include any NIST enhancements.

3. NUREG Ignored reporting issues and time trends Fixed 2 deg boxes with broad area tornado

data… not fully site-specific No windfield model Many other details in data analysis Target = 200ft x 200ft building

4. For many plant site analyzed, NUREG is always less conservative than TORRISK/TORMIS site-specific results.

5. NIST research is pointing toward unconservatisms in many of the tornado modeling inputs that have been used in the past.

Nominal EF Probabilistic Analysis

Point vs. Plant

NUREG

22

HW PRA Family of Tornado Hazard Curves

1.E-08

1.E-07

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

0 50 100 150 200 250 300 350

Exce

edan

ce P

roba

bilit

y (p

er y

ear)

Peak Gust Windspeed (mph)

Tornado Plant (Aleatory + Epistemic)

5th

25th

50th

Mean

75th

95th

Tornado Plant Hazard Family

1. Tornadoes are characterized by extremely large epistemic uncertainties.

2. Many factors affect the uncertainties 1. SPC database includes

many eras of changing reporting guidance. Numerous biases exist and must be corrected.

2. Path area uncertainties 3. Windspeed relationship to

F/EF scale. EF scale is not the gospel. There are many issues. Needed modeling/analysis.

23

Nominal vs Derived Mean Hazard 1. TORRISK Derived Mean is

much larger than nominal probabilistic analysis curve at high winds

2. Detailed simulation models are needed to develop tornado hazard curves and propagate uncertainties.

3. Industry lags behind in state of art modeling (see recent research and publications).

4. Tornado hazard curves are different for wind pressure vs missile fragilities (larger strike area) 1.00E-08

1.00E-07

1.00E-06

1.00E-05

1.00E-04

1.00E-03

- 100 200 300 400

Exce

edan

ce P

roba

bilit

y (p

er y

ear)

Windspeed (mph)

Tornado Derived Mean vs Nominal Mean Comparison (point)

Point Derived Mean

NUREG F

Nominal Point EF

NUREG EF

Plant Derived Mean

Difference factor = 10 to 100 at 200 mph


Year One NIST Tornado Research Plan

1. Develop Framework

2. Detailed literature review and sensitivity analysis

3. Many database issues

4. Update TORRISK

5. Spatially dependent targets

Task 3. Quantify Tornado Risk MetricsPoint Probability Sensitivity AnalysisSpatial Probability

• Municipality• EOC• Shelters• Infrastructure• Critical Facilities

Sensitivity Calculations • Rank Order• Compare

Task 5. Stakeholder Inputs(Workshop in WDC Area

~ September 2014)

Task 4. Document Results(Framework)

Tornado Risk Maps for Building DesignYear One: Framework

Task 1. Project Plan

Existing Models

TORRISK (Tornado Risk)TORMIS (Missile Risk)TORLOSS* (Tornado Loss)TORLINE (Outbreak/Lifeline)CLSPT (Statistics)SAS (Statistics)ARC-GIS (Mapping)FCALC (Fragility)NIFF MSTAT (Missile Stat Model)HAZUS (Vulnerability/Loss)* Under Development

Task 2. Assess Tornado Data and Risk MethodologiesLiterature/Data Review

1. Database2. Occurrence Rate/Trends3. Wind Speeds & F/EF

Damage Scales4. Radar Data/Findings

5. Path Length Intensity Variations

6. Path Variables7. Windfield Variables8. Mapping Approaches

Analysis Plan1. Findings2. Preliminary model updates3. Sensitivity Study Inputs

Year Two


Tornado Summary 1. Industry update of tornado hazard is needed 2. The NIST project is integrating many new data sources, including Radar

and other data sources provide the technical basis for significantly improved tornado hazard modeling. Key updates: 1. Reporting eras 2. Population bias, random damage encounters 3. Path length intensity – Catalogue of 1000s of tornadoes 4. EF0 bias 5. Other Source Rating Bias 6. Windfield model parameter updates 7. Probabilistic damage to windspeed development with engr models. 8. Radar data on windspeeds and windfield parameters

3. NIST project provides a good opportunity with sufficient resources for NRC to leverage 1. Significant update, next generation modeling 2. APC loads 3. Windspeed damage relationships for purposes of using database

and developing wind hazard risk 4. Could develop specific NRC deliverables, such site evaluation, APC

load risk modeling of uncertainties


Combined Wind Hazard

1.E-08

1.E-07

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

1.E+00

0 50 100 150 200 250 300 350

Annu

al E

xcee

danc

e Pr

obab

ility

(per

yr)

Windspeed (mph)

Family of All Wind Hazards (Plant)

5th

25th

50th

75th

95th

Mean

Gap-----Missing SCS events-downbursts, tornadoes?

