See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/285929282

Modeling and Simulation for Nondestructive Testing with Applications to

Aerospace Structures

Article  in  Materials Evaluation · January 2008

CITATIONS READS

26 1,032

2 authors, including:

John C. Aldrin
Computational Tools
183 PUBLICATIONS   1,058 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Forward Models for Extending the Nondestructive Evaluation Capability of Resonant Ultrasound Spectroscopy View project

Imaging and characterization of fatigue cracks around fastener holes View project

All content following this page was uploaded by John C. Aldrin on 28 November 2017.

The user has requested enhancement of the downloaded file.

 Modeling and Simulation for Nondestructive Testing
with Applications to Aerospace Structures
John C. Aldrin* and Jeremy S. Knopp†
*
Computational Tools,
4275 Chatham Ave., Gurnee, IL 60031, USA
Phone: (847) 599-1213
†
Air Force Research Laboratory (AFRL/MLLP),
Wright-Patterson AFB, OH 45433, USA

ABSTRACT

Modeling and simulation provide the NDT practitioner with the capability to better understand the

physics of complex inspection problems necessary for improved NDT technique development. The

state-of-the-art on NDT modeling and simulation is presented with emphasis on ultrasonic and eddy

current testing. The practical advantageous and challenges for various numerical methods such as the

finite element method are also discussed. To highlight the potential value of models in NDT, two case

studies are presented for aerospace structural applications. The finite element method is used to provide

insight into the complex interaction and scattering of ultrasonic waves around fastener sites with radial

cracks in three dimensions. A demonstration is also presented of the potential for numerical simulations

to assist in the evaluation of the probability of detection (POD) for an eddy current inspection technique.

The model-assisted POD results were found to be in good agreement with experimental POD results.

Lastly, future research and development of modeling tools for wider use within the NDT community is

proposed.

KEYWORDS

aerospace structures, eddy current, finite element method, nondestructive evaluation, ultrasonic

testing.

1
INTRODUCTION

Although empirical approaches have been successful in the past, existing trends have made the

development of new inspection techniques more difficult. Such trends include an increase in the use of

advanced materials, aging aircraft and infrastructure requiring the testing of components not designed

for inspection, the need for damage characterization, the goal to reduce inspector variability through

automation, and greater cost scrutiny. Given this environment, improving existing nondestructive

techniques and developing new techniques requires a better understanding of the physics of NDE

measurement techniques.

An NDT measurement model functions to predict the measurement response of a material under test

for an NDT technique using the appropriate physics. NDT simulation is defined as the operation of the

NDT model across the measurement domain of interest, for example in the time, frequency and/or

spatial domains. The components of an NDT technique model are shown in Figure 1 for the case of an

ultrasonic technique. A complete system model encompasses three major components: a source, a

sample, and a receiver. The source component of the model defines the incident field in the sample

through representation of the input signal, source hardware, electrical connection, source transducer, and

the transducer interface condition with the sample. Given the incident field, the scattered field can be

calculated by the sample component model, given the material properties, sample geometry and

discontinuity (often cracks or corrosion) characteristics. Depending on the measurement technique,

significant material properties can include elastic properties, electrical conductivity, thermal

conductivity and density. Sample geometry including domain size (finite or infinite) guides the

selection of the appropriate model. Discontinuity characteristics include type (cracks, voids, porosity,

corrosion, disbonds), geometry, and condition (such as the interface condition between crack faces).

The reception component transforms the scattered field into measurement data through a relationship

2
defined by the transducer interface condition, receiver transducer, electrical connections and data

acquisition hardware. The model for each component is typically designed in order to provide the most

accurate representation of a measurement technique while minimizing computational effort.

Models can produce significant benefits at several stages of the NDT technique development process

(Figure 2). First, models can be used to aid in the interpretation of raw measurement data. With this

understanding, modeling can be beneficial in selecting the appropriate features of the measurement for

evaluation and classification. The inspection design including for example, ultrasonic transducer

characteristics and eddy current coil parameters can greatly benefit from parametric studies using

accurate measurement models. For the development of automated inspection techniques, models can

expand the training data set, thus reducing sample costs. Models can also be directly incorporated into

automated signal classifiers through inverse methods. During the validation process, models can be

used to verify the robustness of the classification technique while again reducing sample costs. Model-

assisted probability of detection (MAPOD) procedures have the potential to evaluate the reliability of an

NDT technique at a lower cost with respect to conventional POD studies. Lastly, concerning the

transition of a new technique to a practical application, models can be quite beneficial in displaying a

fundamental understanding of the inspection problem to project sponsors and helpful during the

instruction of the inspectors.

In this paper, the state-of-the-art on NDT modeling and simulation is presented with emphasis on

numerical models for ultrasonic testing and eddy current testing. The practical advantages and

challenges for various numerical methods such as the finite element method are also discussed. To

highlight the potential value of models in NDT design, two case studies are presented for aerospace

structural applications. The finite element method was used to evaluate the complex interaction and

scattering of ultrasonic waves around fastener sites with radial cracks in three dimensions. Lastly, a

3
demonstration is presented of the potential for numerical simulations to assist in the evaluation of the

probability of detection (POD) for an eddy current inspection technique.

