Development of Advanced Risk Assessment Methodologies for Aircraft - - PowerPoint PPT Presentation

Development of Advanced Risk Assessment Methodologies for Aircraft Structures Containing MSD/MED

• M. Liao, Y. Bombardier, G. Renaud,
• N. Bellinger, T. Cheung (DTAES/DND)

Structures and Materials Performance Laboratory Institute for Aerospace Research

2

Acknowledgements

This work was performed with financial support from the DRDC-NRC collaborative project “Quantitative Risk Assessment of CF Aircraft Structures”

Project members:

• Dr. G. Renaud, Mr. Y. Bombardier, Dr. M. Khan, Dr. G. Li, Dr. M. Liao
• Dr. A. Fahr, Mr. N. Bellinger

DND support:

• Mr. K. McRae of DRDC
• Mr. T. Cheung, Mr. Y. Caron, Mr. J. Gaerke of DTAES
• Capt. T.J. Cadeau, Sgt. M. Bunn of ATESS/DND

3

Contents

• Risk Management for CF Air Fleets
• NRC Risk Analysis Methods/Tools
• MSD Damage Tolerance Analysis

– MSD/MED crack growth analysis – MSD/MED residual strength analysis

• Risk Analysis for MSD/MED Structures

– ICSD/EIFSD – Monte Carlo MSD crack growth analyses – Maximum Stress Distribution

• Probability of Failure (PoF) Results
• Concluding Remarks
• Future work

4

Risk Management for CF Air Fleets

RARM (Record of Airworthiness Risk Management)

• Hazard Id. Risk Ass. Risk Ctrl. RARM Approval Risk Tracking
• Affecting all CF fleets (DND-AD-2007-01)

When “sufficient” data is available, Quantitative risk assessment (QRA) substantiates the assignment of a risk number in Qualitative risk assessment

TAM, C-05-005-001/AG-001, DTAES/DND, 2001 TAM, C-05-005-001/AG-001, DTAES/DND, 2001 TAM, C-05-005-001/AG-001, DTAES/DND, 2001

5

• NRC developed methods and tools to calculate the single flight

hour probability of failure (PoF, ~hazard rate) based on extensive durability and damage tolerance analysis (DaDTA) and stress-strength interference model

NRC Risk Analysis Methods

hour flight per

• n

distributi stress maximum the is ] [ where )] ( [ 1 ) ( : criterion strength residual For )]) , ( [ 1 )( ( ) ( : criterion For ) ( ) ( )]

• r

, ( [ ) ( σ σ σ σ σ σ

σ σ σ

H a H a POF dK K a H K f a POF Kc da a POF a f K a P t PoF

RS C C C C K RS C Critical Max

C

− = − = = ≥ =

• ∞

∞

• Crack size distribution update based on NDI and repair

) , ( )] ( 1 [ ) , ( ) , ( ) ( ) , (

, , ,

t a f a POD t a f da t a f a POD t a f

before a RCSD before a after a

− + ⋅ =

∞

6

NRC Risk Analysis Tools ProDTA

• ProDTA calculates the PoF using probability integration method or

Monte Carlo technique

• ProDTA is under development, aiming to become a tool for CF fleets

Maximum stress (Gumbel / others) Initial crack size distribution (ICSD/EIFS) Crack growth curve and β-solution NDI POD (Log-logistic / others) Failure criteria (KC, ac, σRS)

ProDTA

Maximum pit depth (Gumbel) Corrosion growth rate (Weibull / database) Corrosion protection breakdown time (Normal) Corrosion POD/NDI error (Normal)

PoF

Fatigue inputs Corrosion inputs

• Re. ICAF

2005 paper

7

Case Study: CC-130 Centre Wing MSD/MED Issue

The crisis C-130A catastrophic failure in Walker, CA. 2002 The causes “fatigue cracks in the lower wing skin” and “multiple site fatigue damage/ MSD” (NTSB)

The method needed

Advanced DaDTA and Risk Assessment Methodologies for Aircraft Structures Containing MSD/MED

8

CC-130 Center Wing Lower Surface Panel

CC-130 Center Wing, Lower Surface Panel, Location CFCW-1 Standard Crack (SC) scenario: single dominant crack, phase-by- phase (PBP) approach (OEM DTA)

∅ 0.339” (BBR=1.587) ∅ 0.267” 7075-T7351 0.22” thick VIII VII VI V IV III I II Phases I & II Phases III & IV Phases V & VI Phases VII & VIII Multi-phase single crack growth analysis:

SC-PBP (OEM analysis, duplication)