Recommended research project to fill gap for under-reported downbursts,micro-bursts climatology

Current gap produces reverse curvature in combined wind hazard curves

27

2. HW PRA Plant Walkdowns

28

Walkdown Planning Walkdown Objectives – Meet Requirements of ASME Standard

• Use plant-specific data (WFR-A1) to evaluate SSCs for inclusion in wind pressure and missile fragility analyses (WFR-A2)

• Use a high wind missile hazard analysis methodology (WHA-A4) • Inventory potential sources of wind-borne missiles (WHA-A5)

Include Team Members Familiar with: • Wind engineering and wind failure of structures • Equipment list and Plant Response model • History of wind damage at the plant • Typical plant operation including Planned outages Planned construction Planned changes in plant configuration

29

SSC Walkdown List

Locked Open Manual Valve included in walkdown list but

not the piping

Air Operated Valve (Fragile)

Initial walkdown lists generally include hundreds of basic events from the plant’s internal events PRA and were not necessarily developed with wind analysis in mind.

30

As-Built Details

Panel Width = 0.6 m

0.2 m 0.4 m 0.4 m 0.2 m

Roof Deck

Roof Purlin

Weld Locations

12” weld spacing, but center weld sees

33% more wind load!

Wall not connected to building frame

31

Structural Interaction Deepwell Pump and Generator located next to

Fabrication Shop

Cladding and contents treated as

missiles

Frame treated as structural interaction

32

Vulnerability to Water Exposure

Spray shields protect from drips, but not wind driven rain following cladding failure

33

Condition Assessment – Siding Fasteners

25% of connection points had at least 1 missing screw 50% at Turbine Deck Level

34

Category I Structures – Doors

Verify Wind, APC and Missile Design on Plant Documents and Drawings

Look for Missile Barrier behind Door & Evaluate Missile Paths through Door

35

Category I Structures – Vents, Intakes & Exhausts

36

Missile Zone Identification Consider:

• 2500’ Radius • Homogenous areas • Local landmarks

like road and fence lines

Organization is Key: • Ensure all areas and

non-Cat I Structures considered

• Know where you are in the plant at all times

37

Plant Specific Missile Identification

Thousands of pavers found on Aux Bldg Roof Site specific missile survey requirement from ASME TORMIS RIS and SER requirements

38

Operating vs. Outage Conditions

39

SSC Screening Results

TB

Fra

me

TB

Wal

l

TB

Roo

f

SB F

ram

e

SB W

all

SB R

oof

SSF

Fram

e

SSF

Wal

l (bl

ock)

SSF

Roo

f

SSF

Lou

vers

Modeling Approach / Screening Reason

Hit

(V>V

tg)

Perf

orat

ion

Spal

l

Cru

shin

g

Cru

shin

g (L

g M

issi

le)

Pipe

Pen

etra

tion

TO

RM

IS T

arge

ts

Gro

up N

umbe

r

Gro

up D

escr

iptio

n

16 6900/600 V ac Transformer 1SLXB Fails 1STXB Yes Service Building 760 X X X286 600 V ac Load Center Bus 1SLXB (added following walkdown) 1SLXB No Service Building 760 X X X11 600 Vac Load Center Bus 1SLXC 1SLXC No Service Building 760 X X X17 6900/600 Vac Transformer 1SLXC Fails 1STXC No Service Building 760 X X X12 600 V ac Load Center Bus 1SLXD 1SLXD No Service Building 760 X X X18 6900/600 V ac Transformer 1SLXD Fails 1STXD No Service Building 760 X X X13 600 V ac Load Center Bus 1SLXF 1SLXF No Service Building 760 X X X19 6900/600 V ac Transformer 1SLXF Fails 1STXF No Service Building 760 X X X