OVERVIEW OF MODELING AND SIMULATION IN NDT

A critical initial step in NDT modeling concerns the determination and evaluation of the critical

characteristics of the inspection problem that must be accurately represented in the model. This

evaluation will determine how each model component can be practically solved, by either analytical,

asymptotic, numerical or empirical methods. Analytical approaches evaluate the exact solution to

fundamental problems based on first-principles. Asymptotic methods formulate expressions that

approximate the solution to problems where exact solutions are unavailable. Numerical modeling

approaches are applied to complex problems to transform the model space to a series of coupled

fundamental problems that can be efficiently solved by a computer. Numerical methods like the finite

element method are particularly powerful since they can more easily address NDT simulations for

complex part geometries and unique material constitutive relationships. This overview of modeling for

ultrasonic and eddy current techniques also highlights simulation tools that provide a complete NDT

measurement model.

Ultrasonic Testing

The basis for ultrasonic NDT modeling is the elastodynamic equations of motion derived from the

theory of elasticity,

(λ + µ )∇∇ ⋅ u(x, t ) + µ∇ 2 u(x, t ) = ρu(x, t ) . (1)

where u is the displacement vector, λ and µ are the Lame constants, and ρ is density. Although this

equation can be reduced to two wave equations for longitudinal and shear waves, the challenge in

4
modeling ultrasonic wave propagation originates from their coupling at the boundaries. For example,

the interaction of ultrasonic waves between parallel surfaces and at corners will produce complex guided

wave modes and diffracted waves respectively. Research into ultrasonic NDT measurement models has

concentrated on two fundamental components, 1) the ultrasound source (i.e. the transducer model), and

2) the propagation and scattering of waves in a specimen for a variety of material properties, sample

geometries and discontinuity characteristics. Excellent introductions to modeling in ultrasonic

nondestructive evaluation can be found in the following works: Achenbach (1973, 1992), Thompson

and Thompson (1985), Gray et al. (1989), Schmerr (1998, 1999), and Spies (1999).

A straightforward means of representing elastic wave propagation is through ray theory. Thus, the

direction of propagation of disturbances corresponds to the direction along rays. The theory for

modeling waves in elastic solids as rays can be found in the text by Achenbach et al. (1982). Many

software packages have been implemented based upon ray theory, the geometric of diffraction and

paraxial beam models (Schmerr, 1998). The Kirchhoff approximation and asymptotic approaches such

as geometric theory of diffraction have been found to be useful models for many crack inspection

problems. Several NDT simulation packages based on these approaches include: CIVA (Champ-Son,

Méphisto) by CEA (Lhémery, 1999), UTSIM by CNDE of Iowa State University (Schmerr, 1999),

CADMUS by Fraunhofer-Institute for Nondestructive Testing (Spies, 1996), SUNDT (UTDefect) by

Dept. of Mechanics - University of Chalmers, Continuum Ultrasonic Modeler by Acoustic Ideas Inc.,

and Imagine3D by UTEX Scientific Instruments Inc.

A variety of numerical methods have been developed and applied to address more complex

ultrasonic inspection problems including the scattering of surface and Lamb waves. The finite-

difference time-domain (FDTD) method solves the elastodynamic differential equation of motion by

approximating the derivatives with finite differences. The FDTD formulation is quite straightforward in

5
implementation but can have difficulty in accurately modeling irregular geometries and complex

boundary conditions. The local interaction simulation approach (LISA) is similar to FDTD but allows

for implementation of nonlinear contact models and representing heterogeneous media (Delsanto, P. P.

and Scalerandi, 1996). The finite element method (FEM) has been a workhorse for ultrasonic NDT

modeling over the years due to its ability handle problems with awkward geometries, inhomogeneous

media and nonlinear materials. A few examples of the numerical formulation of FEM with applications

to ultrasonic testing can be found in Lord et al (1988) and Wojcik et al (1993). For certain problems

with localized boundaries or domains that extend to infinity, formulations such as the boundary element

method (BEM) or boundary integral equation method (Niwa et al, 1986, Aldrin, 2001) and

elastodynamic finite integration technique (EFIT) (Schuhmacher et al., 1994) can provide significant

computational advantages. Hybrid approaches combining multiple solution methods for reduced

computational time and memory requirements have also been developed for ultrasonic NDT application

(Liu and Datta, 1993; Chang and Mal, 1999). Hybrid methods often employ numerical methods like

FEM for the solution of the local scattering problem while applying semi-analytical solutions to obtain

the far-field response. Packages that both employ numerical methods and have been designed to

simulate NDT measurements include: Wave2000/Wave3000 by CyberLogic using the FDTD (Kaufman

et al., 1999), and PZFlex by Weidlinger Associates Inc. using FEM (Wojcik, 1993).

6
Eddy Current Testing

A variety of approaches have been developed over the years to model eddy current NDT

measurements. Fundamentally, eddy current NDT can be modeled by a diffusion partial differential

equation in terms of the magnetic vector potential derived from Maxwell’s equations,

∂A
∇2 A = µJ 0 + µσ + µ∇(1 / µ ) × (∇ × A ) , (2)
∂t

where A is the magnetic vector potential, J0 is the applied current density, σ is the electrical

conductivity and µ is the magnetic permeability. This equation can be solved analytically for canonical

geometries (Dodd and Deeds, 1968; Theodoulidis and Kriezis, 2006) or numerically for complex

domains and geometries containing cracks. Once the magnetic vector potential is determined, many

electromagnetic quantities such as coil impedance, magnetic fields, and current densities can easily be

calculated. Introductions to modeling of eddy current measurements can be found in works by Gray et

al. (1989), Auld and Moulder (1999).

