© U.S. Air Force, Liquid Penetrant © U.S.

Air Force, Magnetic Particle Testing

Airbus S.A.S Carbon fibre

© Wolfram Schroll

American Airlines Boeing 737 © AA Media

This proceedings is intended for the sole use of registered attendees. No license is granted to reproduce, copy, forward, or
redistribute this proceedings in any form or fashion and is exclusively intended for personal use only. No part of this publication or
its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.

Copyright 2018 by The American Society for Nondestructive Testing.

Exclusive of those papers that are a work of the federal government and not subject to copyright.

The American Society for Nondestructive Testing, Inc. (ASNT) is not responsible for the authenticity or accuracy of information
herein. Published opinions and statements do not necessarily reflect the opinion of ASNT. Products or services that are advertisedor
mentioned do not carry the endorsement or recommendation of ASNT.

No part of this publication may be reproduced or transmitted in any form, by means electronic or mechanical including
photocopying, recording or otherwise, without the expressed prior written permission of The American Society for Nondestructive
Testing, Inc.

IRRSP, NDT Handbook, The NDT Technician and www.asnt.org are trademarks of The American Society for Nondestructive Testing,
Inc. ACCP, ASNT, Level III Study Guide, Materials Evaluation, Nondestructive Testing Handbook, Research in Nondestructive
Evaluation, and RNDE are registered trademarks of The American Society for Nondestructive Testing, Inc.

First printing 10/18

Errata, if available for this printing, may be obtained from ASNT’s website, www.asnt.org.

ISBN: 978-1- 57117-455-0

Printed in the United States of America

Published by: The American Society for Nondestructive Testing, Inc.

1711 Arlingate Lane
Columbus, OH 43228-0518
www.asnt.org

ASNT Mission Statement: ASNT exists to create a safer world by advancing scientific, engineering, and technical knowledge in the
field of nondestructive testing.

ii
 TABLE OF CONTENTS
PAPERS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Nondestructive Testing of Single Crystal Alloy by X-ray Diffraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Alexiane Arnaud, Edward Romero, Clément Remacha, David Murer,
David Just, Vincent Revol, and Henry Proudhon

Inverse Wave Field Extrapolation, the Next Generation of Ultrasonic Inspection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Domenico Bellistri, Harvey Haines, Jeff Vinyard, and Chase Collins

In-Line Inspection of Friction Stir Welds Using Laser Ultrasonics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Bradley Bobbs, Max Wiedmann, and Marvin Klein

New Tools in CIVA for Model Assisted Probability of Detection (MAPOD) to Support NDE Reliability Studies . . . . . 32
Fabrice Foucher, Roman Fernandez, Stéphane Leberre, Pierre Calmon

Flaw Detection Capabilities in Aerospace with Eddy Current Array Technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

Jesse Herrin, Nicholas Cardillo, Steve Timm, and Tom Rohlfing

Quality Classification of Adhesive Bonds in Composite Structures by Single-sided

Air-coupled Ultrasonic Testing Using Linear Phased Array Probes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Mario Kiel, Stefan Scheunemann, Maria Grötzner, Manuel Lucas, and Andreas Mück

Camera Encoded Phased Array for Semi-Automated Inspection of Complex Composite Components . . . . . . . . . . . . 59
Borja Lopez

Implementation of Special PA UT Techniques for Manufacturing Inspection of Friction Stir Welds. . . . . . . . . . . . . . . 65

Michael Monette, Dany Devos, and Daniel Richard

ABSTRACTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .76

Commercial, Accelerator Driven Neutron Radiography Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Brad Bloomquist, Michael J Taylor, Katie Rittenhouse, and Evan Sengbusch

Faster, Easier, and Securer Inspection than Using Mercury Vapour Lamps in Fluorescent Magnetic
and Penetrant Testing UV LED Lamps by Using UV LED Lamps Fulfilling the Newest Aerospace Standards . . . . . . 77
Marc Breit

Predicting Bond Performance of Composites Through Nondestructive Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Brooke Campbell and Giles Dillingham

Improvement to a Novel Ultrasonic Spectroscopy Technique and Its Applications to Bond-Line Analysis . . . . . . . . . . 78
Janelle Chambers

Coherent Adaptive Focusing Technology for the Inspection of Variable Geometry Composite Material. . . . . . . . . . . . 79
Andre Lamarre and Etienne Grondin

Eddy Current Array for Aircraft Engine Component Inspection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

Andre Lamarre

iii
Recent Advances in Computed Radiography Plate Grain Pattern Compensation
and Frame Averaging Techniques with Related Technologies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Sean Larkin, Timothy J. Holmes, Melena Abijaoude,
Robert Schurz, Zhen Ji, and Joseph Goldfeld

Structural Health Monitoring of Aerospace Composite Structures Through Integrated Carbon Nanotube Sensors. . 81
Hao Liu

Detection of Defects in RAM-Covered CFRP Structures by Active Microwave Thermography. . . . . . . . . . . . . . . . . . . 81

Ali Mirala, Kristen Donnell, and Mohammad Tayeb Ghasr

A Novel CT Reconstruction Algorithm for Compensation of Scatter and Beam Hardening . . . . . . . . . . . . . . . . . . . . . . 82

Miguel Pineros, Sean Larkin, Timothy J Holmes, Joseph Goldfeld,
Robert Schurz, and Melena Abijaoude

Electromagnetic Characterization of Advanced Composites by Voxel-Based Inverse Methods. . . . . . . . . . . . . . . . . . . . 82

Harold A. Sabbagh, Elias H. Sabbagh, Kim Murphy, and Russell A. Wincheski

NDE of Aerospace Materials and Composites using Acoustography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Jaswinder S. Sandhu and Charles Pergantis

Magnetic Rubber Testing for Structural Health Monitoring of Military Aircraft. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

Khalid Sulman Shabbir

Thermographic Characterization of Weak Defects in Cured and Uncured Composites. . . . . . . . . . . . . . . . . . . . . . . . . . 83

Steven M. Shepard and Maria Frendberg Beemer

Real-Time Measurement of Aerospace Fastener Loading Using Laser Ultrasonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

Max Wiedmann, Bradley Bobbs, and Marvin Klein

In-Line Inspection of Additive Manufactured Parts Using Laser Ultrasonics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

Max Wiedmann, Bradley Bobbs, and Marvin Klein

Production CT Inspection in Aerospace. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

Jay P. Wood

AUTHOR INDEX. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

iv
PAPERS

1
 Nondestructive Testing of Single Crystal Alloy by X-ray Diffraction
Alexiane Arnaud1, 4, Edward Romero1, Clément Remacha1, David Murer2, David Just2,
Vincent Revol3, and Henry Proudhon4

1
Safran Tech
Nondestructive testing ofRue
single crystal
des Jeunes Boisalloy by X-ray diffraction
78117 Châteaufort, France
email alexiane.arnaud@safrangroup.com;
Alexiane Arnaud1, 4, Edward Romeroedward.romero@safrangroup.com;
, Clément Remacha1, David Murer2clement.remacha@safrangroup.com
, David Just2, Vincent Revol3,
1

Henry Proudhon4,
DECTRIS Ltd. 2
1
Safran Tech
Taefernweg 1
Rue des Jeunes Bois, 78117 Châteaufort, France
5405 Baden-Daettwil, Switzerland
alexiane.arnaud@safrangroup.com, edward.romero@safrangroup.com, clement.remacha@safrangroup.com
+41 56 500 21 00; fax +41 56 500 21 01; email david.murer@dectris.com; david.just@dectris.com
DECTRIS Ltd. 2

CSEM
Taefernweg 1, 5405 3Baden-Daettwil, Switzerland
Technopark
+41 56 500 21 00; fax +41 56 500 21 01; david.murer@dectris.com, david.just@dectris.com
Technoparkstrasse 1, 8005 Zürich, Switzerland
email vincent.revol@csem.ch
CSEM
3

Technopark, Technoparkstrasse 1, 8005 Zürich, Switzerland

4vincent.revol@csem.ch
Centre des Matériaux
Centre desMines Paristech
Matériaux,
4
Mines Paristech
63-65 ruerue
63-65 Henri-Auguste
Henri-AugusteDesbruères, BP-87,91003
Desbruères, BP-87, 91003 Evry,
Evry, France
France
+33 (0)1 60(0)1
+33 76 30 00;30fax
60 76 00;+33
fax (0)1 60 76
+33 (0)1 60 31 50;50;
76 31 email henry.proudhon@mines-paristech.fr
henry.proudhon@mines-paristech.fr

ABSTRACT
In the aerospace industry, quality control of mono-crystalline metallic alloys, such as those found in turbine blades, is
mandatory. Control can be performed with a variety of X-ray methods known as Laue diffraction techniques. These
methods allow to infer structural information of a mono-crystalline material, such as its crystal orientation or structural
defects, through the reflection or transmission of polychromatic X-ray beams.

Laue diffraction methods are categorized according to the relative position of the X-ray source, the detector and the
sample. The back- or side reflection method allows to determine defects on the surface of the sample material.
However, the advent of mono-crystalline structures featuring more complex geometric shapes has rendered such
methods partly inadequate for efficient and accurate quality control. As internal defects within the sample volume
cannot be detected with this method. These drawbacks are overcome with the Transmission Laue technique, in which
the detector is placed behind the crystal to record X-ray beams transmitted through it.

In this work, after a brief introduction to Transmission Laue, we explore in detail the impact of detector choice on the
performance and feasibility of the Transmission Laue method. To this end, a series of similar tests have been
conducted on a classic CsI flat Panel and at the CSEM facilities in Zurich with a DECTRIS PILATUS 3 CdTe detector.
A quantitative comparison of the efficiency of these detectors for this imaging modality is derived. Several parameters
such as the sample thickness or the exposure time have been compared in order to evaluate the accuracy of the
acquisition.

Keywords: X-ray diffraction, single crystal, Laue transmission method, nickel based superalloys.

INTRODUCTION
An X-ray system, to perform crystalline quality control on internal cavities of high pressure turbines blades, is being
developed under the framework of a PhD thesis between Mines ParisTech and Safran. The system uses the Laue
transmission technique to display, into a bidimensional detector, the blade diffraction pattern generated by a
polychromatic X-ray beam. Then the diffraction pattern is used to identify the blade crystalline orientation. Better is
a diffraction pattern (intensity and diffraction spots quantity), easier is the identification process. Then the detector
capability to display correctly a diffraction pattern is a key aspect of the system configuration. First tests of feasibility

2
Copyright 2018. This paper is intended for the sole use of registered attendees. No part of this publication or its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.
have been performed with indirect detectors [1]. Diffraction patterns have been obtained, nevertheless the combination
of low intensity of the diffracted beam and limited dynamic range of this kind of detectors, lead to low contrast
diffraction image. Direct detectors [2] can offer, theoretically, a layer dynamic range and better contrast due to the
absence of background noise, which leads to better diffraction images.

