ndt performance demonstration using simulation

NDT PERFORMANCE DEMONSTRATION USINGSIMULATION

N. Dominguez, F. Jenson, CEAP. Dubois, F. Foucher, EXTENDE

[email protected]

| PAGE 1CEA | 10 AVRIL 2012

NDT PERFORMANCE DEMO. USING SIMULATION

OVERVIEW

• Why using simulation for NDT performance demonstration

• NDT performance demonstration approaches

• Examples & tools for NDT performance demonstration

• Challenges

CEA | 10 AVRIL 2012 | PAGE 2


OVERVIEW




• Challenges


WHY USING SIMULATION FOR NDT PERFORMANCE


THINK NDT BEFORE IT’S TOO LATE• Prepare NDT before parts are manufactured• Possible use of digital mock-ups• Evaluate changes of material, geometry• Early feed-back to design teams


ANTICIPATE NEEDSNDT often comes late in manufacturing processes.When parts arrive it is often too late to launch R&D or investments .=> Early performance evaluation can help.


Composite parts inspectionComplex geometry inspection

Phased array solution


UNDERSTAND NDT PROCESS BEHAVIOURPropagation behaviour



REDUCE COSTS• Samples


Coupons

= $, €


OVERVIEW




• Challenges


NDT PERFORMANCE DEMONSTRATION

DEMONSTRATIONS OF PERFORMANCES

• NUCLEAR QUALIFICATION RSEM97• Comprehensive NDT

• Deterministic proof: worst case

• PROBABILITY OF DETECTION (POD)• Intensive NDT

• Statistical analysis



NUCLEAR QUALIFICATION (France RSEM97)EDF philosophy (ENIQ based)

Demonstrate the ability for a process to detect and/or characterize flaws of a given size in a given component taking into account environment.

• Comprehensive study• Includes worst case• Every defect must be detected



PROBABILITY OF DETECTION (POD)Damage tolerance design rules impose production of POD values as input for maintenance interval justification

Quantify the ability of a process, implemented in operational conditions, to detect flaws of different sizes.• Intensive NDT• Statistical analysis• Defects are detected with a given POD and given

confidence



ANALYSIS OF PARAMETERS VARIABILITY ON NDT RESULT


List influent parameters (explicitly)• nominal value• range of variation [min; max]

For each parameter, quantify influence on inspection result and demonstrate that the targeted defects are still detected.

Realize NDT according to the procedure under evaluation:Influent parameters are integrated implicitly.

Global analysis of influence of parameters by statistical estimation.

Deterministic approach Probabilistic approach


OUTPUT OF NDT PERFORMANCE DEMONSTRATION


The NDT procedure allows sure detection of flaws of size a within the specified range of application.

The NDT procedure allows detection of flaws of size a with a probability p.

0 0.5 1 1.5 2 2.50

10

20

30

40

50

60

70

80

90

100

Crack length (mm)

Pro

babi

lity

Of D

etec

tion

(%)

Estimated POD

POD 95% confidenceEmpirical POD

The probability of missing a defect isnot considered.


NUCLEAR & AERONAUTICS: WHAT IN COMMON?


Both approach• Require production of a large number of data (samples and NDT)• Include sources of variability in the NDT results (implicitly or explicitly)

Simulation can help in producing data for NDT performance demonstration


OVERVIEW




• Challenges


EXAMPLE OF PERFORMANCE DEMONSTRATION

DETERMINISTIC ANALYSISExample: UT of Thermal Sleeve of Steam Generator

Bring elements for Technical Justification in Qualification Dossier

Reduce the number of mock-ups or their complexity



DETERMINISTIC ANALYSISExample: UT of Thermal Sleeve of Steam Generator• What is the response if the defect takes different positions around the sleeve?


0°θθθθ

0°

Some positions can be studied with a mock-up,

But the variation will not be linear…

How choosing positions?