To practically address the solution for a variety of complex probe designs and test sample

configurations, several numerical methods have been developed to solve the field equations. Finite-

difference time-domain (FDTD) method solves the differential equation over a domain by

approximating the derivatives with a finite difference equivalent (Dodd and Deeds, 1967). This finite

difference relation is then applied at a grid of nodes that result in a system of algebraic equations to

solve. The primary advantage of the FDTD method is its simple efficient formulation. The numerical

formulation of FEM for eddy current NDT is well established in the literature (Palanisamy and Lord,

1979; Ida et al., 1983, 1985). The governing equations for each element are derived using either

Galerkin’s method or a variation approach. The resulting system of equations for the meshed domain is

then solved most often in terms of magnetic vector potential. The measurement model for an eddy

current coil can then be expressed in terms of changes in resistance and inductance associated with

7
dissipated energy in the region of the conductor and the stored energy in the whole solution domain

respectively. Advantages of the FEM model include the fact that the solution is valid and available in

the entire domain, and the ability to address problems with awkward geometries, inhomogeneous media

and nonlinear materials.

For certain problems such as domains extending to infinity, formulations such as the boundary

element method (BEM) and volume integral equation (VIE) method can be more efficient than FDTD

and FEM. The boundary element method (BEM) or boundary integral equation method solves an

integral form of Maxwell’s equations where only the boundary of the solution domain must be solved

through application of equivalent surface sources (Beissner, 1986). The volume integral equation (VIE)

or volume integral method (VIM) models anomalous regions as a fictitious current source, which is

directly related to departures in electrical conductivity and magnetic permeability in the host material .

Green’s functions for infinitesimal current sources in multi-layered materials are applied and the method

of moments (MoM) is used to solve the volume integral equation. The advantage of this formulation is

that only the region of the scatterer needs to be discretized. Because of the regular grid, the resulting

matrix has an advantageous structure, being Toeplitz (convolutional) and Hankel (correlational), where

very large problems can be solved efficiently using the conjugate-gradient method (Bowler et al., 1989,

Murphy et al., 1997). Note, both BEM and VIE can be difficult to apply to problems with large

complex structures.

Several software packages have been developed for simulation of eddy current measurements. They

include OPERA-3D by Vector Fields using FEM, ECSIM by CNDE at Iowa State University using

BEM, and VIC-3D by Victor Technologies and CIVA (MESSINE) by CEA using the MoM/VIE.

Lastly, many numerical methods such as FEM require very fine meshes for an accurate solution to

problems with complex geometries that result in very large systems of equations to be solved. Research

8
has investigated meshless and hybrid methods to better address such problems. For example, the

element-free Galerkin model can be used to represent both the geometry and basis over a simple grid of

points. This meshless approach eliminates the need for fine meshing of complex geometries, improving

the speed and accuracy of the model solution (Xuan, 2004).

APPLICATIONS FOR AEROSPACE STRUCTURES

Due to the potential for catastrophic failure, fatigue cracks in aircraft structures are of significant

concern. Given inherent stress concentrations, fastener holes in aircraft structures are primary locations

for crack initiation. The early detection of cracks at maintenance intervals is necessary to maintain

safety and minimize repair cost. The bolt hole eddy current technique has the capability to detect small

cracks, but requires the removal of the fastener. Thus, the development of robust NDT techniques that

detect subsurface cracks and do not require the removal of the fastener are of particular interest. For this

inspection problem, numerical simulations are presented here to 1) explore the unique challenges of

detecting sub-surface cracks around fastener sites in multilayer structures using ultrasonic techniques

and 2) demonstrate the potential for model-assisted POD evaluation for validation of an eddy current

testing application.

Spiral Creeping Waves in the Ultrasonic Inspection of Fastener Sites

Ultrasonic angle-beam shear wave inspections have been developed to address crack detection in

multilayer structures without fastener removal. Early work to develop a wide-area automated ultrasonic

technique for corrosion detection in the DC-9 wing box laid the foundation for future applications using

angle-beam shear waves. The first ultrasonic inspection procedure for the detection of cracks around

fasteners in multilayer structures was developed for the C-141 spanwise splice joint incorporating

automated scanning and imaging. This technique was validated through a probability of detection

9
(POD) study and implemented for regular depot maintenance (Mullins and MacInnis, 1997). However

for this case and other recently developed applications (Lindgren et al, 2006), full 360° coverage around

the hole was not required given a priori knowledge of the expected crack location due to deterministic

loading conditions. For inspection problems requiring full 360° coverage, several approaches have been

investigated including mechanical rotary scanners and 360° phased array transducers, but each require

careful centering around each fastener site. A linear phased array transducer with a single-axis linear

scanner has been used to improve the scan speed and coverage versus 2D raster scans of single element

transducer (Smith et al, 2005); however, the proposed approach does require six separate scans to ensure

full 360° coverage.