THEORETICAL BACKGROUND

Single crystal turbine blade

Turbine Blades Generalities

Turbine blades are made of nickel based superalloy. Currently, two different superalloys are used by Safran: the AM1
alloy for aircraft blade, the CMSx4 for helicopter blade. The composition of both superalloys are described in the
Table 1 [3].

Turbines blades are manufactured by investment casting based on lost-wax casting, using Bridgman technique to
obtain oriented single crystal solidification. This provides high strength with low thermal creep [4]. The strength of
the blade is enhanced by the absence of grain boundary.

Table 1: Composition in weight % of alloys used by Safran

Density
Alloys Ni Cr Co Mo Re W Al Ti Ta Hf
(g/cm3)
AM1 Base 7.8 6.5 2 - 5.7 5.2 1.1 7.9 - 8.6
CMSx4 Base 6.5 9 0.6 3 6 5.6 1 6.5 0.1 8.7

Crystalline defects
During the solidification process, foundry defects such as disoriented or parasite grains, can appear. This type of defect
involves a loss of performance because of the presence of grain boundaries. Crystalline defect may come from
inhomogeneities in the solidification furnace, or during heat treatment of the blade.

The likelihood of defects is increased by geometry complexity of blades. If blade cooling system design includes
internal walls, their crystalline orientation cannot be controlled by the existing NDT: the side-reflection Laue method.

Laue transmission method

Laue transmission is a technique that allows to generate diffractions patterns, irradiating a stationary single crystal
with a polychromatic radiation (Bremsstrahlung spectrum, all possible wavelengths within a given range [λmin, λmax])
[5]. A pattern is composed by a set of diffraction spots distributed on the detector. A diffraction spot is observed only
if the relative angle between the single crystal and the X-ray beam satisfied the Bragg law:

(Eq. 1) 2𝑑𝑑ℎ𝑘𝑘𝑘𝑘 sin 𝜃𝜃 = 𝑛𝑛𝑛𝑛

where
(hkl) designates a particular lattice plane,
dhkl is the spacing between the lattice planes,
θ is the angle between the planes and the X-ray beam in the reflection plane,
λ is the wavelength of the diffracted beam,
n is the order of the reflection.

The pattern consists in arrays of high intense spots known as Laue spots [6]. The Laue spots array are due to the
polychromatic beam, where many wavelengths are diffracted by sets of (hkl) planes from the illuminated crystal.

3
Two techniques were developed by Max von Laue 1: a back reflection and a transmission technique. Back reflection
is used to generate diffraction pattern from the surface of the crystals, while transmission allows to generate diffraction
patterns from the interior of the crystals. For applications such as volume NDT on high pressure turbines blades (were
internal structures needed to be controlled), the transmission technique is the most appropriate.

A characteristic of the Laue transmission patterns is that the diffraction spots appear along ellipses curves running
through the projection of incident beam on the detector, see Fig. 1. Spots located on the same ellipse prove to be the
reflections of planes that lie in the same zone i.e. which contain a common crystallographic direction: the zone axis.
Every zone axis with an angle up to 45° with the incident beam result in the formation of a cone of 2ψ aperture angle,
and each diffracted beam on the cone will have 2θ Bragg's angle with the incident beam. The combination of all spots
detected by the detector forms the diffraction pattern.

Figure 1: Elliptical curves which corresponds to the diffracted

beam of planes in a given zone.

Detectors technologies
Two main types of X-rays detectors are available: indirect or direct.

Indirect detector
It is the most commonly used for X-ray imaging systems. The detector works converting the X-rays that reach its
surface into visible light via a scintillator, usually a layer of CsI. Then the light signal is converted into electric signal
by an amorphous silicon semiconductor. The first feasibility test have been performed with an indirect detector: the
PaxScan 2520V from VARIAN Medical System, see Fig. 2.

Figure 2: Varian PaxScan 2520V.

1
Max von Laue, German physicist who won the Nobel Prize in Physics in 1914 for his discovery of the diffraction of
X-rays by crystals.

4
Direct detector
Hybrid Photon Counting (HPC) detectors are comprised of a sensor part and a readout part, with the sensor bump
bonded to the readout ASIC. The sensor material can be chosen for optimal sensitivity in a given energy range, for
higher energy applications, as in this work, CdTe is used. With a dedicated signal processing circuitry on each pixel,
the energy of an X-ray event is analyzed and compared to a reference threshold. For events exceeding the chosen
threshold, a digital counter is incremented by one. This makes the detector virtually noise-free. Due to absence of an
indirect conversion step of the X-ray energy into visible light, no signal blurring occurs. In addition, counting detectors
allow for an almost unlimited dynamic range due to the digital nature of the signal accumulation. These properties
result in a very high signal-to-noise ratio, even when the X-ray signal is very low.

Because of these abilities, HPC detector became the state-of-the-art over the past 10 years for many X-ray techniques
in reciprocal space, for example in macromolecular crystallography or in SAXS/WAXS [7]. Transmission Laue has
similar signal properties, thus the HPC detector technology is a promising candidate to support such NDT methods.
The model used for the present work was the DECTRIS Pilatus 3X 300K, see Fig. 3, normally designed for
synchrotron beam lines [8].

Sensor pixel Readout pixel

X-ray
+
–

CMOS readout chip

Direct conversion of
X-rays in sensor
(1) ( (2) ( (3)
Figure 3: at left: DECTRIS HPC detector technology; at right: DECTRIS Pilatus 3X 300K.

Contrast to-Noise Ratio Calculation

Objectives have been to test the detector DECTRIS Pilatus III under different acquisition conditions, and to compare
the detector performance against an indirect detector. The evaluation criterion was simple: the more intense the
diffraction spots were, the better the systems performs. In order to compare the diffraction spots from images obtained
under different conditions, the Contrast to Noise Ratio [9] has been chosen as the criterion to evaluate how much a
diffraction spot is visible or not.

̅
𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 −𝑆𝑆̅𝑏𝑏𝑏𝑏𝑏𝑏
(Eq. 2) 𝐶𝐶𝐶𝐶𝐶𝐶 =
𝑆𝑆𝑆𝑆𝑆𝑆(𝑆𝑆𝑏𝑏𝑏𝑏𝑏𝑏 )

where
̅
𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ̅
and 𝑆𝑆𝑏𝑏𝑏𝑏𝑏𝑏 are the mean of the intensity of the spot and the background respectively,
𝑆𝑆𝑆𝑆𝑆𝑆(𝑆𝑆𝑏𝑏𝑏𝑏𝑏𝑏 ) is the standard deviation of all pixels composing the background signal.

In order to extract the different values of the signals from the images, a square region around selected spots was
̅
analyzed. The spot was segmented from the background and average intensity values in the two regions 𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 and
̅
𝑆𝑆𝑏𝑏𝑏𝑏𝑏𝑏 were derived to compute the CNR value.

5
EXPERIMENTAL SETUP
The experimental setup will be divided in 3 parts: the X-ray system, the positioning sample equipment, and the detail
of the sample used.

X-ray system
The X-ray system is composed of a dual spot X-ray tube and a direct detector. The X-ray tube used was the COMET
MXR160HP/11 (available at the test bench of CSEM 2 Zurich). It is a unipolar water-cooled tube assembly with
integrated radiation protection. It comprises dual focal spot, a tungsten anode and a directional beam [10]. The direct
detector used is the DECTRIS PILATUS 3X 300K (see Fig. 3). The most relevant technical specifications for the
diffraction application are detailed in the Table 2.

Table 2: Technical specifications of the X-ray system.

Sensor thickness 1 mm of CdTe
Pixel size 172 x 172 µm²
Detector specifications Dynamic range 20 bits
Area 83.8 x 106.5 mm²
Adjustable threshold range 2.7 – 18 keV
Nominal tube voltage 160 kV
Filament current 4.1 A
X-ray tube specifications Continuous rating 1800 W
Target material W
Target angle 11°

Specific positioning equipment

The position sample equipment has been conceive, to perform Laue Transmission technique with a conventional X-
ray system. It contains a lead collimator plate of 10 mm thickness and 1 mm diameter, an association of linear stages
which allows to move the sample in front of the incident beam, and a beam stop aligned with the collimator's pinhole,
see the schema Fig. 4. This equipment has been designed to easy align collimator, sample and beam-stop with the
incident beam.

d3

d2
d1
Diffracted
beam

Beam stop
X-ray cone beam

Blade
Z Collimators

Bidimensional
X
Diffraction setup Detector

Figure 4: Experimental setup schema.

2 Swiss Center for Electronics and Microtechnology

6
The positions configuration used were:
 d1 ≈ 8 cm, from the source window to the end of the second collimator;
 d2 ≈ 5.4 cm, due to the setup and the support sample;
 d3 = 7 cm.
To limit the diffusion of X-rays, a first collimator (5 cm Pb thickness, 3 mm diameter) have been placed in front of
the windows of the X-ray tube, see Fig. 5.

Figure 5: Experimental diffraction setup, from left to rigth: X-ray tube,

collimator first step, diffraction setup, DECTRIS Pilatus III detector.

Test description
The experimental conditions (acquisition parameters) have been established in order to:
 Generate Laue diffraction patterns under similar acquisition conditions that have been used, previously, with
indirect detectors.
 Test the limits of the image acquisition capabilities of the DECTRIS detector.

Images were taken varying the acquisitions parameters: acquisition time, response of the detector with different sets
of current and voltage combined with the use of the different samples.

Two representative samples have been chosen to perform this test, see Fig. 6. The first sample is a CMSx4 step wedge,
whose thickness varies from 2 mm to 10 mm. In order to explore the detector sensitivity to the flux, 3 acquisition
times and 3 thicknesses of the step wedge have been used: 0.1 s, 1 s, 10 s, and 2 mm, 4 mm, 6 mm respectively, at
120 kV, 5 mA. For higher thicknesses such as 8 mm and 10 mm, X-ray energy has been improved at 160kV.

The other sample is a helicopter blade, tests have been performed on the tip of the blade, not on the hub. The thickness
of the tip of the blade is up to 2 mm of superalloy.

Figure 6: Representative samples, at left a step

wedge, at right a helicopter turbine blade.

7
DIFFRACTION IMAGING RESULTS AND ANALYSIS

Experimental results

Step Wedge sample

The detector capability to show a signal also depends on the time of acquisition, and on the path of the X-ray through
the object. According to this, diffraction pattern has been acquired varying the acquisition time and the thickness of
the X-ray path through the material (see Fig. 7 and Fig. 8). As expected the observed spots brightness of the patterns
have been higher for 2 mm thickness, than for 4 mm thickness; at 6 mm the diffraction patterns have almost
disappeared. At the same time, the 0.1 s pattern is noisier than the 10 s pattern due to the low number of photons on
the detector.