54 Manual Valve 1CA225 1CA225 No Service Building roof X X X X

55 Manual Valve 1CA226 1CA226 No Service Building roof X X X X

38 120 V AC Power Bus SKG SKG Yes SSF 760 X X X X X14 600 V ac Load Center Bus 1SLXG 1SLXG Yes SSF 760 X X X X X20 6900/600 V ac Transformer 1SLXG Fails to Function 1STXG No SSF 760 X X X X X39 600 V ac MCC SMXG Bus SMXG Yes SSF 760 X X X X X40 Ac Power Bus SMXG-1 SMXG-1 Yes SSF 760 X X X X X46 Check Valve 1AD2 1AD2 Yes SSF 760 X X X X X47 Manual Valve 1AD4 1AD4 Yes SSF 760 X X X X X44 Manual Valve 1AD18 1AD18 Yes SSF 760 X X X X X45 Manual Valve 1AD19 1AD19 Yes SSF 760 X X X X X48 Standby Shutdown Fuel Oil Day Tank AD00DT Yes SSF 760 X X X X X49 Standby Shutdown Fuel Oil Day Tank Pump AD0DTPP Yes SSF 760 X X X X X50 Standby Shutdown Fuel Oil Filter AD0FOFFL Yes SSF 760 X X X X X51 600/120 V ac Transformer SKG SKGX Yes SSF 760 X X X X X88 Battery SDSP1 SDSP1 Yes SSF 760 X X X X X90 Battery SDSP2 SDSP2 Yes SSF 760 X X X X X186 SSF Diesel Generator 0ADGE0005 Yes SSF 760 X X X X X X187 Battery (24 V Diesel Start) SS0SSBAT Yes SSF 760 X X X X X188 Battery Charger Failure Prior to Event Drains Battery SS24VDBAT Yes SSF 760 X X X X X185 600/120 V ac Transformer 1SSFARC 1SSFARC Yes SSF 760 X X X X X254 Tank (Day Tank) Vent Pipe 1A Fails Due to Damage by Missile ADVPDTA Yes SSF 760 X X X X X255 Tank (Day Tank) Vent Pipe 1B Fails due to Damage by Missile ADVPDTB Yes SSF 760 X X X X X89 Battery Charger SDSP1 SDSP1-BC No SSF 777 X X X X X91 Battery Charger SDSP2 SDSP2-BC Yes SSF 777 X X X X X87 125 V dc Distribution Center SDSP Bus SDSP Yes SSF 777 X X X X X

5 6900/4160 V ac Transformer 1ATC 1ATC Yes U1 Turbine Building 760 X X XHit with Minimal Velocity on

Electrical EquipmentX 77 27

1ATC Transformer in U1 TB

6 6900/4160 V ac Transformer 1ATD 1ATD Yes U1 Turbine Building 760 X X XHit with Minimal Velocity on

Electrical EquipmentX 78 28

1ATD Transformer in U1 TB

Electrical vulnerable to failure of TB Wall Cladding, Roof Deck, or

Frame.

Equipment located directly below SB roof is Vulnerable to Roof

Deck or Frame Failure. Walls of Room are Interior to SB and TBs.

Electrical equipment vulnerable to failure of exterior walls, louvers, and roof deck. Also vulnerable to

SSF Frame failure.

Wind Pressure FragilityWind Missile Fragility (Damage)

Direct wind for CACSTs, Interaction with SB Roof and

Frame for CACSTs and PipingE

leva

tion

(ft)

Ori

gina

l Seq

uenc

e N

umbe

r

HW Description HW ID

Cre

dit D

urin

g L

OO

P

Location

62-74 24 SSFPerforation of Walls, Roof, Doors, and Louvers. Crushing of Diesel

Generator Exhaust

55-56 22Service Building Electrical Rooms

Perforation of Concrete Block Walls and Metal Roof Deck

2357-61CA Condensate

Storage Tanks on Service Building Roof

Model CACSTs and Associated Piping for Perforation

Dir

ect W

ind

Interactions

Modeling Approach / Screening Reason

Example Plant – 289 SSCs Walked Down – 72 screened out of both wind pressure and missile fragility, 70 included in both, 95 missile only & 52 wind pressure only

40

Walkdown Summary High wind walkdowns require detailed reviews of the entire site

• Wind design and/or analysis experience is required • Starting walkdown lists can be extensive and not based on wind resistance • Include a mapping of wind pressure and missile fragilities to PRA events • As-built details and conditions have to be documented • Missile survey must address all potential missiles and operating conditions

Industry Needs • Develop industry guidelines and training materials for HW PRA walkdowns

that address: SSC identification and evaluation Consistent screening criteria for wind pressure, wind missiles and APC Missile identification and survey procedures Standardized data collection forms and databases Water exposure criteria for electrical systems