To supplement traditional methods for crack detection based on the corner reflection signal and

diffraction at the crack tip and/or root, the concept of generating and detecting ‘spiral’ creeping waves at

a fastener site is proposed. A diagram of this concept is presented in Figure 3. The use of creeping

waves for the inspection of cracks around holes in vertical riser structures has been explored for empty

(weep) holes (Nagy et al, 1994; Aldrin et al, 2001), holes with coatings or filled with fluid (Aldrin and

Achenbach, 2003), and holes containing fastener with sealant (Aldrin et al, 2004). In Figure 3, a spiral

creeping wave is generated by an incident shear wave on a curved hole surface, which propagates in a

spiral direction to cracks located in the shadow region. Due to the curvature of the hole, any reflected

spiral creeping wave from a crack will leak away from the hole and can be measured in either pulse-

echo or pitch-catch modes. Recently, Michaels et al (2006) demonstrated the potential for in situ sizing

of cracks around fastener using an ultrasonic angle-beam method generates and detects creeping waves

around the hole. Potential benefits of the approach include greater sensitivity to cracks of varying

angular location and improved sensitivity to skewed cracks not perpendicular to the hole. Also, this

10
approach is expected to reduce inspection time, given the potential to eliminate scans with varying head

position (for linear scans) and single hole centering (for circumferential scans).

Although analytical and 2D BEM models have provided some insight into the nature of spiral

creeping waves as they propagate around a cylindrical hole (Aldrin et al, 2006), 3D models are essential

to fully study the physics of this complex ultrasonic scattering problem when fatigue cracks are present.

A 3D explicit FEM model is used here to solve the localized scattering from a crack located in the

shadow region of a fastener site with respect to the transducer location. This approach provides an

accurate measurement model that best addresses the complexity, speed, and memory requirements for

the 3D problem. The FEM package FEAP was used to solve the 3D explicit finite element method

problem (Taylor, 2007). Perfectly matched layers were implemented as absorbing boundary conditions

to help reduce the size of the meshed domain and thus minimize solution time. Model parameters for

the simulation are as follows. A 4.0 mm (0.16") by 4.0 mm square element transducer was modeled at

the surface with a shear wave angle of incidence of 45º and a center frequency of the ultrasonic pulse of

3.3 MHz. The material properties of aluminum were used for the part. The diameter of the hole was 5.0

mm (0.20") and the total thickness of the plate was 4.0 mm (0.16"). The corner crack was modeled as a

finite width notch with a square crack face of 2.0 mm (0.079") by 2.0 mm located at the bottom surface.

The plate was truncated around the fastener site to a square region of 20 mm (0.79") by 20 mm in the x

and y-directions centered around the hole. In addition, 1.0 mm (0.039") thick perfectly matched layers

were implemented around the truncated plate and free boundary conditions were applied on the top and

bottom surfaces of the plate. A uniform mesh, which is typically used for explicit FEM simulations, was

not used here because the representation of the smooth curved boundary was deemed critical for the

spiral creeping wave model to avoid spurious diffractions that would be present with a truncated

uniform boundary. Instead, the domain was modeled using eight sections extending from the hole

11
surface to the truncated (square) edge with 9 node quadrilateral elements of varying dimension

providing a smooth surface of elements at the hole boundary. Future work is planned to fully explore

the differences between a uniform mesh truncated at the hole versus an irregular mesh that is smooth

mesh at the hole boundary on the propagation of creeping waves. The number of elements was set to

maintain at least 6 nodes per wavelength. For this problem, the shortest wavelength in the sample is

associated with the creeping wave that propagates around the hole (Aldrin et al, 2006) which is slower

than the Rayleigh wave speed in aluminum by a factor of approximately 0.92. Lastly, the maximum

time step was set according to the Courant stability criteria using the minimum grid spacing around the

hole and the longitudinal wave speed.

Figure 4 presents the transient finite element response of an ultrasonic wave scattering around a

fastener site with a radial notch (representing a crack) in 3D. Presenting transient 3D responses with 2D

representations is inherently difficult. To provide an adequate perspective, the root mean square (RMS)

displacement response (for the three directions) is presented from several views and at three different

times-of-flight. The left column displays the RMS displacement response on the bottom surface, and the

center and right columns present the RMS displacement response around the hole surface. Three select

times-of-flight are also displayed: (a) the main shear wave incident at the hole surface, (b) the

propagating spiral creeping wave reflecting from the hole bottom edge and (c) the reflected spiral

creeping wave propagating upward from the far surface. The notch is located at the bottom surface and

positioned in the shadow region of the hole with respect to the transducer. This is the most severe

location for inspection, located 90º around the hole from the normal location for angled-beam shear

wave crack inspection. In Figure 4(a), the incident shear wave on the hole surface generates a surface

wave that both propagates around and leaks away from the hole. Although not clearly observed in

Figure 4(a) due to the saturation of the response, figures 4(b) and 4(c) more clearly present the reflection

12
and spiral propagation of the creeping wave with respect to time. In addition, the leaking of the wave is

observed in Figure 4(c) through both the decay in magnitude of the spiral wave (in the hole surface

view) and the leaky wave propagating away from the hole (in the bottom surface view). Another

interesting observation is the existence of a corner creeping wave that propagates around the bottom

hole corner and scatters from the crack. Clearly, the position of the transducer and selection of the angle

of incidence was not ideal for achieving a significant scattering of the spiral creeping wave by the crack

that would return to the transducer. Given that simulation time was on the order of several days for the

high frequency ultrasonic problem, only a few simulations have been run to date. Prior work has

demonstrated that these cracks are observable at locations with the angular crack shifted as much as 40°

from the normal position, where corner and diffraction effects are not the source for the measured

signals since the crack root and tip are located in the shadow region of the hole (Aldrin et al, 2006). As

observed here, beyond 40° from the normal position, signals are present but very difficult to measure

with respect to background measurement noise. Future work is planned to present more extensive

experimental studies and model comparisons for optimal incident angle, varying transducer location and

ideal transducer focusing.

