applied mathematical modelling of ndt and iqnde · 2017. 1. 8. · – the procedure specifies;...

Applied mathematical modelling of NDT and

IqNDE

ADVANCED NDT �� MATERIALS AND MANUFACTURING TECHNOLOGY

Håkan WirdeliusAdvanced NDT/SCeNDT

SCeNDTScientific Centre of NDT

Advanced NDT, Materials and Manufacturing Technology• Håkan Wirdelius, director, theoretical mechanics, MM UT/MM DRT • Gert Persson, ass. Prof., theoretical mechanics, MM UT/FEM• Kenneth Hamberg, ass. Prof., material science, Material/NDT• Peter Hammersberg, ass. Prof., material science, POD/MM DRT• Lars Hammar, graduate student, DRT/ ET • Anders Rosell, graduate student, MM ET (VAC)• Lars Larsson, graduate student, MM ET • Erik Lindgren, graduate student, DRT

Organisation


• Erik Lindgren, graduate student, DRT• NN, graduate student, MM UT

Surface and microstructure engineering, MMT• Lars Nyborg, Prof., material science, Material

Dynamics, Applied Mechanics• Anders Boström, Prof., theoretical mechanics, MM UT• Peter Bövik, ass. Prof., theoretical mechanics, MM UT/MM ET• Per-Åke Jansson, ass. Prof., theoretical mechanics, MM UT

NDE group, Signals and Systems, UU (Ångström)• Tadeusz Stepinski, Prof., measurement engineering, ET

SCeNDT

Research area• Applied mathematical modelling of NDE methods

• Develop and experimental validation of mathematical models of different NDE methods

• Apply mathematical modelling in the development of new NDT techniques• Generate Probability of Detection curves (POD) based on simulated data

• Integrity and quality assessment by NDE (IqNDE)


• Risk Based Quality Assessment (identification of risks, classification, probability of occurrence and probability of detection)

Education• ISI Technologies (graduate and PhD)• Education for industry (graduate level)

The NDT laboratory• High resolution X-ray technique (HiReX)

• simSUNDT 1.2 delivered at a workshop 2004 (SQC)

• SCOD, a method to calculate the orientation of the dendrites in a weld based on UT data (Q. Liu, PhD 2007)

• Quantification of the reliability of flaw detection for non-destructive techniques using probability of detection (POD) based on synthetic data, (Energimyndigheten)

Applied mathematical modelling of NDT methodsH

H

∆

Ζ

SF

n̂


• PICASSO-imProved reliabIlity inspeCtion of Aeronautic structure through Simulation Supported POD,

• Analytical model of eddy current technique (T-matrix formalism) , (L. Larsson, Lic. 2012)

• Modelling of EC applied to weld geometries and properties, (A. Rosell, Lic. 2012)

• Detection and classification of defects with digital RT applied to laser welded titanium, (L. Lindgren, Lic. 2012)

• simSUNDT 1.2 -> 2.0 including immersion testing, phased array probes and anisotropic materials, (SSM)

−30−25−20−15−10−5051015202530354045505560

0

5

10

15

20

Weld: D Wave: S

Probe: −18mm Angle: 45° Frequency: 2MHz Randomness: 0%

Ray Tracing Diagram

15202530354045505560657075808590951001051100

2

4

6

8x 10

20

Position of Receiver (mm)

Reciprocity Output 1

15202530354045505560657075808590951001051100

2

4

6

8

10

Position of Receiver (mm)

Reciprocity Output 2

• Characterisation and fatigue life prediction of laser welded titanium with single pores and chain porosity, (NFFP)

• Product Uniformity Control, 2013-2018, Research Fund for Coal and Steel (RFCS)

Applied mathematical modelling of NDT methods


• Model assisted probability of detection analysis for nondestructive evaluation of complex shaped parts, 2013-2016, (NFFP)

• Volumetric reconstruction of in-service defects based on limited view x-ray tomography, 2013-2018, (VR, DEKRA)

• Validation of a system for lifetime assessment of laser welded titanium components, 2014-2015, (NFFP)

• The procedure was qualified (SQC) 1996 and UT personnel has been qualified according to it since (41 accepted between 1996 and 2004).

– Includes specification of procedures for detection, sizing and characterization.

– Incorporates both fatigue cracks and intergranular stress corrosion cracks (IGSCC).

– The procedure specifies; components, defects, level of competence (personnel),

The UT-01 procedure (manual UT)

Probability of Detection curves (POD) based on simulated data


– The procedure specifies; components, defects, level of competence (personnel), method, equipment, calibration and inspection procedure.

• Data from qualifications of personnel was used 2005 to generate corresponding POD.

– 97 implanted defects (mainly manufactured fatigue cracks).

– 14 were real stress corrosion cracks withdrawn from nuclear plants and thereafter welded into the test pieces.

• When only stress corrosion cracks (IGSCC) were considered:

– POD = Ф(0.1218+0.3720·ln(a))

The UT-01 procedure (manual UT)

• When only fatigue cracks were considered:

– POD = Ф(0.6503+0.3720·ln(a))

From the reports “Probability of Detection for the Ultrasonic Technique according to the UT-01 Procedure” (SKI 2005:03 by T. Jelinek, L. Tidström, B. Brickstad,)



MSWQC45 • conventional 45 degree shear wave transducers, • 6.35 mm in diameter (crystal), • centre frequency 2.25 MHz (44% in bandwidth).