• What is a credible missile path

41

3a. Fragility Analysis-Wind Pressures

42

Code-Based Wind Pressure Fragilities

For a building → 6 Basic Wall Fragilities, 9 Roof

43

Calculational Flow

Enclosure States: Fully Enclosed, Partailly Enclosed, Open

44

Example WP Mean Fragilities

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

50 150 250 350

Frag

ility

Windspeed (mph)

Mean Roof Fragilities

Turbine Building Roof Building 1 Roof Service Building Roof

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

50 150 250 350

Frag

ility

Windspeed (mph)

Mean Wall and Frame Fragilities

Turbine Building Wall Bldg 1 Frame

Service Building Walls/Frame Turbine Building Frame

Building 1 Louvers

45

Example WP Fragility Families

46

Doors vs Roof and Walls Large door much more

vulnerable than walls or roof.

Small door more vulnerable than walls.

Roof more vulnerable than walls and small door.

Large wall uncertainties due to directionally dependent blockage/shielding

Mean curve is derived mean over uncertainties (not nominal)

47

Summary- Code Based WP Fragilities 1. ARA has developed a general code-based methodology for wind and tornado fragilities

with significant enhancements vs. Kennedy and Ravindra approach (seismic) 2. Does not require assumption of lognormal fragilities; produces non-parametric

numerical fragilities. 3. Method uses high level design and construction characteristic inputs, based on code

load/resistance and other probabilistic factors. Calculations are performed to develop inputs to a Monte Carlo code to reflect strength margin or resistance, design code vs modern code, wind, and terrain.

4. ARA FACLC code uses two loop simulations to treat modeling uncertainties and randomness.

5. Treats wind pressure zones, different material types, and enclosure states. Frame failure fragilities reflect geometry change from failure of walls/cladding.

6. Basic fragilities are developed and automatically combined to produce a wall system fragility, roof system fragility and frame fragility. Simple, but physically consistent system model relationship for envelope failures and frame failures.

7. Separate non-building fragilities are developed for doors, openings, towers, tanks, etc 8. Additional work is needed to include WBD effects on wind pressure fragilities. 9. Detailed knowledge of building codes is needed to produce WP fragilities. Does not

typically exist in industry.

48

3b. Fragility Analysis- Progressive Failure Modeling

49

Progressive Failure Modeling Methodology uses 3-D models of buildings and includes the

modeling of beams, girts, columns, purlins, wall cladding, rood decking and doors/windows.

Ideally coupled with design drawings with information on member types and sizes, spans and support conditions. Walkdowns provide much needed information on wall cladding connections, missing fasteners and validation of drawings.

Approach considers that each of the components have different wind resistive capability (e.g., cladding vs. girts vs. main wind force resisting system).

Failure of weaker components leads to failure of additional components (internal pressure)

50

Directionally Dependent Pressure Coefficients (GCp)

###

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

1

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

2

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

3

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

4

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

5

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

6

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

7

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

8

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

9

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

10

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

11

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

12

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

13

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

14

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

15

-1.5

-1

-0.5

0

0.5

1

0 45 90 135 180 225 270 315 360

GCp

Wind Direction

16

51

Time Stepping Approach for Failure Modeling

No

Yes

Begin N storm simulation

Begin M time steps for ramping of wind load

Yes

No

Yes No

Sample:

• Component resistances • Pressure coefficients • Wind load modeling errors

• Compute external pressure • Add internal pressures to get

total loads

• Compare computed loads with resistances

• Check for failure of wall panels, girts and roof panels

Panel failure? = M Time steps?

Compute damage statistics for a given storm

Recompute internal pressure

Compute wind speed for the time

step

Check LOOP event

= N Simulations?