Model Assisted POD Demonstration for Eddy Current Testing

A second case study is presented on the application of numerical simulations to assist in the

validation of an eddy current technique for the detection of cracks around fastener holes in multilayer

structures. To validate the performance of new inspection techniques, probability of detection (POD)

studies can be performed. The preparation of POD samples with real fatigue cracks and the process to

acquire a statistically significant number of measurements is often very time consuming and expensive,

providing a significant burden for the validation of new inspection techniques. A model-assisted

13
strategy for the design and execution of POD studies has been proposed to help mitigate the validation

costs and to improve POD evaluation quality by addressing a wider array of inspection variables (Gray

et al, 1989; Thompson, 2001). Using computer and empirical models to address variables that cannot be

easily recreated in experimental samples is a significant opportunity. Two methodologies have been

proposed to perform model-assisted POD evaluations: the transfer function (XFN) approach and the

full-model assisted (FMA) approach. The transfer function approach transforms an existing POD model

for one inspection to address another similar inspection based on a limited number of new experimental

measurements studying a specific varying parameter. This approach has been successfully demonstrated

to transform POD results for EDM notch measurements in engine components (Smith et al, 2004) and

aircraft structures (Harding et al, 2007) to parts with real cracking conditions. The full-model assisted

approach uses advanced computer simulations to model the inspection process and determines the POD

for the inspection technique through a combination of experimental and simulated data. This study

investigates a FMA POD methodology incorporating computer simulation for the inspection of cracks

around fastener sites in a two layer aircraft structure inspection performed with an eddy current

technique.

For any POD study, all critical factors for the NDT technique, part material, part geometry, and

discontinuity characteristics that control signal and noise must be fully considered through expert

knowledge, experimental results and/or model-based studies. For example, the probe characteristic

response, lift-off, and scan resolution were identified as important factors concerning the NDT

measurement. The surface condition of the samples, thickness of the layers, type of fastener, fastener

fit, and distance between adjacent fasteners and edges were identified as significant factors related to

part geometry and material. Lastly, the dimensions, aspect ratio, location around fastener site, and

14
morphology of the crack were identified as important characteristics to be considered. Figure 5 shows a

diagram of the fastener crack inspection problem for both (a) corner and (b) through crack problems.

Evaluation of the quality of the NDT simulation is another critical step in applying the model-

assisted POD approach. Two numerical models were initially evaluated for the FMA study, the finite

element method implemented in Opera3D® (Carpenter, 2000) and the volume integral method (VIM)

implemented in VIC-3D® (Murphy et al, 1997). Additional details on experimental and simulated

comparison studies are presented in Knopp et al (2006). Comparisons between the experimental data

and simulated data for the two numerical methods for the fastener crack problem are presented in Figure

6. Good agreement between the experimental data and the two models was achieved in terms of the

magnitude, phase and general shape of the curves in the impedance plane. Given uncertainty in probe

parameters and noise in experimental measurements, these remains deviations are quite reasonable. For

the FMA study, the volume integral method (VIC-3D®) was chosen since (1) meshing was only

required for the hole and crack region resulting in significantly shorter solution times and (2) meshes are

regular in all three dimensions providing less numerical error when probes are scanned across a sample.

Simulated parametric studies were performed for varying crack size and type, considering both

through and corner crack geometries. Experimental measurements were also incorporated in the

MAPOD study to address noise in the eddy current measurement. Details on the sample characteristics

and eddy current measurements are as follows. The dimensions for the top and bottom layers measured

3.96 mm (0.156") and 2.54 mm (0.100") respectively. Conductivities of 1.87 x107 S/m for the

aluminum layers and 1.79 x106 S/m for the titanium fasteners were used in the model. The size of the

hole was set to a radius of 4.04 mm (0.159") respectively. The probe frequency was set to 600 Hz and

the permeability of 1000 was used for the ferrite cup core. The coil dimensions had a height of 6.0 mm

(0.236"), an inner radius of 3.0 mm (0. 118"), and an outer radius of 6.0 mm (0. 236"). Two crack

15
conditions were simulated: a through notch and a corner notch with the aspect ratio length to width to be

1:1. For this simulated study, all cracks were treated as notches of a finite width of 0.25 mm (0.010"),

so it was assumed that there is no electrical contact between crack faces. Crack lengths in the

experimental samples were available between 0.000" to 0.169" while notch lengths in the simulations

were run between 0.000" to 0.200". Additional information on the experimental and simulation studies

can be found in Knopp et al (2007). After the simulations were conducted, the feature extraction

technique described in Aldrin et al (2007) was used to extract the crack signal response.