=> Make one mock-up with 2 defects at different positions and get the experimental amplitudes


Experimental data

-70,0

-60,0

-50,0

-40,0

-30,0

-20,0

-10,0

0,0

-40 -20 0 20 40 60 80 100

Defect angular position(°)

Am

plitu

de (

dB)

Experience

Then what is amplitude for other positions?

Guess? Other mock-ups?

Or…



=> Complete the data set by using simulation!


Threshold: -38 dB

Experimental data

-70,0

-60,0

-50,0

-40,0

-30,0

-20,0

-10,0

0,0

-40 -20 0 20 40 60 80 100

Defect angular position(°)

Am

plitu

de (

dB)

Simulation

Experience

θ = -20°

−29dB

θ = -10°

−21dB

θ =0°

-11dB

θ =10°

-5dB

θ = 20°

−8dB

θ = 30°

-14dB

−12dBθ = 40°

0dB

θ = 50°

−3dB

θ = 60°

−4dB

θ = 70°

-15dB

EX. & TOOLS FOR PERFORMANCE DEMONSTRATION

VARIATION SCENARIOSTools to achieve deterministic variation analysis1. List influent parameters and give a range of variation [min;max]2. Compute NDT response for each combination of parameters3. Draw resulting curves


EX. & TOOLS FOR PERFORMANCE DEMONSTRATION

POD ANALYSISApply specific variation scenario to produce POD curves


Input parameters:• Part

• Dimensions• Conductivity

• Probe• Dimensions• Number of turns• Frequency

• Inspection• Start scan position• Scan increment• Lift-off

• Flaw• Shape• Length • Height• Width

-1 -0.5 0 0.5 10

0.2

0.4

0.6

0.8

1

1.2

Conductivity variation (MS/m)

Pro

bability

den

sity

-1 -0.8 -0.6 -0. 4 -0.2 0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

Y start position (mm)

Pro

bab

ility

dens

ity

0 10 20 30 40 50 60 70 80 90 1000

0.005

0.01

0.015

Lif t-off (µm)

Pro

bability

den

sity

0.1 0.2 0.3 0. 4 0.5 0.6 0. 7 0.8 0.90

0.5

1

1.5

2

2.5

3

3.5

4

Crack height (ratio of length)

Pro

bability

den

sity

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

20

40

60

80

100

Crack length (mm)

Sig

nal r

espo

nse

(%F

SH

)

With uncertainties

Deterministic

0 0.5 1 1.5 2 2.5 3 3.50

10

20

30

40

50

60

70

80

90

100

Crack length (mm)

PO

D (%

)

Estimated POD(simulation with uncertainties)

POD 95% confidence(simulation with uncertainties)Step detection function(no uncertainty)

Explicit influent parameters description needed for POD using simulation

TOOLS FOR PERFORMANCE DEMONSTRATION

POD simulation by uncertainties propagation


-4 -3 -2 -1 0 1 2 3 40

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Probe angle (°)

Pro

ba

bili

ty d

en

sity Input parameter values are picked up

« randomly » following the probability density function

a

SY

*

*

**

Input: uncertainty Output: variability

POD EVALUATION USING SIMULATION


AUTOMATED PHASED ARRAY INSPECTION OF ELECTRON BEAM WELDING (STEEL)

Part NDT

Material: SteelElectron Beam WeldingGeometry: Locally flat areasDefects: Voids

Configuration: Phased array UT Multi-points focusing along the weld (0.2 mm pitch)0.2 mm increment on external radius

Probe: Linear array 32 elements, pitch 0.3 mm, 10 MHz central frequencyConditions:In-plant (automated rotation)


POD simulation: probabilistic variation scenario



POD Analysis



OVERVIEW




• Challenges


CHALLENGES IN PERFORMANCE DEMONSTRATION

ZONE COVERAGE FOR COMPLEX PARTS

How to ensure zone coverage of areas to be inspected, witha specified sensitivity?

POD WITH CONFIDENCE

How to provide simulated POD with confidence?