52

Modeling Internal Flow and Internal Pressure

Volume flow in/out of opening

Total Volumetric Flow Must be Zero

Background leakage to allow flow into building with one opening

53

Example Fragilities (Building A)

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

1.E+00

100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300

Cond

ition

al P

roba

bilit

y of

Fai

lure

Peak Gust Wind Speed (mph) at 10 m in Open Terrain

Wall Cladding

Roof Deck

Frame

Roof

54

Example Fragilities (Building B)

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

1.E+00

100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300

Cond

ition

al P

roba

bilit

y of

Fai

lure

Peak Gust Wind Speed (mph) at 10 m in Open Terrain

Wall CladdingRoof DeckFrameRoof

55

Performance Summary Building A:

• Wall cladding attached to light gauge sub girts and the attachment. • Cladding fails by fastener pull trough • Failure of sub-girts/cladding begins at peak gust wind speeds of 120 mph • Roof beck begins to fail at ~180 mph

Building B • Some long span roof beams begin to fail at wind speeds of 120 mph, taking

large sections of roof decking • Wall Cladding begins to Fail at 140 mph • Building contains safety related equipment

56

Wall Cladding Fragility Curves by Wind Direction

Results compared well to past panel failures in 70-80 mph winds

57

Bare Frame Loads

58

Bare Frame Loads Can be Higher than a Fully Clad Build Building

Ratio of Fully Clad to Bare Frame Loads

59

Summary Progressive element-by-element failure models reveal

weaknesses in the system that code-based approaches can miss The approach readily deals with construction/maintenance issues

(missing fasteners leading to panels failing at low wind speeds) We have found flaws in in most plants we analyzed Fragility functions often not well representative with log-normal

distributions Bare frame loads often greater than loads on fully clad structures.

The cladding blow-off practice used to design building in the 1960’s and 70’s and later was ill advised

60

Progressive Failure Modeling

TB with 1000+wall panels Directionally-dependent Pressure Coef. for 16 Wall Zones

Sequential failure with changes in enclosure state, geometry, and loads on the structure

61

3c. Fragility Analysis- Missiles

62

Modeling of Missile Fragilities Missile fragilities require detail modeling (winds, aerodynamics,

large number of missiles, 3D geometries, the dynamics of missile impacts,…..)

Simplified methods are prone to enormous errors (orders of magnitude 10 to 1000 or so) for any individual target.

We use TORMIS to develop missile fragilities. Advances include: Missile source modeling Stochastic modeling to represent different plant states (outage, non-

outage) Modeling of equipment inside vulnerable structures Missile impact damage models Modeling of “pipe-penetration” type SSCs Use of statistical replication to quantify statistical convergence of missile

fragilities

63

Tornado Wind Model Simulations Tornado Wind Model Descriptive Parameters 1. Rmax 2. Translational

Speed 3. Radial,

Horizontal, and Vertical Velocity Components

4. Central Pressure Drop

5. Velocity Profile

64

Missiles and Trajectory Model


Example Missile Geometries Basic

Aerodynamic Shape Set

Basic Missile Sets Subset on d (in.) General

Description Cross-Section b/d Variation

Impact Material

Final Set Number a b c d e

Cylinder

Rod

Steel 1 <1 (1,2] (2,12] (12,20] >20

Wood 2 <13 (13,17] (17,49] >48

Pipe

Steel 3 <3 (3,6] (6,12] (12,24] >24

Concrete 4 >0

Rectangle

Box, Beam

Steel 5 <24 (24,48] >48

Wood 6 <6 (6,12] >12

Steel 7 <4 >4 Concrete 8 >0

Wood 9 <12 >12

Plate

Steel 10 <24 (24,72] >72

Wood 11 >0

Steel 12 <36 (36,72] >72

Wood 13 <48 >48

I-Shape Wide Flange

Steel 14 <6 (6,12]

Angle Angle

Steel 15 >0

Channel Channel

Steel 16 <8 (8,20] >20

Concrete 17 >0

Frame, Truss

Pipe Frame

Steel 18 >0 Rect. Frame Steel 19 >0

Rect. Frame Wood 20 >0

Pipe Frame

Steel 21 <54 >54 Rect. Frame Steel 22 <48 >48 Rect. Frame Wood 23 >0

Sphere Sphere

Steel 24 >0

Vehicle Auto, Trailer

Steel 25 >0

Tree Tree

Wood 26 >0

d

d

d

d

b=d/4d

d

b=d/10

d

b=d/50

d

d

d

d

d

d

d

d

d

d

d

• TORMIS has a built in aerodynamic Library for:

•Cylinders •Spheres •Box Shapes •Auto Vehicle •Tree/branch •Wide Flange •Angle •Channel •Plates •Truss (box with open areas)

• New missiles can be added according to these shapes •The missiles are flown with a RO-6DOF model, verified in the research phase

66

TORMIS Modeling • Includes Safety (red), Shielding (grey), and

Missile Source (orange) targets • Plant specific tornado hazard inputs and

missile zone layouts • Produces missile fragilities conditioned on

wind speed

• Follows TORMIS SER and RIS


Missile Sources: Zone and Structure Sources

68

Crimping Targets on DG Bldg Roofs

69

What’s Behind/Beneath Openings?