Figure 7 presents a comparison of the experimental and simulated data for varying crack length and

crack type. In general, there is good agreement between the two data sets with some error for select

experimental data points. In particular, there are a few weaker than expected responses for some of the

larger cracks that are likely lower due to differences in the crack aspect ratio or in some cases close

proximity to adjacent steel fasteners in the test samples. A Monte Carlo simulation was then performed

using the noise measurement distribution for the fastener sites with no cracks, the corner crack model,

and the through crack model, to populate the full-model assisted data sets for POD evaluation. The

Monte-Carlo study was performed using 5000 data points to generate the model-assisted POD data set

for statistical evaluation. Finally, the POD curves were fit for the experimental and model-assisted data

sets using the conventional maximum likelihood approach for hit-miss data. The confidence bounds

were determined using the loglikelihood ratio method. Figure 8 presents a comparison of the POD

results for the experimental and full-model assisted approaches. The curve generated with the MAPOD

approach is within the 95% confidence bounds of the empirical curve, demonstrating good agreement

between the two approaches. These results are very encouraging for the greater application of numerical

models in NDT technique validation.

16
CONCLUSIONS AND RECOMMENDATIONS

Models for nondestructive testing have reached a level of sophistication such that the study of a wide

variety of inspection problems can be addressed computationally. Two examples were introduced

highlighting the benefits of simulation in the design and validation of ultrasonic and eddy current

techniques respectively. However, it must be noted that the wide acceptance and everyday use of

modeling as a practical tool in NDT has not occurred throughout the industry. The following issues still

exist: 1) modeling tools for unique NDT applications require excessive development cost and time, 2)

accurate numerical simulations are often computationally intensive in terms of excessive time and

computing power, and 3) models lack consistent accuracy with experimental results due to known model

approximations and/or unknown experimental factors. To properly address these issues, a strategic

collaboration between the NDT leadership / funding organizations, the NDT research community, and

the NDT applications community over the long-term is needed to both address the key challenge

problems of the application community and transition simulation tools that can be used by NDT

practitioners. Simulation capability must also improve to better address advanced array transducers, in

situ sensing, scattering from complex discontinuities, composite materials, and emerging methods like

terahertz NDT. Lastly, the goal to move beyond detection and achieve quantitative discontinuity

characterization through the application of models in robust inverse method schemes should continue to

be pursued.

ACKNOWLEDGEMENTS

Original funding for work on computational methods in NDE has been provided by the Air Force

Office of Scientific Research and the Air Force Research Laboratory – NDE Branch, including support

and valuable discussions with Eric Lindgren, Kumar Jata, Gary Steffes, and Charlie Buynak. Recent

17
support and encouragement was provided by NASA including Don Roth, Bill Winfree, Bill Prosser and

Phillip Williams. The authors also thank Claudia Kropas-Hughes, Larry Dukate of AFRL and Dick

Martin of UDRI for prior acquisition of the eddy current experimental data.

REFERENCES

Achenbach, J.D., Wave Propagation in Elastic Solids, North-Holland, 1973.

Achenbach, J.D., A.K. Gautesen, and H. McKaken, Ray Methods for Waves in Elastic Solids, Pitman,
London, 1982.
Achenbach, J.D., “Measurement Models for Quantitative Ultrasonics,” Journal of Sound and Vibration,
Vol. 159, No. 3, 1992, pp. 385-401.
Aldrin, J.C., J.D. Achenbach, G. Andrew, C. P’an, B. Grills, R.T. Mullis, F.W. Spencer and M. Golis,
“Case Study for the Implementation of an Automated Ultrasonic Technique to Detect Fatigue Cracks
in Aircraft Weep Holes,” Materials Evaluation, Vol. 59, n 11, 2001, pp. 1313-1319.
Aldrin, J.C. and J.D. Achenbach, “Models for Crack Detection in a Cylindrical Hole Containing an
Elastic Layer and a Fluid-Filled Cylindrical Hole,” Rev Prog Quant Nondestr Eval, Vol. 22, 2003,
pp. 181-189.
Aldrin, J.C., J.R. Mandeville and C.V. Kropas-Hughes, “Ultrasonic Detection of Cracks in a Complex
Aircraft Structure Using a Local Correlation Method for Signals from a Moving Transducer,” Rev
Prog Quant Nondestr Eval, Vol. 23, 2004, pp. 565-572.
Aldrin, J.C., J.S. Knopp, D. Judd, J.R. Mandeville and E. Lindgren, “Spiral Creeping Waves in
Ultrasonic Angled-Beam Shear Wave Inspection of Fastener Holes In Multilayer Structures,” Rev.
Prog. Quant. Nondestr. Eval, Vol. 25, 2006, pp. 187-194.
Aldrin, J.C. and J.S. Knopp, “Case Study for New Feature Extraction Algorithms, Automated Data
Classification, and Model-Assisted Probability of Detection Evaluation,” Rev. Prog. Quant.
Nondestr. Eval., Vol. 26, 2007, pp. 257-264.
Auld, B.A. and J.C. Moulder, "Review of Advances in Quantitative Eddy Current Nondestructive
Evaluation," Journal of Nondestructive Evaluation, Vol. 18, 1999, pp. 3-36.
Chang, Z. and A. Mal, “Scattering of Lamb Waves from a Rivet Hole with Edge Cracks,” Mech Mater,
Vol. 31, No. 3, 1999, pp. 197-204.