The Pilatus III has also shown diffraction signal up to 10 mm of nickel superalloy with some of the more powerful
acquisition parameters: 160 kV, 5 mA, 10 s time of acquisition (see Fig. 9).

Helicopter blade
Diffraction patterns have been acquired on the upper part of the blade for different acquisition times, from 10s to 0.1s,
with classical parameters such as 120 kV, 5 mA. All patterns present several and clear spots (see Fig. 11).

Figure 7: Diffraction patterns for various thickness of the step wedge, 0.1 s of acquisition time.

Figure 8: Diffraction patterns for various thickness of the step wedge, 10 s of acquisition time.

8
 Figure 9: Diffraction patterns performed through 8 mm (left) and 10 mm (right) of nickel
superalloy. Patterns obtained with powerful parameters: 160 kV, 5 mA, 10 s, without beam stop.

Figure 10: Diffraction patterns of CMSx4 blade, 120 kV, 5 mA, without beam-stop, at left, 10 s
of acquisition time, at right, 0.1 s of acquisition time.

CNR analysis

Step Wedge sample analysis

Thicker is the irradiated object in the beam direction, higher is the absorption and lower is the signal intensity that
reaches the detector. Higher is the acquisition time, higher is the photon number that reaches the detector and higher
is the signal intensity.
CNR vs Thickness curves for different acquisition time, have been performed over the diffraction patterns as observed
on Fig. 11. As expected, CNR values decrease as the material thickness increase, as well as CNR values decrease as
the acquisition time decrease. In both case this is a consequence of the loss of the intensity of the diffraction spots. In
this sense the best CNR value corresponds to the maximum acquisition time and minimum traversed thickness. Note
that the CNR values converge as they decrease. This is due to the fact that the variations of the background signal Sbkg
are smaller than the variations of the diffracted signal Sspot.

9
Blade analysis
CNR calculation results for the blade sample are of 116.2 at an acquisition time of 10 s and 46.5 at an acquisition time
of 0.1 s.
The CNR value obtained with acquisition time of 0.1 s is lower than the one at 10 s. Nevertheless all the observed
patterns displays many diffraction spots which, in theory, are sufficient to identify the crystalline orientation of the
samples, and then satisfy the industrial requirements.

Figure 11: CNR estimation for 3 thickness values over

the step wedge and 3 acquisition time.

Comparison with flat panel detector

The most similar image in term of acquisition parameters for this comparison have been realized with a third
superalloy single crystal sample. Acquisition parameters for each detector were:
 Pattern from Pilatus: 120 kV, 0.5 mA, d3 = 10 cm, 1 s ;
 Pattern from flat panel: 120 kV, 0.9 mA, d3 = 9.4 cm, 2 s exposure time, 4 averaged frames.

Laue diffraction patterns observed at the Fig. 13, are not the same due to the difference of the sample position on the
support. To calculate the CNR two spots have been chosen at each image, each couple of spots placed more or less at
the same distance from center of the pattern: two high intensity spots (see the square in Fig. 12), and two low intensity
spots (see the circle in Fig. 12).

For the background calculation, an area of 50x50 pixels has been selected at the left corner of the Pilatus image (see
the crosses at Fig. 12). And a similar area at a similar position respect to the center of the pattern of the Varian image.

The CNR results are shown at the Table 3. Note that the CNR values obtained with the direct detector are slightly
lower than those obtained with the indirect Varian detector. Nevertheless, the acquisition time and the current intensity
used with the DECTRIS detector were lower than those used with the Varian detector, which were meant to improve
the CNR value. Overall it can be considered that the DECTRIS detector can obtain diffraction patterns (with less
acquisition time and less current) equivalent to those obtained with the Varian detector (with more acquisition time
and more current).

10
 Table 3: CNR result for the comparison of the direct and indirect detector
CNR Pilatus Flat Panel
Shiny spot 47.15 55.08
Low-light spot 8.37 15.38

Figure 12: Laue diffraction patterns of the same sample (not oriented on the same direction) used to compare
direct, and indirect detector.

The direct detector DECTRIS Pilatus III allows to perform the acquisition without beam stop, with a minor influence
over the image quality of the diffraction pattern. The equivalent test with a flat panel detector could not be achieve
because of the saturation due to the incident beam.

CONCLUSION
The direct detector DECTRIS Pilatus III allows to detect a significant diffraction signal up to 4 mm of nickel
superalloy with standard parameters (120 kV) while 4 mm is the current thickness limit of detection with a flat panel
detector. The acquisition time is drastically reduced with direct detector. 0.1 s acquisition time is short enough to
perform an efficient diffraction control and no beam stop.

The direct detector DECTRIS Pilatus III has shown better performance regarding Laue Transmission technique
applications, than the indirect detector VARIAN PaxScan 2520V. It is important to mention that other indirect
detectors (such as Perkin Elmer models, not used for this test) have been also used in the framework of the
development of the NDT system driving this study. Qualitatively speaking, these indirect detectors are also
outperformed by the DECTRIS Pilatus III, regarding the considered diffraction application.

Further experiences with other direct detectors models could help to confirm the pertinence of this technology for
NDT applications using the Laue Transmission technique in particular regarding the detection of crystalline defect
within a known orientation.

ACKNOWLEDGMENTS
The authors would like to acknowledge Safran Tech for its financial support, DECTRIS for the detector technology
and the CSEM for the test facilities.

11
REFERENCES :
[1] Esnouf, C. (2011) Caractérisation microstructurale des matériaux: Analyse par les rayonnements X et
électroniques, PPUR Presses polytechniques.
[2] Schmitt, B., Brönnimann, C., Eikenberry, E. F., Hülsen, G., Toyokawa, H., Horisberger, R., ... & Tomizaki, T.
(2004). Development of single photon counting detectors at the Swiss Light Source. Nuclear Instruments and
Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment,
518(1-2), 436-439.
[3] ONERA. Journée 3AF sur les superalliages, Juin 2016.
[4] Reed, R. C. (2008). The superalloys: fundamentals and applications. Cambridge university press.
[5] Amorós J.-L., Buerger M.-J. and De Amorós M.-C. (1975). The Laue Method, Academic Press.
[6] Cullity B. D. (1978). Elements of XRD. USA Edison-Wesley P Inc.
[7] Brönnimann, C., & Trüb, P. (2014). Hybrid pixel photon counting X-ray detectors for synchrotron radiation.
Synchrotron Light Sources and Free-Electron Lasers: Accelerator Physics, Instrumentation and Science
Applications, 1-29.
[8] DECTRIS. https://www.dectris.com. Consulted in August 2017.
[9] Gonzalez, R. C., & Woods, R. E. (2012). Digital image processing.
[10] COMET. http://www.comet-xray.com. Consulted in August 2017.

12
 Inverse Wave Field Extrapolation, the Next Generation of Ultrasonic Inspection
Inverse Wave field Extrapolation, the next generation of Ultrasonic Inspection
Domenico Bellistri, Harvey Haines, Jeff Vinyard, and Chase Collins
Domenico Bellistri1, Harvey Haines1, Jeff Vinyard1, and Chase Collins1
Applus RTD USA
Technology Center Center
1
Applus RTD USA, Technology
11801 S Sam 11801 S Sam Houston Pkwy
Houston Pkwy, Houston, TX 77031, USA
Houston, TX 77031
(832) 295-5023; fax (832) 295-5001;
(832) 295-5023; fax (832) 295-5001; email domenico.bellistri@applusrtd.com;
domenico.bellistri@applusrtd.com; harvey.haines@applusrtd.com; jeff.vinyard@
harvey.haines@applusrtd.com;jeff.vinyard@applusrtd.com;
applusrtd.com; chase.collins@applusrtd.com
chase.collins@applusrtd.com

ABSTRACT
NDE in industry has a vital role in assuring integrity of assets and safety for people and environment.
Friction Stir Welding (FSW) is among the welding technique aerospace industry uses. It is used in a range of
industries including military, naval, and rail. FSW allows for rapid and high quality welding of previously difficult
to weld materials such as 2XXX and 7XXX aluminum alloys, among the others. Phased Array (PA) has been
proven effective in inspecting FSW and most of the aerospace companies have it among their advanced NDE
techniques, but improvements in characterization of defects may help.

The new Full Matrix Capture (FMC) family of ultrasonic imaging measurements is showing great potential and is
considered the natural evolution of PA. The majority of FMC techniques in the market uses a single array probe
with only one acquisition mode at a time and has slow data processing. Inverse Wave Extrapolation (IWEX) is a
fast processing technique for FMC data that uses two array probes and up to 13 simultaneous acquisition modes,
with each mode providing a different inspection angle of an anomaly. With this paper we describe the IWEX
technique and share the results and benefits of using it over PA, such as straightforward interpretation including
automation of identification and sizing of indications, flaws and defects.

Keywords: friction stir welding, full matrix capture, inverse wave extrapolation, phased array, aerospace

INTRODUCTION
PA is the ultrasonic inspection method currently used to inspect FSW. Although it represents a remarkable
advancement compared with the single UT (Ultrasonic Technique) A-Scan, it still has some limitations i.e. focusing
depth and difficulty to show vertical features, making interpretation something for highly skilled and very
experienced personnel. IWEX belongs to the FMC family of ultrasonic imaging techniques and, although it makes
use of array probes, it creates the images differently than steering the beam and stacking multiple A-Scans together:
it still makes use of A-Scans but it combines all of them to build every single point of the final image that resembles
the cross-section of the volume under examination, see Figure 1 [1]. For this paper, both PA and IWEX have been
used to see how each of the techniques images the flaws in FSW samples.

2mm

(a) (b)
Figure 1: 1mm side drill hole (a), IWEX cross-section view (b)

Copyright 2018. This paper is intended for the sole use of registered attendees. No part of this publication or its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.

13
METHODS

Phased Array
PA has improved the way UT technology can be used allowing multiple beams to be generated nearly
simultaneously from a single point from a single array probe. Encoding capabilities let dedicated software stack the
received signals in a four dimensional format: x, y, z, with a color scale for amplitude. The resulting images are
more understandable than an amplitude vs. time chart and can be re-analyzed at will once the array probe is no
longer on the specimen.