70

Large Doors: Major Vulnerability

71

Valve Examples

Manual Valve

Motor Operated Valve Check Valve

Solenoid Operated Valve

Pipes

72

Air Intake Labyrinth

73

Missile Fragility Example for 12” Concrete Wall Target and Door (Vul Area = 1100 sq ft)

74

Masonry Wall Perforation Fragility (Vul Area = 10,500 sq ft)

75

Dynamic FEA- Crimping of Exhaust

For complicated targets, failure modes are analyzed and the critical impact velocities are input to TORMIS for each missile type

76

Example Missile Fragilities

Perforation of Small Generator Building Crimping of Vent Stack

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

1.0E-03

1.0E-02

1.0E-01

1.0E+00

0 50 100 150 200 250 300

Frag

ility


Missile Fragility -- Small Generator Building

Low (5th Percentile) Nominal Mean High (95th Percentile) Nominal Mean (Hit)

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

1.0E-03

1.0E-02

1.0E-01

1.0E+00

0 50 100 150 200 250 300

Frag

ility


Missile Fragility -- Crimping of Vent Stack


77

Pipe Penetrations

29

30

31

33

34

3536

3738

39

40

NRC Inspectors raised issue

Over 40 Pipe Penetrations modeled for some plants

78

Pipe Penetration Fragility

PP Opening – Door behind a door PP Fragility

EssentialEquipment

“Pipe Penetration” opening1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

1.0E-03

1.0E-02

1.0E-01

1.0E+00

0 50 100 150 200 250 300

Frag

ility


Missile Fragility -- Pipe Penetration Target


79

Missile Fragility Convergence

1.0E-04

1.0E-03

1.0E-02

1.0E-01

0 10 20 30 40 50 60

Mis

sile

Fra

gilit

y

Number of Replications

Convergence for 135 mph

Individual Runs Running Mean Mean - 2 Std Err Mean + 2 Std Err

80

Simplified Missile Statistical Models Have Large Errors

Missile Hit Missile Damage

81

Missile Fragility Summary (1 of 2)

Detailed 3D modeling is necessary for accurate missile fragilities(statistical models have large potential errors).

TORMIS needs to be updated and made into a usable code. Many missile related analyses with simple tools,

assumptions or hand calculations are not reliable or necessarily conservative.

Performing 3D missile simulations is computationally cheap, industry lacks widely available tools.

82

Missile Fragility Summary (2 of 2) A missile project could improve guidance and inputs for plant

screening, missile protection needs and HW PRAs: • Expansion of damage calculation abilities, including: Crimping velocities for typical exhausts, vents, and intakes Fragility of heavy equipment like pumps, motors, and transformers Fragility of conduits and cable trays

• Use TORMIS to develop missile fragility screening criteria Minimum thickness of concrete or steel walls/roofs/missile barriers for

exclusion from missile fragility analysis Base screening criteria on wind speed – for example, minimum steel thickness

of target to exclude from fragility analysis for F1 level wind speeds Line of sight and non-line of sight vulnerabilities

• Update TORMIS to include straight line wind fields to develop straight line and hurricane missile fragilities

• Such guidance could help teach industry personnel about the fundamentals of wind missile fragility analysis

83

4. HW Produced Loop

84

LOOP Failure Mode Analysis 1. High Winds at the site, above a probabilistic threshold, produce a

LOOP 2. The power lines, supporting structures, and SY structures were

typically designed for winds with return period of about 1E-02 3. None of these exposed SSCs, including transformers, are

designed for missiles. 4. Hence, there is a serious LOOP vulnerability to HW events 5. Not well understood in industry; many want to dismiss the LOOP

since “meteorological events” are represented in the weather caused loops in Interanal Event PRAs.

6. Metal clad structures near the offsite power SSCs add to the problem. The metal panels are typically 20-30 ft long, whereas the separation of the

conductors is much less.