18
Carpenter, D.C., “Use of the Finite Element Method in Simulation and Visualization of Electromagnetic
Nondestructive Testing Applications,” Materials Evaluation, Vol. 58, No. 7, 2000, pp. 877-881.
Beissner, R.E., “Boundary Element Model of Eddy Current Flaw Detection in Three Dimensions,” J.
Appl. Phys., Vol. 60, 1986, pp. 352-360.
Bowler, J.R., L.D. Sabbagh and H.A. Sabbagh, “A Theoretical and Computational Model of Eddy-
Current Probes Incorporating Volume Integral and Conjugate Gradient Methods,” IEEE Trans
Magn, Vol. 25, No. 3, 1989, pp. 2650 -2664.
Delsanto, P.P. and M. Scalerandi, "A Spring Model for the Simulation of the Propagation of Ultrasonic
Pulses through Imperfect Contact Interfaces," Journal of the Acoustical Society of America, Vol.
104, No. 5, 1998, pp. 2584-2591.
Dodd, C. V. and W.E. Deeds, “Electromagnetic Forces in Conductors,” Journal of Applied Physics, Vol.
38, No. 13, 1967, pp. 5045-5051.
Dodd, C. V. and W.E. Deeds, “Analytical Solutions to Eddy-Current Probe-Coil Problems,” Journal of
Applied Physics, Vol. 39, No. 6, 1968, pp. 2829-2838.
Gray, J.N., T.A. Gray, N. Nakagawa and R.B. Thompson, “Models for Predicting NDE Reliability,”
Nondestructive Evaluation and Quality Control, Metals Handbook 17, ASM International, Ohio,
1989, pp 702-715.
Harding, C., G. Hugo and S. Bowles, “Model-assisted Probability of Detection Validation of Automated
Ultrasonic Scanning for Crack Detection at Fastener Holes,” Proceedings of the 10th Joint
FAA/DoD/NASA Conference on Aging Aircraft, Palm Springs, CA, April 16-19, 2007.
Ida, N., R. Palanisamy and W. Lord, “Eddy Current Probe Design Using Finite Element Analysis,”
Materials Evaluation, Vol. 41, No. 12, 1983, pp. 1389-1394.
Ida, N. and W. Lord, “Finite Element Model for Three-dimensional Eddy Current NDT Phenomena,”
IEEE Trans Magn, Vol. 21, No. 6, 1985, pp. 2635-2643.
Kaufman, J.J., et al, “Computational Method for NDT,” Proc. SPIE, Vol. 3585, 1999, p. 173-181.
Knopp, J.S., J.C. Aldrin and P. Misra, “Considerations in the Validation and Application of Models for
Eddy Current Inspection of Cracks around Fastener Holes,” Journal of Nondestructive Evaluation,
Vol. 25, No. 3, 2006, pp. 123-138.
Knopp, J.S., J.C. Aldrin, E. Lindgren and C. Annis, “Investigation of a Model-Assisted Approach to
Probability of Detection Evaluation,” Rev Prog Quant Nondestr Eval, Vol. 26, 2007, pp. 1775-1782.

19
Lhémery, A., “Multiple-Technique NDT Simulations of Realistic Configurations at the French Atomic
Energy Commission (CEA),” Rev Prog Quant Nondestr Eval, Vol. 18, 1999, pp. 671-678.
Lindgren, E., et al., “Validation and Deployment of Automated Ultrasonic Inspections for the C-130
Center Wing,” ASIP Conference, Savannah, Georgia, December 2 - 4, 2003.
Liu, S.W. and S.K. Datta, “Scattering of Ultrasonic Wave by Cracks in a Plate,” J Appl Mech Trans
ASME, Vol. 60, No. 2, 1993, pp. 352-357.
Lord, W., Z. You, M. Lusk and R. Ludwig, “Numerical Predictions of Surface Wave Phenomena for
Ultrasonic NDE,” Ultrason Symp Proc, Vol. 2, 1988, pp. 1065-1068.
Michaels, J.E., T.E. Michaels and B. Mi, "An Ultrasonic Angle Beam Method for in situ Sizing of
Fastener Hole Cracks," Journal of Nondestructive Evaluation, Vol. 25, 2006, pp. 2-15.
Mullis, R.T. and T. MacInnis, “C-141 Spanwise Splice Advanced NDI Method,” Proceedings of the
First Joint DoD/FAA/NASA Conference on Aging Aircraft, Ogden, Utah, July 8-10, 1997.
Murphy, R., Sabbagh, H., Chan, A. and Sabbagh, E., “A Volume-integral Code for Electromagnetic
Nondestructive Evaluation,” Proceedings of the 13th Annual Review of Progress in Applied
Computational Electromagnetics, Monterey, CA, March 1997.
Nagy, P.B., M. Blodgett and M. Golis, “Weep Hole Inspection by Circumferential Creeping Waves,”
NDT&E International, Vol. 27, No. 3, 1994, pp.131-142.
Niwa, Y., S. Hirose and M. Kitahara, “Application of the Boundary Integral Equation (BIE) Method to
Transient Response Analysis of Inclusions in a Half Space,” Wave Motion, Vol. 8, 1986, pp. 77-91.
Palanisamy, R. and W. Lord, “Finite Element Modeling of Electromagnetic NDT Phenomena,” IEEE
Trans Magn, Vol. MAG-15, No. 6, 1979, pp. 1479-1481.
Schmerr, L.W., Fundamentals of Ultrasonic Nondestructive Evaluation, Plenum Publ., New York, 1998.
Schmerr, L.W., “Modeling and Simulation of NDE Inspections,” Rev Prog Quant Nondestr Eval, v 18,
1999, pp. 679-686.
Schuhmacher, S., P. Zanger, and K.J. Langenberg, “System Model to Predict the Results of Ultrasonic
Scattering Experiments,” Journal of Nondestructive Evaluation, Vol. 13, No. 3, 1994, pp. 147-154.
Spies, M., “Simulation of Ultrasonic Testing of Complex-structured Materials and Components,” Proc
IEEE Ultrason Symp, Vol. 1, 1999, pp. 791-800.
Smith, K. D., Thompson, R. B. and Brasche, L., “Model-Based POD: Successes and Opportunities,” 1st
Meeting of the MAPOD Working Group, Albuquerque, New Mexico, September 23-24, 2004, web
site: http://www.cnde.iastate.edu/MAPOD/.