9

Crack Growth Analysis Scenarios

MSD scenario: MSD approach Standard Crack (SC) scenario: MSD approach

Primary crack (0.050”) Secondary cracks (0.005”)

SC-MSD MSD

Primary crack (0.050”) Secondary cracks (0.005”)

10

Good agreement between NRC closed-form equations, OEM, and FEA (StressCheck)

β β β β-Library

• Currently available & validated -functions:

c a φ Thickness (T) B

σbearing σtotal

W c

σbypass

BBR=σbearing/σbypass D B

σtotal

W c

σtotal

B D=2R 2c

Corner crack Radially crack at hole with bearing load

σtotal

Load path Plate Crack Stiffener

c

Ligament failure

Stringer/Cap effect Edge crack through hole Crack approaching a hole

11

β β β β-Library

• Additional available & validated -functions:

2a1 2a2 A B C D b Gap B

σbearing σtotal

W c1

σbypass

c2 BBR=σbearing/σbypass B1

σtotal

W c1 c2

σtotal

B2 D2 D1 W ci c2

σtotal * W/(W-Σci) σtotal * W/(W-Σci)

Diametrically cracks at hole with bearing load Crack interaction effect Linked-up crack Net section effect (under investigation)

Good agreement between NRC closed-form equations, OEM, and FEA (StressCheck)

12

Verification of MSD β β β β-Solutions

MSD

• solution from a benchmark MSD problem was verified with

FEA (StressCheck) results (ICF12 paper, Ottawa, 2009)

1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 10 20 30 40 50 a0 (mm) β β β β-solution CGCC130MSD (A11) CGCC130MSD (A12) STRESSCHECK (A11) STRESSCHECK (A12) a12 and a21 merged a22 and a31 merged 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 50 100 150 200 250 300 a0 (mm) β β β β-solution CGCC130MSD (A11) CGCC130MSD (A12) STRESSCHECK (A11) STRESSCHECK (A12) a11 merged with left edge a32 merged with a41 and a41 merged with a51

• solutions for the lead

crack a0 (< 50mm)

• solutions for the lead crack a0

(50mm<a0<300mm)

13

CC-130 Global and Local FE Modeling

Full aircraft Center wing Lower panel

(βas2) Local model (βas1)

β β β β-solution for adjacent structural effect and MED β β β βas= β β β βas1 * β β β βas2

14

Effect of Load Re-distribution (β

β β βas2)

• Methodology:

– Detach elements in global FEM

• Crack faces
• Stringers when failed

– Sum of loads across WS61

• skin, cap, stringer

Detailed FEM is needed to refine the results

a = 20 in, no stringer failure

β β β βas2

15

Effect of Cap/Stringer and Load Re-distribution

Assumption: Stringer #24 fails when the lead crack reach 12-inch; stringer #23 fails at 17-inch

9) . ( 2 1 2 1 Fig reduction Load as

• u

K s K a Tu

• a

Ts

• as
• as
• as
• as
• =

= = ∗ =

16

Crack Growth Analysis Tool

• CGA Software: NRC Crack Growth Software, CGCC130MSD

– β-library (or user defined β) – Standard crack problem (single dominant crack, phase-by-phase ) – MSD problem – Forman Equation and Retardation (Hsu model) – Monte Carlo simulation – In-service finding regression

• Spectrum: Medium usage spectrum developed by L3-Spar and used by

QETE for coupon testing of CFCW-1

17

SC vs. MSD: β β β β-Solutions

18

SC-PBP

SC MSD

SC vs. MSD: Life Prediction

~25%

Using NRC Crack Growth Software, CGCC130MSD

OEM DTA Duplicating

19

MSD/MED Residual Strength Analysis

• RS failure criteria used:

– Ultimate or yield strength (σult, σys) – Fracture toughness – Abrupt Fracture (Kcr)

Stringer #24 failed Stringer #23 failed

Residual strength (normalized to ys) curves for SC and MSD/MED scenarios

• =

a (a) K ,

• (a)

RS

• C

ys

min

20

ICSD/EIFSD Methodologies

• Approach 1 (ICSD/EIFSD): with a small sample

size (n < 40) of crack data from service, full scale test, and/or teardown

• Approach 2 (ICSD/EIFSD): with an extremely

small sample size (n<5) of crack data from service

• r full scale tests
• Approach 3 (IDS/HOLSIP): with no crack data

available from service, material and/or coupon test data can be used to determine an ICSD Affecting Factors

• DaDTA vs DTA

curve

• Lognormal vs.