Validation of simSUNDT

Experimental data available by WCNDT benchmark 2009, CEA



Identified essential parameter in UT-01

that are included in simSUNDTSet 1 parameters Set 2 parameters

Defect size/width (a) 2-20 mm Centre frequency 2.25 Mhz

Object thickness (t)* 35-50 mm Banwitdth 44%

Defect tilt angle (α) 0°(perpendicular)

±15° Couplant 0.4

Defect skew angle 0° ±20° Wave speed (CL) 5720 m/s

Probe angle (γ) 45° ±2.5° Wave speed (CS) 3120 m/s

Effective area (circular) Diameter 6.5 mm ±15% Density 7950 kg/m3



Effective area (circular) Diameter 6.5 mm ±15% Density 7950 kg/m

The DAC calibration in the procedure excludes thickness (t) as an essential parameter. The procedure also specifies compensation of losses caused by damping in the material which explains why this parameter is excluded

t

a

α γ

Procedure to generate synthetic based POD•A central composite designs method and a multi level full factorial sequence

-> 98 simulations were executed (simSUNDT). •A response surface was then created with multiquadratic surface fitting

-> the meta-model. •The stochastic simulation was made by using a Latin hypercube sampling (LHS) for creation of 5000 random designs for each data set, data according to:

Stochastic distributed parameters Distribution Mean Delta SD

Probability Density Function



Stochastic distributed parameters Distribution Mean Delta SD

Defect tilt angle (α) Uniform 0 15

Defect tilt angle (α) Normal 1 0 5

Defect tilt angle (α) Normal 2 0 6.67

Defect skew angle Uniform 0 20

Defect skew angle Normal 1 0 6.67

Defect skew angle Normal 2 0 10

Probe angle (γ)* Normal 1 45 0.8

Probe diameter Normal 1 6.5 0.32

σ3±

σ2±

Procedure to generate synthetic based PODThe POD then is the percentage of 5000 that is above detection level (-6dB UT-01)

70

80

90

100

Normal distr. 1 (-6dB)

Normal distr. 1 (0dB)

Probability Density Function



0

10

20

30

40

50

60

0 2 4 6 8 10 12 14 16 18 20


Normal distr. 2 (-6dB)


Uniform distr. (-6dB)

Uniform distr. (0dB)

Crack width

Comparison with POD based on qualification results (1996-2005) according to SKI 2005:3


?


Normalized signal response (simSUNDT) vs crack widthdifferent tilt angles (2.25/1.12 MHz)



• Develop an algorithm for detection and positioning of sub-millimeter pores

• The algorithm is based on a small number of rotation angles, a pore detection filter and a multiple hypothesis tracker

• Mathematical model for synthetic radiographs and the tracker

• Connection to existing fatigue models, non-destructive evaluation output to be

PhD - Detection and classification of defects with digital RT applied to laser welded titanium

Mathematical modeling of Digital RT (NFFP)

Examples of IqNDE


• Connection to existing fatigue models, non-destructive evaluation output to be predicted lifetime

PhD - Characterisation and lifetime assessment of laser welds using digital radiography

Examples of IqNDE

Mathematical modeling of Digital RT (NFFP)


PUC – Product Uniformity Control (EU-RFCS)

Mathematical modeling of UT

1 TATA STEEL NEDERLAND TECHNOLOGY, NL 2 ARCELORMITTAL MAIZIERES RESEARCH, FR

3 THYSSENKRUPP STEEL EUROPÉ, DE 4 SALZGITTER MANNESMANN FORSCHUNG, DE

5 THE UNIVERSITY OF MANCHESTER, UK 6 CENTRO DE ESTUDIOS E INVESTIGACIONES TECNICAS, ES

7 SWEREA KIMAB AB, SE 8 TNO BUILT ENVIRONMENT AND GEOSCIENCES, NL

9 CHALMERS TEKNISKA HÖGSKOLA, SE 10 SCUOLA SUPERIORE DI STUDI UNIVERSITARI E

11 the UNIVERSITY OF BIRMINGHAM, UK DI PERFEZIONAMENTO, IT

13 UNIVERSITE JOSEPH FOURIER GRENOBLE, FR 12 ARCELORMITTAL EISENHÜTTENSTADT, DE

15 CEDRAT TECHNOLOGIES, FR 14 COMMISSARIAT A L'ENERGIE ATOMIQUE CEA, FR

Examples of IqNDE


• Optimization of the product uniformity by using non-destructively measure electromagnetic and ultrasonic parameters that relate to the steel microstructure.

• Detail the relation between microstructural uniformity and online measured (non-) uniformity.

• Developments of methods for non-destructive texture evaluation, the challenging application of microstructure characterising sensor systems in the continuous annealing section and the use of operational online systems for sample selection.

15 CEDRAT TECHNOLOGIES, FR 14 COMMISSARIAT A L'ENERGIE ATOMIQUE CEA, FR

PUC – Product Uniformity Control (EU-RFCS)

Mathematical modeling of UT

Examples of IqNDE

2.25 MHz 20 mm


2.25 MHz 100 mm 5 MHz 100 mm

Evaluate the ability of eleven different NDE techniques to detect and characterize PWSCC in nickel alloy components and welds

PARENT - Program to Assess the Reliability of Emerging Non-destructive Techniques


P28 MN 220 AD U6 spline 8 bit 512 inv.avi

applied mathematical modelling of ndt and iqnde · 2017. 1. 8. · – the procedure specifies;...

Documents