20
Smith, R., D. Edgar and L.D. Jones, “Rapid Second-layer Crack Detection Using Phased-arrays and
Automated Analysis,” Proceedings of the 8th Joint FAA/DoD/NASA Conference on Aging Aircraft,
Palm Springs, CA, January 31 - February 3, 2005.
Taylor, R.L., FEAP - A Finite Element Analysis Program, Theory Manual, University of California,
Berkeley, 2007, web site: http://www.ce.berkeley.edu/~rlt.
Theodoulidis, T.P. and E.E. Kriezis, Eddy Current Canonical Problems (with Applications to
Nondestructive Evaluation), Tech Science Press, 2006.
Thompson, R.B. and D.O. Thompson, “Ultrasonics in Nondestructive Evaluation,” Proc IEEE, Vol. 73,
No. 12, 1985, pp. 1716-1755.
Thompson, R.B., “Using Physical Models of the Testing Process in Determination of Probability of
Detection,” Materials Evaluation, Vol. 69, No. 7, 2001, pp. 861-865.
Wojcik, G.L., D.K. Vaughan, N.N. Abboud and J. Mould, “Electromechanical Modeling Using Explicit-
time-domain Finite Elements,” Proc. IEEE Ultrason. Symp., Vol. 2, 1993, pp. 1107-1112.
Xuan, L., Z. Zeng, B. Shanker and L. Udpa, "Element-free Galerkin Method for Static and Quasi-static
Electromagnetic Field Computation," IEEE Trans Magn, Vol. 40, No. 1, 2004, pp. 12-20.

21
 NDE Measurement Technique Model

Source Components Sample Components Reception Components

• Input Signal • Incident Field • Interface Condition

• Source Hardware • Material Properties • Transducer
• Electrical Connection • Sample Geometry • Electrical Connection
• Transducer • Flaw Characteristics • DAQ Hardware
• Interface Condition • Scattered Field

Figure 1. Components of NDE measurement technique model.

22
 NDE Measurement Technique Development Process

Phases Role of Modeling

• Determine feasibility of inspection method(s) for

particular inspection problem
1. Design • Aid interpretation of raw data
• Provide information for setup, transducer design
• Identify features for measurement, classification
• Expand training data set for automated classifier
• Use as feature classifier through inverse methods
• Estimate technique probability of detection (POD)
2. Validation
• Provide data for model-assisted probability of
detection (MAPOD) validation study
• Demonstrate physical understanding of
technique to project sponsors
3. Implementation
• Use simulation to aid instruction of inspectors

Figure 2. Role of modeling in NDE technique development process.

23
 z z

y x

layers

spiral creeping wave

y
fastener site
transducer θ
x

corner crack

Figure 3. Schematic diagram of the ultrasonic inspection of fastener sites for corner cracks in multilayer
structures using spiral creeping waves to detect corner cracks in the shadow region of the hole.

24
 bottom surface around hole (near side) around hole (far side)

transducer (on top)

(a)
notch
notch

(b)

(c)

Figure 4. Simulated response of ultrasonic wave scattering around a fastener site with a notch in 3D
from several views (left column: bottom surface; center and right column; hole surface) and at select
times (a) main shear wave incident at hole surface, (b) spiral creeping wave reflecting from hole bottom
edge and (c) spiral creeping wave propagating upward with corner surface wave at scattering at notch.

25
 eddy current fastener site
probe z
x
c

corner crack b

a (a)

z
x

through crack

a (b)

Figure 5. Diagram of fastener site model with (a) a corner crack and (b) a through crack.

26
 10 VIM
FEM
experiment
8
Re(∆Z) (Ohms)

-5 -4 -3 -2 -1 0 1
Im(∆Z) (Ohms)

Figure 6. Comparison of simulation and experimental results in the impedance plane for the case study
eddy current inspection problem.

27
 1.2E-01
transition (2nd layer
thickness = 0.10")
1.0E-01 a
through
'crack'
8.0E-02 model
â (V)

6.0E-02
experimental
measure

4.0E-02

2.0E-02 a
corner
'crack'
model
0.0E+00
0.00 0.05 0.10 0.15 0.20 0.25
a (inches)

Figure 7. Comparison of experimental and simulated data for varying crack length and type.

28
 Probability of Detection (

1.0

0.8

model-assisted POD curve

0.6 (from Monte Carlo
simulation)

0.4

0.2
experimental POD curve
90/95% confidence bounds
0.0

0.00 0.05 0.10 0.15 0.20

target size (inches)

Figure 8. POD evaluation results for experimental and full model-assisted POD studies.

29

View publication stats