Weibull

• Uncensored vs.

censored sample

• Confidence bands
• Effect of NDI

uncertainty

Ref: RTO-MP-AVT-157 (Montréal, 2008)

21

ICSD/EIFSD Approach 1

• Direct regression in-service findings to EIFS, and then find a

best-fit statistical distribution

0.00001 0.0001 0.001 0.01 0.1 1 10,000 20,000 30,000 40,000 Flight hour Crack Length (in)

In-service finding

x x x x

EIFS

For small sample (n<40) crack data from service/full scale test/teardown Regression (back calculation) methods: a) Using DaDTA/DTA curve (Master curve) b) Using the calibrated crack growth program Similar results are obtained using both methods

22

MSD/MED Monte Carlo Simulation

Monte Carlo Random EIFS generator Crack growth from EIFS Crack size (a) vs. time (t) Crack size distribution at time ti ,F(a) Probability of Failure (PoF) x N

t a t1 t2 t3 t4

START

23

EIFSD for MSD and Monte Carlo F(a)

EIFSD and MSD/MED Monte Carlo crack size distribution F(a) matched in-service findings

24

MSD/MED Monte Carlo Simulation

• Challenge: TIME!

– 3.5 min./trial x 1,000,000 trials (Laptop) > 6 years!!!

• Strategy

– Reduce number of trials to 100,000: 8 months – 10% tails results (10,000 trials only): 24 days – Parallel computing (NRC’s Linux cluster):

• 24 days / 84 CPU = 7 hours
• 24 days / 25 CPU = 23 hours

Still has room to improve!

25

N=10,000 of 100,000 runs

Matched Matched

Crack Size Distribution F(a) Tail Sampling

10% tail sampling 100% sampling

N=10,000 runs

26

• Max. Stress Distribution

(per flight hour)

Example stress exceedance curves

• Table look-up format was used

as they fit the data better than a Gumbel distribution

1E-11 1E-10 1E-09 1E-08 1E-07 1E-06 1E-05 1E-04 1E-03 1E-02 1E-01 1E+00

0.2 0.4 0.6 0.8 1 Maximum stress as a ratio of limit stress Probability of exceedances per Hour (1-CDF) Gumbel fit Table look-up data (CF2004)

• Max. stress distribution

Max Min

27

PoF Results for SC and MSD/MED

• PoF(MSD) is significantly HIGHER than PoF(SC), especially after a

certain point in the end of service life

• Maintenance actions should be adjusted according to MSD/MED

based crack growth, residual strength, and risk analyses

1E-12 1E-11 1E-10 1E-09 1E-08 1E-07 1E-06 1E-05 1E-04 1E-03 1E-02 1E-01 1E+00 10000 20000 30000 40000 50000 60000 70000 80000 EBH Single hour PoF, PoF(t) Standard crack scenario (Monte Carlo, ProDTA) MSD scenario (Monte Carlo, ProDTA)

~24%

28

Discussion: Master Curve vs. Monte Carlo

• The master curve and Monte Carlo approaches gave similar PoF. Since

the master curve approach is significantly faster than the Monte Carlo approach, further investigation is worthwhile for MSD/MED risk analysis

1E-12 1E-11 1E-10 1E-09 1E-08 1E-07 1E-06 1E-05 1E-04 1E-03 1E-02 1E-01 1E+00 10000 20000 30000 40000 50000 60000 70000 80000 EBH Single hour PoF, PoF(t) ProDTA: SC (Monte Carlo, same EIFSD) ProDTA: SC (Master curve approach) ProDTA: MSD (Master curve approach) ProDTA: MSD (Monte Carlo, same EIFSD)

29

• Developed advanced MSD/MED analysis methodologies and tools,

including beta library, crack growth analysis, residual strength analysis, and Monte Carlo simulation to support DaDTA of build-up structures of aircraft like CC-130, CP140

• Developed ICSD/EIFSD using CF in-service damage data
• Improved NRC-ProDTA software to calculate the PoF for MSD/MED

scenario, using Monte Carlo based MSD crack size distributions

• Results showed that the PoF of MSD is significantly higher than the

PoF of SC (standard crack), especially after a certain point in the end

• f service life. The maintenance actions can be adjusted according to

the MSD/MED risk analysis, crack growth, and residual strength analysis results.

Concluding Remarks

30

Future Work: Quantitative Risk Assessment

(Project funded in 2008-2011)

• Objectives:

– Continue DaDTA and PoF studies for other locations in CC-130 and CP-140 aircraft – Support the CF life cycle management

• Partners:

– Structures, NDE, DTAES/DND, IMP Aerospace …