Depending on the orientation of the flaws, different sizing techniques can be used. For flaws parallel to the surface a
0⁰ wedge and array probe can be used for detection when the geometry of the examination piece allows for the
placement of the probe/wedge above the area of interest. When the geometry of the examination piece, such as weld
reinforcement, doesn’t allow the use of a zero degree approach, angle beams must be relied on to detect flaws
parallel to the scan surface with greatly reduced detection capabilities. Flaws with a pronounced vertical component
may be detected by looking for corner traps if surface breaking, and tip diffractions if not surface breaking and
vertically aligned. An alternative to detect and size vertical flaws is using DPA (Dual Phased Array) probes with a
TRL (Transmit Receive Longitudinal) setup [2]. Angled flaws with surface perpendicular to the beam(s) are easier
to detect and are sized using amplitude drop, maximum amplitude or tip diffractions techniques having the PA probe
on the opposite skew for the last one. Results may change if using a sectorial scan vs. a linear scan and flaws may be
missed, though: linear fixed angle scans should be used and should be complemented by at least one or two other
angle beams, and sectorial swept angle scans should not be used alone [3].

To ensure the best coverage for the inspection multiple passes on both sides of the weld and at least two different PA
wedges, one 0⁰ and one angle wedge, are required.

Full Matrix Capture

When pulsing a single element of an array probe and simultaneously recording with all elements in the array
multiple time domain A-Scans are generated. For a N-element array probe we are going to have N A-Scans. When
we repeat this process pulsing each element of the N-element array probe N2 A-Scans will be recorded and saved in
what is called Full Matrix Capture (FMC). The size of a FMC file of N2 A-Scans with 4096 samples of 2 bytes
generated by two 64 elements array probes on a given position along the weld is 128 megabytes (2x4096x1282
bytes/10242 = 131072 kilobytes/1024). If this process is repeated for each position of the array probe along the weld,
a 1000mm (~3ft) long scan is going to be 125 gigabytes (128 megabytes x 1000/1024).

The collected A-Scans can be processed using methods like Total Focusing Method (TFM) and IWEX, with the first
developed largely heuristically and the second based on more explicit mathematical models of the wave propagation
and scattering processes: the received wave field is back propagated using the angular spectrum decomposition [4].

INVERSE WAVE FIELD EXTRAPOLATION

Although IWEX as imaging technique is not new [5] [6], it only has been used in the field since 2016 thanks to the
achievement of an acceptable scan speed of 5mm per second, and improved portability of the system. Examinations
have been carried out for new construction girth weld inspection and for vintage pipelines inspection of seam weld
areas identified by In Line Inspection (ILI) tools [7].

IWEX is a direct application of the Rayleigh integral for wave field extrapolation. This integral enables the
calculation of the sound pressure at any point in an object based on measurements of the sound pressure distribution
at the surface of the object. The FMC dataset is processed by the IWEX algorithm, which effectively corrects the
amplitude and phase of the signals for propagation effects and generates an image by focusing virtually on a number

14
of image points in the region of interest [8]. Thanks to its capability of generating images that resemble the cross-
section of the volume under examination, characterization of anomalies is a straightforward process.

Principles
Each cross-section image is a combination of up to 13 images, each one related to the path of each A-Scan gathered
during the FMC and named “mode-k” where k is equal to the number of skips or reflections off the surfaces, see
Figure 2.

Figure 2: wall thickness/time vs modes

The modes with an odd number of skips are used to display defects with vertical orientation: the sound wave path is
similar to a tandem configuration in which the sender probe and receiver probe have a specific offset [9]. The even-
numbered modes are typically used to display the corner traps and the tip diffractions of imperfections and the
reflections of imperfections when they are inclined to the surface and approximately perpendicular to the sound
wave. The "cross" modes, with an even number of skips, have a pitch-catch configuration and may detect flaws
parallel to the surfaces, Figure 3.

For applications where the weld geometry is fairly uniform (parallel surfaces, no high-low), the number of applied
modes is normally 10: 4 modes from the left probe, 4 modes from the right probe and 2 cross modes by both probes
for ID (Inner Diameter) and OD (outer Diameter) surface imaging. The mode-0 left and the mode-0 right are used
for the image of inclined reflectors, tip diffractions and corner traps near the ID surface. Mode-1 left and mode-1
right are used to imagine the vertical component of reflectors near the ID surface. Mode-2 left and mode-2 right are
used for the detection of inclined reflectors, tip diffractions and corner traps of defects near the OD surface. Mode-3
left and mode-3 right are used to imagine the vertical component of reflectors near the OD surface. Cross mode-0 is
used for reflectors parallel and near to the ID surface, and the cross mode-2 is used to image defects surface parallel
and near to the OD surface. Three additional modes, mode-4 left, right mode-4 and cross mode-4, are used for the
examination of welds influenced by the geometry: non-parallel surfaces, high-low.

Each image corresponding to an IWEX mode is flipped vertically according to the number of skips and aligned with
the images of the other IWEX modes.

Figure 3: IWEX modes

15
Calibration / Setup
The IWEX method doesn’t require a calibration block: the values of the parameters necessary to align the cross
modes and compute the focal laws are calculated by a software calibration tool once the two probes are sitting on the
specimen to be examined. Two array probes with typically with 37⁰ plastic wedges are placed on the surface of the
specimen to be examined. For weld examinations, the probes will be placed on both sides of the weld. The
parameters are calculated in two steps: first the acquisition system parameters and then the examination specimen
parameters. Probe latency, wedge velocity and wedge dimensions are calculated in the first step. Wall thickness,
material velocity, wedges offset (distance between the two wedges), and probe rotation are part of the second step
and it is necessary to have the two probes and wedges sitting on the specimen with couplant.

Scan
Once the setup is complete the scan speed depends upon the size of the cross-sectional image, the resolution of the
image and the number of active modes. Image width and height is up to 1024 pixels times the resolution. Image
horizontal and vertical resolution can be as high (low value) as the total pixels count reaches 65536. For example,
25.5mm (~1”) by 25.5mm (~1”) image will have a resolution of 0.1mm (0.004”). Typical scan speed with these
parameters is 5mm per second.

While the size of the raw FMC data acquired for a single cross-section is 128 megabytes, each FMC data scan is
processed into a set of images, one per mode, and stored in an image file which is less than 1 gigabyte of storage
capacity for 1000mm (~3ft) scan length. The raw FMC data is discarded once the images are built, reducing the data
storage requirements by over 2 orders of magnitude.

TEST
Tests with IWEX has been carried out on FSW samples using two 37⁰ Rexolite flat wedges and two 10MHz 64
elements array probes. Image resolution was 0.1mm. The same samples were tested with PA using a 5MHz,
32 elements (Olympus 5L32-A11) array probe with a SA11C-N55S-IHC-NORING wedge for angled scans and a
5MHz, 64 elements (Olympus 5L64-A12) probe with a SA12-0L wedge for 0⁰ scans.

The results of one sample, aluminum to aluminum FSW with 11.8mm wall thickness, are shown.
Figure 5 is a comparison between the specimen cross-section of the FSW area of an aluminum sample (a) and
the IWEX cross-section view (b). Figure 4 shows IWEX left modes (a), cross modes (b) and right modes (c)
images: Figure 5 (b) is the combination of all these modes.

(a) (b) (c)

Figure 4: left modes signal (a), cross modes signal (b), right modes signal (c)

Figure 6 is a comparison between PA results and IWEX results respectively:

• (a) and (b) on unwelded area.
• (c) and (d) on area with partial penetration with void.

16
(e) is the PA 0⁰ scan on partial penetration and void area.

(a) (b)
Figure 5: alu to alu 11.8mm wt FSW cross-section (a), IWEX cross-section view (b)

(a) (b)

(c) (d)

(e)

17
Figure 6: PA S-Scan (35⁰ to 70⁰) on corner traps from unwelded area (a), IWEX cross-section on unwelded
area (b), PA S-Scan on partial penetration and void (c), IWEX cross-section on partial penetration and void
(d), PA 0⁰ scan on partial penetration and void (e)

It is possible to see how the PA results require some interpretation time to be mapped with the cross-section in
Figure 5(a) whereas with the IWEX images the correspondence is almost immediate. Figure 6 (a) shows the corner
traps of a total lack of penetration. Figure 6 (c) shows corner traps and tip diffractions of the lack of penetration and
the void. Figure 7 show a possible interpretation of the PA sectorial scan.

Reflection

Tip Diffraction Tip diffraction

OD Corner trap

Figure 7: PA S-Scan on partial penetration and void in Figure 6

PA is able to see only inclined flaws perpendicular to its beam, corner traps and the tip diffraction, and only an
experienced and expert eye can give the correct characterization of the flaw. On the other hand the results obtained
with IWEX contain enough information not only for inclined flaws but also for flaws vertically and horizontally
oriented.

In Figure 8 we compare the data from another position on the weld: in (a) we have the PA data, in (b) the IWEX
mode-2 and mode-3, in (c) all the IWEX modes.

18
 OD Corner trap
Reflection

Tip Diffraction

(a) (b) (c)

Figure 8: PA sectorial scan (a), IWEX mode-2 left and mode-3 left (b), IWEX all modes (c)

A certain level of data redundancy is guaranteed thanks to the use of multiple modes from both sides of the weld,
see Figure 4, since most of the times a flaw can be sized using more than one mode. Figure 9 and Figure 10 contain
the strip charts from both IWEX and PA. In Figure 9 (a) is the C-Scans of a volume from 3mm above the ID surface
down to the rest of the cross-section. In Figure 9 (b) is the PA C-Scan with the gate starting at 9.2mm.
In Figure 10 (a) and (b) we have the IWEX left and right side views respectively. In Figure 10 (c) we have the PA
B-Scan with the angle cursor on the ID corner trap of the unfused face. With the label A is indicated area with the
total lack of penetration.

A A

A A

(a) (a) (b) (b)

Figure 9: IWEX C-Scan
Figure (a), PA C-Scan
9: IWEX (b) (a), PA C-Scan (b)
C-Scan

19
 A A A

(a) (b) (c)

Figure 10: IWEX left side view (a), IWEX right side view (b), PA B-Scan (c)

The straightforward interpretation allows shorter learning time therefore a certain level of “human” redundancy for
the interpretation process is achievable.

Since the IWEX results are focused on each point of the image according the given resolution, automation based on
image amplitude can be used to detect and characterize the flaws: intersection points between tandem modes as well
as cross modes and skip modes define the start and the end of vertical and horizontal flaws respectively.

Sizing based on corner traps and tip diffractions has proven to be more effective than sizing based on amplitude
attenuation. Sizing errors standard deviations have been as low as ±0.2 mm [7].

CONCLUSION
The techniques belonging to the FMC family, compared with PA, provide a new way of displaying the scanned data
thanks to the simultaneous use of multiple “modes” that make use of direct insonification and indirect measurement
via the back wall, giving a better representation of flaws of any orientation and improved sizing accuracy. Current
usage of the IWEX technology is for verification of ILI calls for crack-like features such as ERW/FW seam welds
and SCC calls. Metallographic sections have been produced to verify the validity of the method confirming its
potential in generating images closer to actual flaw configuration. Further testing with typical FSW flaws is
recommended to reach an acceptable level of confidence.