ZONE COVERAGE (SENSITIVITY)


Context:Inspection of a part with a complex geometry

Objective:For a given zone, ensure that the probe allows detection of a given defect located somewhere in the zone.

Simulation can help , providing tools to• Define the appropriate probe trajectory• Determine the appropriate delay laws associated to the computed trajectory• Evaluate the sensitivity of the inspection everywhere in the inspected zone


First step: Definition of the zone to be inspected and defect type to be detected

Computation of optimal probe positions and corresponding optimal delay laws


Second step: Determination of the appropriate probe trajectory

• The cloud formed by the optimal probe positions may be difficult to describe with a parametric trajectory (such as those used by a robot for instance)

• Necessary to simplify the problem


Third step: Computation of sensitivity maps

• Using the cloud of optimal probe positions • After simplification, using a parametric probe trajectory

Further works: • Propose GUI for the 3 step process• Sensitivity maps in 3D views, projected onto the specimen


CONFIDENCE IN POD ESTIMATION WITH SIMULATION

Difficult point : fill in the uncertain parameters description in input


?

What influence on POD result?


CONFIDENCE IN POD ESTIMATION WITH SIMULATIONOne POD scenario is:1. Define uncertain parameters2. Calculate POD curve


POD study

Scan position

Crack height

Probe angle

Min Max µ σσσσ µ σσσσ

1 -0.5 0.5 0.5L 0.13L

0 1.6

2 -0.5 0.5 0.5L 0.09L

0 0.9

3 -0.5 0.5 0.5L 0.11L

0 0.5

… … … … … … …

k -0.5 0.5 0.5L 0.17L

0 1.2

POD study

Scan position

Crack height

Probe angle

Uniform Gaussian Gaussian


1 -0.5 0.5 0.5L 0.13L

0 1.6

2 -0.5 0.5 0.5L 0.09L

0 0.9

3 -0.5 0.5 0.5L 0.11L

0 0.5

k POD curves

� Confidence approach: calculate a beam of scenarios and use the scattering on the POD curves

to estimate a confidence

POD scenario

Scan position

Crack height

Probe angle

Uniform Gaussian Gaussian


1 -0.5 0.5 0.5L 0.12L

0 1.0

Sure Not sure


CONFIDENCE IN POD ESTIMATION WITH SIMULATIONCERTAINTY INDEX DEFINITIONApplies on the probability distribution parameters


� Index 5No uncertainty on the distribution parameter

� Index 4The distribution parameter follows a gaussian law with

(µ=GV*, σ=20% GV)� Index 3

The distribution parameter follows a gaussian law with (µ=GV, σ=50% GV)

� Index 2The distribution parameter follows a gaussian law with

(µ=GV, σ=100% GV)� Index 1

The distribution parameter follows a uniform law with(min=GV-3*GV, max=GV+3GV) Not sure at all

Sure

Not so sure

Rather sure

Almost sure

* GV stands for « guessed value »


CONFIDENCE IN POD ESTIMATION WITH SIMULATIONThe 95% lower confidence POD curve is the curve which leaves

on the left 95% of the « POD curves » of the beam.


In practice the POD curve with confidence can be built pointwize, looping on POD values :

For p in [0.01;0.99]

� Find the POD curve leaving 95 % of the curves on the left at ordinate p

� Pick-up the corresponding flaw size ap

End

The POD curve with confidence is the set of points (ap, p).


CONFIDENCE IN POD ESTIMATION WITH SIMULATION


a90/95 = 1.51 mma90/95 = 0.26 mm

Manual HFET of fatigue cracks on TiAutomated Phased Array UT of EBWeld

The more the NDT process is mastered and repeatable, the better the confidence.


SUMMARY

• Simulation can help in making performance demonstrations

• Variation tools allow for deterministic performance analysis

• Probabilistic variation tools allow for POD evaluation

• Zone coverage tools allows design of NDT including a target of sensitivity




Visit our new website: www-civa.cea.fr

ndt performance demonstration using simulation

Documents