85

Example LOOP Fragility Analysis TB, Conductors, and Main Transformers

Missile Produced Electrical Shorting Fragility

86

Missile Proximity to Typical Switchyard

87

Example LOOP Contributors

1E-4

1E-3

1E-2

1E-1

1E+0

75 125 175 225 275

LOO

P C

ontri

butio

n

High Wind Interval Midpoint (mph)

High Wind LOOP Contribution

Tornado PressureNon-Tornado PressureCombined PressureMissileCombined LOOP

88

LOOP Summary High wind induced LOOPS have had a significant impact on CDF

produced for several HW PRAs Power lines, towers, switchyards, transformers are vulnerable to

high wind effects. Performance of transmission towers are typically marginal with

respect to design and they have little redundancy beyond the design basis.

Many unknowns on tower capacity for plants built 30+ years ago. Plant HW PRAs don’t have resources to assess these fragilities in detail.

Industry study is needed to analyze these typical systems and how large an area/grid around the plant can be affected for different types of windstorms.

• Duration of outage • Fragility methods • Example results

89

5. Summary

90

Item Number Error Comment

5 ~ 100%

10 ~ 35%

15 ~ 10%

2. Windspeed Range negative Always an underestimation error. Use a range from 50 to 300+ mph.

3. Modeling Separate Wind Hazards vs Combined Wind Hazard

< 10% Does not include errors from CAFTA approximations.

4. Use of Derived Mean vs Integration of Family

< ±15%Approximate quantification with

only 4 cases. Examples herein suggest an overestmation error.

10 -34.9%

100 -36.6%

1000 -36.8%

10 900%

100 9900%

1000 99900%

1. No. of Windspeed Hazard Intervals

5. M Negatively Correlated Failure Modes

6.M Positively Correlated Failure Modes

Maximum Theoretical Error Peaks at 1/M, Approaches → -1/e (Actual errors may be much less)

Maximum Theoretical Error (M-1)*100%

(Actual errors may be much less)

Variable spacing is suggested

Summary of Max FF Calc Errors (see PSA2015 paper by Twisdale, Lovelace, and Slep)

Bound is -1/e

Unlimited bounds; must be estimated with physical simulation model.

91

FF Calcs and PRA Integration-Summary

1. Most sites will have 2-3 wind hazards that dominate the wind hazard risk. Hence the HW PRA must combine results across multiple wind hazards.

2. The number of wind speed intervals should be ≥ 10. We estimate that about 15 properly spaced intervals will produce over-estimation errors < 10%.

3. The use of the derived mean hazard and derived mean fragility curves to estimate failure frequency is accurate to about ± 15% when compared to an exact integration of the families of curves.

4. Due to presence of both weak and strong SSCs at a plant the range of windspeed intervals should be broad, from about 50 mph to 300+ mph.

92

FF Calcs and PRA--Summary (Cont’d) 5. Wind hazards should be modeled separately within CAFTA to account

for different hazard dependent fragilities. Tornado hazards for wind pressure and missile fragilities should reflect the differences in point vs plant area strike probabilities.

6. An analysis of bounding errors associated with the assumption of SI across components and failure modes (typical in HW PRA using CAFTA) produced the following conclusions: SI assumption is conservative for positively correlated failure

modes. The conservatism may be significant and should be examined for the dominant cutsets.

SI is unconservative for negatively correlated failure modes. This un-conservatism is never greater than -1/e (37%).

Amount of correlation is expected to be dependent on the particular wind speed interval ( ie, quantify for each calc point).

7. Physically-based simulation models are needed to quantify the correlations.

93

Summary HW PRAs involve multiple hazards and require computational tools

to meet ASME/ANS requirments. Major needs for wind assessments of nuclear plants:

Lack of industry expertise and experience in wind effects (few wind engineers) Lack of research—no guidance documents Most companies do not have needed computational tools Lack of qualified reviewers-need guidance Need to require PE stamp on HWPRA docs (as is done in Canada) Walkdowns need to be done by experienced wind engineer

Technical/Research Issues 1. Tornado wind hazard modeling/guidance needs updating 2. Straight wind gap wind hazard analysis- downbursts 3. Need broad physical based missile tool-----(update to TORMIS) 4. Validate failure frequency as part of HWPRA 5. Need missile fragility tests

94

Summary 6. Correlated hazards Wind Driven Rain Storm Surge Wind induced fire/steam line breaks

7. Missile Screening Guidance Line of sight? Fragility for exhaust pipes, vents

8. Flow within buildings after loss of cladding 9. Multi-unit Analyses 10. Guidance on use of CAFTA 11. WBD clogging intakes 12. LOOP 13. HW HRA Guidance