20
REFERENCES

[1] N. Pörtzgen, Imaging of Defects in Girth Welds using Imaging of Defects in Girth Welds
using Ultrasonic Data, Delft: N. Portzgen, Laboratory of Acoustical Imaging and Sound
Control, Faculty of Applied Sciences, Delft University of Technology, Delft, The
Netherlands, 2007.
[2] A. Van Den Biggelaar, F. Dijkstra and K. Chougrani, MODIFIED PHASED ARRAY
CONCEPT FOR AUT OF LNG STORAGE TANKS, Rotterdam, The Netherlands: APPLUS
RTD, 2010.
[3] T. Armitt, Phased Arrays Not The Answer To Every Application, ECNDT, 2006.
[4] P. D. Wilcox, ULTRASONIC ARRAYS IN NDE: BEYOND THE B-SCAN, Bristol, UK:
American Institute of Physics, 2013.
[5] N. Pörtzgen, F. H. Dijkstra, A. Gisolf and G. Blacquière, "ADVANCES IN IMAGING OF NDT
RESULTS," in Proceedings of the 16th WCNDT, Montreal (Canada), 2004.
[6] K. Chougrani and N. Pörtzgen, "IWEX: A New Ultrasonic Array Technology for Direct
Imaging of Subsurface Defects," in 17th World Conference on Nondestructive Testing,
Shanghai, China, 2008.
[7] H. Haines, L. Hörchens and P. Tomar, IDENTIFYING AND SIZING AXIAL SEAM WELD
FLAWS IN ERW PIPE SEAMS AND SCC IN THE PIPE BODY USING ULTRASONIC
IMAGING, Calgary: ASME, 2016.
[8] X. Deleye, L. Hörchens and K. Chougrani, "Experimental Comparison of Wave-field
Based Ultrasonic Imaging with other Advanced Ultrasonic Weld Inspection
Techniques," in 18th World Conference on Nondestructive Testing, Durban, South Africa,
2012.
[9] L. Hörchens, X. Deleye and K. Chougrani, Ultrasonic imaging of welds using boundary
reflections, Melville, NY: American Institute of Physics, 2013.

21
 In-Line Inspection
In-Line Inspection of
of Friction
Friction Stir
Stir Welds
Welds Using
Using Laser
Laser Ultrasonics
Ultrasonics
BradleyBobbs
Bradley Bobbs,
1
Max Wiedmann
, Max Wiedmann,1, and
and Marvin
Marvin Klein
Klein1

Optech
Optech Ventures
Ventures
1

2520 West 237th

2520 West Street
237th Street
Torrance,
Torrance, CA,CA, 90505
90505 USA
(424) 263-6326; fax (310) 530-7417; email BBobbs@intopsys.com
(424) 263-6326; fax (310) 530-7417; BBobbs@intopsys.com

ABSTRACT
Friction Stir Welding is widely used for joining of aluminum components. Current post-process quality control
techniques often cannot keep up with the production flow. Thus, there is a need for an in-line inspection technique
that can qualify parts as they are produced. If flaws are detected immediately, corrective actions can be taken,
thereby avoiding scrap and saving otherwise wasted production time.
We have applied laser ultrasonic testing for in-line weld inspection. In this method, a pulsed laser beam generates
ultrasonic waves inside the weld region. The waves interact with internal defects and then return to the surface,
where signal perturbations that characterize the defects are detected by a separate laser receiver. The output signals
are processed using advanced algorithms that can be programmed for automated real-time measurement.
We have performed extensive laboratory and production testing to evaluate our technique. Test samples were
aluminum sheets containing linear friction stir butt welds. Sheet thickness ranged from 3 mm to 10 mm. The
defects of interest for this study were elongated voids, lack of penetration, and porosity defects. We were able to
detect voids below 0.3 mm in cross-section at process speeds.

Keywords: friction stir welding, weld inspection, ultrasonic testing, ultrasound inspection, laser ultrasonics

INTRODUCTION
Friction Stir Welding
The joining of metals for aerospace components and structures has traditionally been performed using fasteners or
by fusion welding. Friction stir welding[1] (FSW) has been introduced as an alternative by The Welding Institute[2],
particularly for seam welds in aluminum. FSW is a solid-state joining process in which a cylindrical shouldered tool
with a profiled pin is rotated and plunged into the joint area between two pieces of sheet or plate material. The parts
must be securely clamped to prevent the joint faces from being forced apart. Frictional heat between the wear-
resistant welding tool and the workpieces causes the latter to soften without reaching their melting point, allowing
the tool to traverse along the weld line. The plasticized material, transferred to the trailing edge of the tool pin, is
forged through intimate contact with the tool shoulder. On cooling, a solid-phase bond is created between the
workpieces. FSW in aluminum is commonly used for plate butt welds, but lap welds and more complicated weld
profiles have also been produced.

FSW was first commercialized for aerospace applications, but the technique has also been widely adopted in the
shipbuilding, automotive, and mass-transit industries. It has numerous advantages over fasteners or fusion welding,
including
 Lower weight joints
 Lower heat deposition and thus less distortion of flat structures
 No consumables such as filler wire or inert shield gas
 Minimization of fume generation, since there is no melted material
 Ability to join alloys that cannot be welded by fusion methods
 Reduced joint preparation and post-weld clean-up
 Capability to be fully automated

Copyright 2018. This paper is intended for the sole use of registered attendees. No part of this publication or its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.

22
Weld Inspection
Many FSW applications, especially in aerospace, are for the assembly of safety-critical, high-value, single-use
components and structures, where the requirements on weld quality are very strict. As a result, considerable efforts
have been invested to optimize the weld process to produce defect-free welds. Volumetric (especially elongated
pores or wormholes) and lack of penetration (LOP) defects are the most common and the greatest threats to weld
mechanical integrity[3]. LOP defects leading to a lack of bonding or gap at the bottom of a single-sided weld are a
particular concern, as they can act as stress risers in fatigue or during forming. Kissing bonds (also known as hook
defects) occur where there is a gap-free contact but no mechanical bond; however these are generally considered to
be less important.

While significant strides have been made in welding process optimization, defects can still arise from dirty surfaces,
inaccurate seam tracking, a worn or misaligned tool, lack of tool penetration, or gaps in the fit-up between the two
parts prior to welding.[1] Hence, critical welds must all be inspected via nondestructive testing (NDT) to qualify as
100% free from critical defects.

There is a great deal of literature related to techniques currently in use for NDT of FS welds.[4-8] The techniques
receiving the greatest attention are radiographic testing (RT), phased array ultrasonic testing (PAUT), and eddy
current testing (ECT). RT is very sensitive for detecting and imaging volumetric defects, but not effective for LOP
defects. ECT is very effective at detecting and imaging surface-breaking defects and LOP defects when
implemented on the root side of the sample, but cannot detect volumetric defects. PAUT is a powerful ultrasonic
technique that can detect volumetric defects with great sensitivity; but it is difficult to automate the interpretation of
sectorial B-scans, especially in thin materials. In addition, PAUT requires a fluid supply to cool the scan head and
provide ultrasonic coupling to the workpiece.

A serious limitation of all these techniques, as well as dye penetrant and others used for weld NDT, is that they
cannot be used in-line during the welding process, but are instead inherently restricted to inspection after a weld is
completed. Such post-process inspection, even when done in an automated fashion, is costly and time-consuming.
These quality control techniques often cannot keep up with the production flow. Thus, there is a need for an in-line
inspection technique that can qualify parts as they are produced. If weld flaws could be detected immediately,
corrective actions could be taken, thereby avoiding scrap and saving otherwise wasted production time.

Laser Ultrasonic Testing

Another NDT method for qualification of welds is laser ultrasonic testing (LUT), shown in Figure 1. In this
technique, a pulsed “generation” laser beam is directed to the surface of the weld. The energy deposition from the
incident beam generates intense bulk longitudinal and shear ultrasonic waves that travel at all angles inside the weld
region, as well as surface and plate-mode waves. The rich selection of wave types and angles allows the probing of
large weld volumes, in which the inspection geometry and the timing of the signal arrivals can be controlled to
optimize the detection of one or more features. By contrast, conventional transducer-based ultrasonic weld
inspection requires complex phased-array equipment or multiple transducers and positions to cover the full volume
of the weld. The ultrasonic waves interact with internal defects and then return to the surface, where signal
perturbations that characterize the defects are monitored by a separate “detection” laser receiver based on
interferometry to see surface motion produced by the ultrasonic waves. The output signals are processed using
advanced algorithms that can be programmed for automated measurement.

23
Figure 1: Illustration of our in-line LUT system integrated onto a welding system. Red and green lines
indicate generation and detection laser beams respectively.

One key to the success of our technique is a novel adaptive interferometric optical demodulator [9] that detects
nanometer-scale surface displacements at ultrasound frequencies while being completely insensitive to large-scale,
low-frequency vibrations and other motions of the tool and workpiece. We summarize here the advantages of our
LUT system for in-line monitoring:
 Noncontact, with only laser beams reaching the part—no couplant of any kind required
 Small beam footprint on the part allows access to small spaces and curved surfaces and very accurate
pinpointing of defects
 Can perform measurements at high process speeds (5 m/s) and at high temperatures (1100C) on rough
surfaces
 Bulk longitudinal and shear ultrasonic waves readily produced over a range of angles, in addition to surface
and plate-mode waves
 High bandwidth enables high defect sensitivity with good spatial and temporal resolution
 Real-time, automated data acquisition possible

EXPERIMENTAL DETAILS
Technical Approach
We produce laser-generated ultrasonic waves in the ablation regime, where the deposited energy is sufficiently
concentrated in space and time to eject a small amount of material from the surface. The directionality of bulk
ultrasonic waves[10] is shown schematically in Figure 2. Longitudinal waves are generated in a cone around the
surface normal that extends to at least 70. Shear waves are generated only off axis, in a range between about 35
and 60.

Figure 2: Directionality of laser-generated bulk ultrasound waves in the ablation regime.

24
The existence of laser-generated waves propagating at a range of angles from the surface normal is critical for weld
inspection. As they propagate and spread, they flood the weld zone with ultrasound. The broad coverage of the weld
zone with a single point of generation is a major advantage over transducer-based inspection. This feature
potentially allows in-line monitoring with no lateral motion or phasing of the beams, thus resulting in reduced
complexity of the inspection system.

For effective detection of volumetric weld defects, we believe that the best LUT configuration for inspecting butt
welds is the symmetric “pitch-catch” arrangement shown in Figure 3, in which the two laser beams straddle the weld
on the tool side, just within the tool marks. Both shear and longitudinal bulk ultrasonic waves originating at the
generation laser spot transit the weld zone at different angles and in different regions, thereby interrogating a
significant portion of the weld volume. If pores or voids are encountered, a portion of the waves will be scattered,
thereby losing amplitude. Our basic approach is to measure the amplitude and arrival time for a number of the bulk
wave returns, and to correlate changes with the presence of defects in or near their paths. Not shown in Figure 3 are
surface (Rayleigh) waves and surface-skimming waves that travel directly from the point of generation to the point
of detection. These waves could be useful for detecting surface-breaking defects on the top (tool) surface. Also not
shown, for simplicity, are paths for mode-converted waves that change between shear and longitudinal at a surface
reflection.

Figure 3: Symmetric pitch-catch laser beam arrangement on a butt weld, optimized for detection of
volumetric flaws. Paths of bulk waves are shown inside the sample, with surface and mode-converted wave
paths omitted for simplicity. In the path labels on the right, the number refers to the number of single passes
and the letter refers to longitudinal (L) or shear (S) waves.

For effective detection of LOP flaws in welds produced by a pin tool, the beam configuration shown in Figure 3 is
not optimum. While it is reasonable to expect ultrasound wave diffraction from the tip of a LOP flaw (not shown in
Figure 3), this signal would be difficult to extract in the presence of the strong back wall reflection that will arrive
only slightly later. To overcome this limitation, we used the asymmetric configuration shown in Figure 4. In this
case, diffracted waves from the tip of the LOP flaw arrive at a time that is free from interference by the 2L back wall
reflection signal. Later back wall arrivals could overlap with the LOP tip-diffracted arrival, but this can be avoided
for a given sample thickness by proper selection of the beam separation.

Figure 4: Asymmetric pitch-catch laser beam arrangement on a single-sided weld, optimized for detection of
a LOP flaw. The 2L back wall reflection path and the mode-converted LS diffracted path are shown.

25
Scan Configurations
Researchers at the Canadian National Research Council (CNRC) have applied LUT to the inspection of friction stir
welds used in aerospace applications[11-13]. In their approach, the generation and detection beams are scanned
together laterally across the weld (with the tool marks removed) to form an image of the weld interfaces. This lateral
scanning process is repeated at equal increments along the weld, and signal processing is applied to provide a two-
dimensional image of defects in the weld. This approach has high sensitivity to defects, but is poorly suited for in-
line usage, as the requirement for lateral scanning greatly slows down the measurement so that it cannot keep up
with the process speed, and so severely slows down processing. Lateral scanning also adds complexity to the
mechanics and requires considerable computing power for data analysis.

Our LUT approach uses instead a linear 1D scanning configuration where the two laser beams move along the weld
(Figure 5). The data are displayed as a conventional B-scan that plots the temporal signal at each stepped scan
position.

Figure 5: Scanning configuration used in the present study, shown here with the two beams straddling the
weld for volumetric defect detection.

Experimental Apparatus
Figure 6 diagrams our LUT system’s functional operation. The generation laser is directed via a fiber umbilical to a
measurement head onto the sample surface, to produce an ultrasonic pulse. The “probe” (detection) laser beam is
divided in optical fiber to go both directly into the receiver[9] as a reference beam and also onto the sample. The
reflection of the detection beam from the sample surface is collected by the measurement head and recombined with
the reference beam inside the receiver interferometrically. The measurement head has translation adjustments along
the sample surface normal direction, so that the beams can be focused for optimal generation spot and maximum
collection of detection light returned from the surface. The receiver output signal, indicating amplitude and time
delay for various return paths of the ultrasonic pulse, is sent to a computer for near-real-time analysis and display.
The data are normally organized into a standard ultrasonic B-scan, where the horizontal axis is time-of-flight (TOF)
for the ultrasonic pulse and the vertical axis is detection position scanned along the weld.

26
Figure 6: Functional diagram of our LUT system.

EXPERIMENTAL RESULTS
Porosity Defects
A test sample containing a defective region of porosity was provided by the U.S. Naval Surface Warfare Center,
Carderock Division (NSWCCD). An example B-scan of this sample is shown in Figure 7. There is a dramatic
difference in the signals acquired within the defective region, resulting from attenuation and scattering by the pores.
While smaller regions may produce less notable signal changes, modern signal processing techniques can often still
detect these changes and correlate the signals to the region size. The signal acquisition and processing can be fully
automated, thereby eliminating the need for a skilled technician to create and interpret the B-scans.

Figure 7: B-scan of sample containing a ~35 mm region of porosity.

Lack of Penetration Defects

A sample containing simulated LOP defects was provided by the Space Exploration Company (SpaceX) in
Hawthorne, CA. The notch-shaped defects were produced on the back side of the weld by electron discharge
machining. We ran a LUT scan using the asymmetric configuration shown in Figure 4, with the detection and
generation spots displaced to one side of the notch position by 1.5 mm and 6 mm respectively. Figure 8 shows a B-
scan that proceeds over a notch-free region of the weld and then encounters one end of the notch at Position 300
mm. Two indications of the notch are seen at positions between 300 and 316 mm. Using a geometric simulation
algorithm, we have identified the arrival at 2.3 µs as an LL notch-diffracted wave and the arrival at 3.4 µs as an LS
notch-diffracted wave.

27
Figure 8: B-scan of a NSWCCD Al FSW sample along the weld, including a section where an EDM notch is
present. The sample is 5.5 mm thick.

In-Line Factory Demonstration of Elongated Void Detection

We have engineered our FSW LUT system for in-line operation at Hydro Extruded Solutions, a manufacturer of
panels produced by welding individual extrusions. A partially welded panel is shown in Figure 9. Figure 10 shows
our measurement head installed onto the weld spindle. Our system acquired data at process speeds up to 50 mm/s.
We developed in-line signal-processing algorithms that analyze the data with no operator intervention.

Figure 9: Partially welded Al panel.

28
Figure 10: LUT measurement head mounted onto a FSW head for an in-line factory demonstration at Hydro
Extruded Solutions in Finspang, Sweden.

For testing purposes, a weld with elongated voids was produced by introducing shims between the faying surfaces.
In Figure 11 we show a B-scan along a 7.4-m weld with no defects (left) and with shim-related defects (right). The
regions of induced defects are clearly identifiable from the reduced amplitude of the 2L and 2S waves.

Figure 11: In-line B-scans over 7.4-m weld length.

A second sample with more controlled defect sizes was produced. Using x-ray imaging, we observed that the weld
regions near the shims contained pores with a uniform gradient in size. LUT scans of such weld regions allowed us
to test our algorithms for measuring the defect size and to determine the minimum detectable defect size. Figure 12
(upper) shows an x-ray image of a 400-mm weld region. Analysis of the x-ray image showed that the defect size at
the middle of the scan was 300 um, with larger defects to the right and smaller ones to the left. In Figure 12 (lower)
we plot the value of a signal processing parameter that was obtained from the amplitude of the 2L and 2S arrivals in
the B-scan. Note that the parameter amplitude varies monotonically with defect size. Thus, with proper calibration,

29
the defect size over a wide range can be determined from the parameter value. If we know what defect cross-section
size poses a risk to the weld’s mechanical integrity, then the data can be thresholded to indicate such a condition. In
Figure 12, for example, the threshold is set for a 300-µm defect.

Figure 12: X-ray image (upper) and defect parameter (lower) for measurement of elongated void defect size.

CONCLUSIONS
We have shown that laser ultrasonic testing is a viable technique for in-line inspection of FS welds. We have
demonstrated the detection of internal porosity, elongated voids and simulated lack of penetration. Factory testing
has demonstrated the ability to collect data at process speeds. Algorithms were developed that will enable real-time
measurement of defect size for elongated pores.

An LUT system for detecting both volumetric and LOP weld defects could combine the beam configurations of
Figures 3 and 4. While this could be done using two generation beams and two detection beams, we plan to develop
a system where a single generation beam would do double duty, for a total of three laser beams.

We also plan to investigate the detection of kissing bonds, notoriously difficult to detect because they are normally
only weak reflectors of ultrasonic waves. However, we expect that the high frequency content (and resultant small
wavelengths) of our laser-generated waves will enhance the sensitivity to kissing bonds.

REFERENCES
(1) "Welding process and its parameters - Friction Stir Welding", http://www.fswelding.com/application-of-
friction-stir-welding-in-aircraft-structures/welding-process-and-its-parameters
(2) W. M. Thomas, E. D. Nicholas, J. C. Needham, M. G. Murch, P. Temple-Smith, C. J. Dawes, 1991, “Friction
welding”, US Patent No. US5460317A, https://patents.google.com/patent/US5460317A/en
(3) D. Lammlein, N. Trepal, M. Posada, A. Floyd, 2011, "Defect significance and detection in aluminum friction
stir welds: a literature search," Technical Report NSWCCD-61-TR-2011/20, Naval Surface Warfare Center,
Carderock Division, Nov 2011
(4) V.D. Ragupathy, M.R. Bhat, M.V.N. Prasad, 2017, "An experimental study of discontinuities in friction stir
welded joints through nondestructive testing," Materials Evaluation, 75(3):406-412.

30
 (5) C. Mandache, D. Levesque, L. Dubourg, P. Gougeon, 2012, "Non-destructive detection of lack of penetration
defects in friction stir welds," Sci. Tech. Welding and Joining, 17(4):295-303.
(6) D.J. Huggett, M.W. Dewan, M.A. Wahab, A. Okeil, T.W. Liao, 2017, "Phased array ultrasonic testing for post-
weld and online detection of friction stir welding defects," Res. Nondest. Eval., 28(4):187-210.
(7) Y. Hovanski, P. Upadhyay, J. Carsley, T. Luzanski, B. Carlson, M. Eisenmenger, A. Soulami, D. Marshall, B.
Landino, S. Hartfield-Wunsch, 2015, "High-speed friction-stir welding to enable aluminum tailor-welded
blanks," JOM, The Journal of the Minerals, Metals and Materials Society, 67(5):1045-1053.
(8) C. Bird, 2004 "Ultrasonic phased array inspection technology for the evaluation of friction stir welds," Insight -
Non-Destructive Testing and Condition Monitoring, 46(1):31-36(6).
(9) Intelligent Optical Systems, “Laser Ultrasonic Receiver”, http://www.intopsys.com/laserultrasonics/lu-products-
and-services/laser-ultrasonic-receiver/
(10) C.B. Scruby and L.E. Drain, 1990, Laser Ultrasonics, Adam Hilger, Bristol.
(11) D. Lévesque, L. Dubourg, M. Lord, M. Jahazi, A. Blouin, 2008, "Defect detection and residual stress
measurement in friction stir welds using laser ultrasonics," 1st International Symposium on Laser Ultrasonics:
Science, Technology and Applications, July 16-18, Montreal, Canada.
(12) D. Lévesque, L. Dubourg, C. Mandache, S. Kruger, M. Lord, A. Merati, M. Jahazi, J.-P. Monchalin, 2008,
"Synthetic aperture focusing technique for the ultrasonic evaluation of friction stir welds," AIP Conference
Proceedings, 975:263-270.
(13) D. Lévesque, L. Dubourg, A. Blouin, 2010, "Nondestructive evaluation of friction stir welds using laser
ultrasonics," Nondestructive Testing and Evaluation.

31
 New tools in CIVA for Model Assisted Probability of Detection (MAPOD)
New Tools in CIVA for Model Assisted
to support Probability
NDE reliability of Detection (MAPOD)
studies
to Support NDE Reliability Studies2
Fabrice FOUCHER , Roman FERNANDEZ , Stéphane LEBERRE , Pierre CALMON
1 1 2

Fabrice Foucher1, Roman

1 Fernandez , Stéphane Leberre , Pierre Calmon
1 2 2
EXTENDE, Massy, France
fabrice.foucher@extende.com
1
EXTENDE, Massy, France
2 fabrice.foucher@extende.com
2CEA Commissariat à l'Energie Atomique, Gif-sur-Yvette, France
CEA Commissariat à l’Energie Atomique, Gif-sur-Yvette, France

ABSTRACT
In the context of the damage tolerance approach used to drive aircraft maintenance operations, it is
essential to demonstrate the reliability of NDE inspections in detecting structural damage. The
Probability Of Detection method, that links the probability to detect a detrimental flaw to its size is
generally used for that purpose by giving the maximum flaw size that a NDE process can miss with a
given level of probability and confidence. To be statistically valid, this approach requires a sufficient
amount of data which is often difficult (and costly) to obtain with a purely experimental approach based
on mock-up tests. Numerical simulation can be particularly useful at that stage thanks to its ability to
give a very large amount of data at a relative low cost, which constitutes the so called Model Assisted
POD approach. Recent developments have even enhanced this capacity thanks to the implementation in
CIVA simulation software of metamodels. On top of providing data for POD curves, simulations and
metamodels can be also used at the design stage to optimize inspection methods and procedures and
target a given POD for a given flaw size. It can also conduct extensive studies on parameters influence
on the result, or to achieve sensitivity analysis. This communication illustrates, with some examples
based on CIVA, how simulation can help to support NDE reliability studies in aerospace applications.

Keywords: POD, Sensitivity Analysis, Simulation, MAPOD

INTRODUCTION

The simulation plays an increasing role in NDE, allowing helping the design of inspection methods, their
qualifications or the analysis and understanding of inspection results, while reducing the number of physical mock-
ups and trials. A lot of validation efforts have been put around the CIVA software to give evidence of models
validity in order to be fully considered as a reliable element to support technical decisions and justifications [1].
In the context of NDE reliability studies, extensive parametric analyses are required in order to identify essential
parameters that can affect the NDE performance. Such studies need a large amount of data which is often difficult
and costly to obtain with a set of purely experimental results. Probability Of Detection methods, that links the
probability to detect a detrimental flaw to its size is generally used for NDE reliability evaluation in the aerospace
sector. The statistical validity of this approach is also dependent on a sufficient amount of data. Numerical
simulation tools can be particularly useful at that stage thanks to its ability to give a very large amount of data at a
relative low cost. It can also help to explore deeper and more precisely some parameters variability that can be
difficult to monitor in an experimental Design Of Experiment. In addition to classical “numerical” simulations,
“Metamodels” or “surrogate models” becomes now available, which drastically ease the capacity to generate an
even larger amount of data. For parametric and sensitivity analyses, or Model Assisted POD studies, such tools give
access to results (such as Sobol Indices, beam of POD curves, non-parametric POD curves) that simply cannot be
reached with experimental studies. This paper will illustrate the benefits of the modelling and metamodeling
approach available in the CIVA simulation platform for sensitivity and POD analyses in the context of aerospace
inspection reliability studies.

Copyright 2018. This paper is intended for the sole use of registered attendees. No part of this publication or its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.

32
Models implemented in CIVA software

Overview of the CIVA software modelling approach

The development of CIVA software started in the early 90s first for ultrasonic application. Then, this package
became commercially available and has started to be widely and even extensively used by the NDE industrial
community from the years 2000s, in different industrial sectors such as power industry, aerospace and
transportation, oil & gas, mechanical or steel industry.
The various modules of CIVA give access to different NDT methods and techniques: Ultrasonic Testing (UT),
Guided Waves Testing (GWT), Eddy Current Testing (ET), Radiographic Testing (RT) & Radiographic Computed
Tomography (CT). All these modules are available in the same environment, bringing to the users a unique NDT
oriented Graphical User Interface. The mathematical formulations used in the different modules often rely on semi-
analytical models. This approach allows solving a large range of applications while offering very competitive
calculation time compared with purely numerical methods (FEA, etc.). For instance, the UT module relies on a rays
theory geometrical approach to compute beam propagation (the so-called “pencil method”). The interaction with
discontinuities involves several models depending on the context. Some of them relies on semi-analytical or
analytical formulations, the Kirchhoff or GTD (which stands for “Geometrical Theory of Diffraction”) model can be
mentioned but other ones have also been implemented to cover several configurations. Such model can for instance
tackle the simulation of ultrasonic waves propagation in composite structures such as Carbon Fiber Reinforced
Polymer ones. The anisotropic nature of the composite medium is accounted for with a homogenized approach as
well as the change of fiber orientation due to the part curvature. In CIVA Eddy Current, the main part involves
Volume Integral and Boundary Element Methods to compute the field/Flaw perturbation phenomenon, which only
requires a numerical sampling of the flaw. The electromagnetic field induced in the work piece will be either
calculated based either on analytical expressions, modal approaches based on truncated regions or more numerical
Surface Integral equations depending on the complexity of the eddy current probe and the component geometry.
In order to continue the extension of the application fields of CIVA, it is sometimes necessary to rely on more
general numerical approaches (FEM, Finite Difference, etc.). To keep the benefits of the semi-analytical strategy,
the current trend within CIVA is to build hybrid models, a part of the computation being done by fast semi-
analytical models, another part being completed by numerical approach when necessary for the validity of the
results. For instance, such a coupling between semi-analytical and Finite Difference or more recently with Finite
Element are used to simulate a composite medium when a ply per ply approach is required instead of the
homogenized one mentioned above. For interested readers wishing to have more information on the models, the
following reference papers are available, [2] and [3] for the Ultrasonic tool, [4] for the Guided Waves module [5] for
the Eddy Current part, [6] for the radiographic one and [7] for the CT module. For more details on composite
modelling, another article can be mentioned [8].

Metamodeling approach in a few words

A metamodel or surrogate model can be defined as a “model of the model” or a “smart interpolator” which is built to
replace a physic-based model. The first step consists in computing a data base of simulation results for a given range
of multi parameters variation. From this data set is built the meta-model which allow ultra-fast exploration of the full
range of parameters variation. Thanks to the computational speed reached with metamodel, it becomes possible to
achieve statistical analysis on data such as sensitivity and POD studies. For instance, Sobol indices can be computed
from metamodel output in order to quantify the relative importance of influential parameters. Various Design of
Experiments methods can be selected to build the data base, based on a fixed number of computed configurations or
based on adaptive sampling. This can be a Full Factorial design (range of variation and number of values for each
parameter explicitly defined) but other drawing schemes based on pseudo random sequences of parameters value
can be also selected (Latin Hypercube Sampling, Sobol, Halton), which generally reaches a better metamodel
accuracy with a much smaller amount of computations. Adaptive Sampling consists in building the data-set by
estimating at each step the accuracy of the meta-model until reaching a given convergence criterion. Also, several

33
interpolators can be applied to build the metamodel from the database (Multilinear, Radial Basis function, Kriging,
etc.). Interested readers can refer to the following paper for more detailed information on the metamodels currently
implemented in the CIVA software [9].

Background in Probability of Detection and MAPOD

POD Methodology
In the aerospace sector, the damage tolerance approach is used to drive aircraft maintenance operations. This
approach requires the knowledge of the defection performance (the reliability) of the NDE process. The Probability
Of Detection method, that links the probability to detect a detrimental flaw to its size is generally used for that
purpose by giving the maximum flaw size that a NDE process can miss with a given level of probability and
confidence. The POD methodology currently adopted by the aircraft industry is described in the Military Handbook
1823A [10]. It is based on a parametric estimation of the POD following Berens models, which is adopted also in
some ASTM standards [11, 12].
Statistical analysis is defined for two different data formats: Either binary information is only provided (defect
detected or not detected), the so-called Hit Miss approach, either the signal amplitude is recorded, the so-called « â
vs a » or « Signal response » approach. The hypotheses to be satisfied as well as the statistical analysis depend on
the selected approach.

Model Assisted POD

Determination of POD curves via a purely experimental approach requires large-scale experiments performed on
representative test-blocks containing representative defects. For instance, the MH1823a states that a minimum
amount of 40 different defects location shall exist in the trial mock-ups when a Signal response analysis is
performed, while this minimum is 60 for a Hit-Miss analysis. Then, to be representative of the “real POD”, this
experiment shall “capture” the variability of the influent parameters in real inspections.
The use of numerical simulation to determine POD curves (known as MAPOD [13]) has been a subject of research
in the past years and has been used in various industrial context (Ref [14] to [19]). Recently, efforts have been done
to fix a recognized methodology and in particular let us mention here the best practice guidance and practical
recommendations published in 2016 by International Institute of Welding [20].
The methodology, as described in this document, aims at using a numerical model which simulates the results of an
inspection in order to reproduce the impact of the variability of the influential parameters on the NDE response. The
key idea consists in introducing variations in the input parameters of the model which lead to the variability on the
output of the simulation. This variability is then analyzed to calculate the POD curve. The estimation of one POD
curve by simulation requires:
1. To define a “nominal” configuration, that is all the parameters needed for simulating one inspection. From this
nominal configuration are derived the configurations which will be computed by considering the variability of some
inputted parameters.
2. To define the characteristic parameter “a” (versus which the POD (a) is calculated) and to identify and
characterize the sources of variability which will be accounted for by the POD:
 To define the “aleatory parameters” whose variability will be taken into account
 To assign a statistical distribution to these parameters
3. To sample the statistical distributions of aleatory parameters and run the corresponding simulations.
4. To compute the POD curve from the set of simulated cases.
The first advantage of using numerical simulation in a POD study is to save time and budget. A second significant
advantage is the possibility offered by simulation to obtain large sets of data, investigating the effects of the
variability of numerous influential parameters. Indeed, with simulation tools, it is generally quite fast & easy and
therefore represents a quite low cost to generate the sufficient amount of data required for POD analyses. This is
even more the case when metamodels are provided. It is also possible to directly and precisely monitor parameters

34
variation while it is difficult to control some of them in an experimental campaign (it can be for instance difficult to
precisely monitor defects orientation and to implement them in a specific zone of the mock-up). Simulation can then
explore a wider range of parameters value which can give more credit to the POD curve. The first limitation of the
MAPOD approach is related to the use of a model which always reproduces only partially the reality. Consequently,
a natural recommendation is to evaluate the accuracy of the predictions provided by the simulation code used in the
study. A second limitation is linked to the necessity to a priori identify and characterize the sources of variability on
the result of the inspection. That means to identify the influential parameters whose variability will be investigated,
and also to have a good knowledge on the statistical distributions describing the variability of those parameters. A
third limitation is that at this stage, even if this is a subject of ongoing research, the human and organizational
factors are not accounted for by this approach.
Besides the calculation of the POD curve entirely from simulation, there are various other possible uses of simulated
data in complement to experiments. Simulation gives the possibility to help determine the most influent parameters
in a given inspection, and which defect sizes correspond to the transition zone of the POD curve. As a consequence,
simulation can be used prior to an experimental campaign in order to help defining the design of experiment as well
as efficient mock-ups, which can help mastering the costs of a POD campaign. Additionally, simulation gives
insights on the results that help understanding the physical phenomena which might be also used to help designing a
more reliable inspection procedure with a given objective of POD.

Illustrative examples
In this section, two illustrative cases are shown to describe the tools implemented in CIVA for the POD simulation
and sensitivity analysis.

High Frequency Eddy Current Inspection Simulation

A first example corresponds to a High Frequency Eddy Current Inspection aiming at detecting surface defects in an
aluminum slab. The testing configuration is represented on the figure below. The component is an aluminum plate of
5 mm thickness. The inspection is made with a pencil sensor built with a 1.4 diameter coil and a ferrite core of 5mm
height working in absolute mode. The operating frequency is 1MHz. Surface defects are modelled with thin
parallelepiped notches. The configuration shown below represents a reference defect of 1mm height, an aperture of
50 microns and a length of 10mm. The figure shows the configuration as well as the simulated signal obtained
(impedance plane and X-Y curve).

Figure 1: High Frequency Eddy current model on an aluminum slab, results obtained on a reference defect

The prior list of influent parameters for this inspection included the specimen conductivity, the coil diameter, the
ferrite core permeability, the lift-off, the orientation of the probe (normal to the surface or with a tilt angle), the

35
scanning path over the defect, and the width and height of the defects. After a first impact analysis, 4 main essential
variables were kept in the design of experiment with the following range of variation:
 Lift-off [0.15mm; 0.5mm]
 Sensor orientation [-5°; +5°]
 Defect height [0.5mm; 3mm]
 Defect aperture [0.03mm; 0.07mm]
The defect length has been also of course investigated and has been even selected as the characteristic parameter, i.e.
the one which will represent the defect size “a” for the future POD analysis. A metamodel has been calculated based
on a sample of 500 computations. The total computation time was about 20 hours on a standard PC (about 2 minutes
for each case scan). The following graph (so-called “parallel plot”) represents the map of the different parameters
computed in order to build the metamodel database and an overview of the results obtained for the whole cases. The
5 first columns represent the values assigned to the variable input parameters (a Sobol sampling scheme has been
used here) and the 6th column on the right shows the corresponding results (amplitude of sensor signal generally).
This parallel plot also helps identifying at a glance for instance which cases give the lowest signals or how is
affected the output variability when you limit one or several parameters to a given range:

a) b)
Figure 2: Parallel plot of the simulations (5 first columns: parameter values; 6 th column: Signal amplitude)
a) All Cases, b) highlighted cases for a limited range of one parameter

In order to use the metamodel results and not only these 500 points of the parametric grid, the metamodel accuracy
should be first evaluated. A graph of the fit obtained between the metamodel and simulated results is provided in the
user interface. This “True vs predicted” graph is obtained with a cross validation methodology. A part of the
samples are used to build the metamodel and these results are compared to the other part of the samples. A
measurement of the error is then performed. In our example, the fit looks correct with 90% of the samples below
10% error. A higher relative error level (points in orange and red) is obtained only for low signal amplitudes.

Figure 3: Validation of the metamodel accuracy

1D or 2D parametric plots can then be obtained from the metamodel database in order to illustrate the impact of a
single of two different parameters on the output signal, while cursors give the user the ability to select which fixed
value is assigned for the other parameters.

36
 a) b)
Figure 4: Parametric analysis a) 1D (e.g. impact of sensor orientation) b) 2D (e.g. impact of defect length
(ordinate) and lift-off (abscissa) on the output signal (color map))

A sensitivity analysis using the Sobol indices (based on variance decomposition computations) can be also
performed. This can be summarized as the graph 5 below where the shared influence of each parameter on the
output variability is quantified. In this case, as expected, the lift-off influence (in purple) is really predominant
compared to the other parameters (Probe orientation in blue, defect aperture in yellow, and defect height in orange,
the defect length has been set constant for this analysis here):

Figure 5: Sensitivity analysis with Sobol indices

This sensitivity analysis uses the statistical distribution assumed for each parameter, which was in our example:

 Lift-off: Exponential distribution with a minimum value of 0.15mm,

a mean value of 0.2mm and a max value of 0.5mm

 Probe orientation: Normal distribution with a 0° mean value, a

standard deviation of 2°, and a minimum and maximum limit of -5°
and +5°

 Defect aperture: Uniform distribution between 0.03mm and 0.07mm

37
  Defect height: Uniform distribution between 0.5mm and 3mm

A POD analysis can then be launched from the metamodel database. The following â vs a graph represents the
obtained data for a set of 20 defect lengths taken from 0.2mm to 3.5mm and with 20 different results obtained for
each defect length. A Signal Response analysis is performed. The detection threshold is defined (set at -12dB versus
the response of an “infinite” length reference defect shown earlier) as well as potentially two other thresholds
representing the saturation threshold (right censoring) and the noise level (left-censoring).

(a)

(b)

Figure 6: “â vs a” graphs, linear regression curves and statistical tests validating the assumption of the Signal
response analysis. a) log/log transform used to represent the data for the linear regression curve
b) log/box-cox transform used

Some coefficients and statistical tests are provided to help validating the different hypotheses that the data should
verify in Signal Response analysis: linearity (the coefficient of determination, closer as possible to 1, a lowest
possible level of quadratic errors), homoscedasticity (variance stability for the whole range of defect lengths) and
normality of the error. These statistical tests are here evaluated both when applied a log/log transform (figure 6a)
and when is applied a second transformation called Box-Cox (figure 6b). Once the model is selected, a POD curve
can be derived and displayed. The a90/95 is here obtained for a defect size very close to the a90 value, illustrating the
really thin confidence bound achieved here. Indeed, as discussed before, in a MAPOD study it may be difficult to
have a precise knowledge on the variability of the influential parameters of the inspection. In the MAPOD
methodology this variability is an input of the process and hypotheses are made on the statistical distributions
assigned to the aleatory parameters. Thanks to metamodel and their ability to provide ultrafast computations, it is
possible in CIVA to estimate the sensitivity of the POD estimation to these hypotheses. Thus has been implemented
the possibility to introduce a variability on the statistical distributions and calculate “beams of PODs” following the
approach proposed by N. Dominguez [21]. The user gives a level of confidence for the input statistical distribution
(here, the mean value of the lift-off as well as the standard deviation of the probe orientation distribution were
considered as uncertain). Then, a Monte-Carlo sampling is performed on these distributions parameters, which lead
to a new POD curve for each test. In the case presented here, a beam of 100 POD curves (representing a set of
40 000 data points!) is shown on Figure 7 to illustrate this functionality. The 100 PODs have been computed and
displayed in only 8 seconds. The variations on the parameters have been chosen here in a somehow arbitrary way.

38
 Figure 7: Simulated POD curve (above) then Beam of POD curves calculated
to estimate the reliability of the simulated POD

Engine disk UT inspection

A second example illustrates an immersion ultrasonic inspection of an engine disk made of nickel super alloy.
Typical defects are represented with Flat Bottom Holes. A Pulse-Echo single element probe is used, operating at
5MHz, working at normal incidence and focused with a 225mm curvature radius. The nominal testing configuration
modelled in the CIVA software is represented below as well as the obtained A-Scan and B-Scan.

Figure 8: Engine Disk UT inspection simulation

The list of influential parameters has been established at the following one with an associated range of variation:
 Probe signal center frequency [4.8MHz; 5.2MHz]
 Incidence angle [-3°; 3°]
 Water path [75mm; 85mm]
 Attenuation coefficient at 5MHz [40dB/m; 60dB/m]
 Defect orientation vs outer surface [-5°; +5°]

39
A metamodel has been computed based on a sample of 800 parametric variations to cover the above range. It took
only 2 hours to compute the full set of simulations. As for the ET example, a first evaluation of the metamodel fit
has to be performed and is shown below as well as a parallel plot of the performed calculations. The True vs
Predicted fit is here very good with 98% of the data below 10% relative error on the signal amplitude.

Figure 9: Validation of the metamodel accuracy for the UT case

A sensitivity analysis is then performed and shows that the flaw orientation (in red) is evaluated as the most influent
parameter with 50% of relative influence on the output variability. But some other parameters show also a non-
negligible impact (incidence angle in light blue, center frequency in dark blue and attenuation level in orange).

Figure 10: Sensitivity analysis with Sobol indices for the UT inspection case

A POD curve is then derived from these results. The following one is plotted by selecting 30 defect sizes (FBH
diameter) linearly distributed between 0.5 and 3.5mm and 40 results per defect size, all data being collected in real
time from the metamodel database. The detection threshold is set at -6dB versus the amplitude level recorded for a
reference FBH of 3mm in a planar block. As it can be observed on figure 11, in this case the statistical assumptions
are not valid for a signal response analysis regarding the normality of the scattering and the homoscedasticity.

Figure 11: “â vs a” graphs, linear regression curves and statistical tests not fulfilled here with a Signal
response analysis leading to selecting a Hit-Miss POD.

40
However, a Hit/Miss analysis can be performed in the same way. Indeed â vs a analysis is, when possible, preferred
in experimental studies because the amplitude information can improve a statistical analysis limited by the number
of available data. Simulation offering the possibility to obtain enough data, Hit miss analysis is generally preferred
when using simulation [IIW]. In Figure 12 is shown the Hit miss POD curve obtained on this illustrative case
performed (using logit link), the obtained POD curve gives a a90/95 value of 1.9mm.
Let us also mention the possibility offered by CIVA to calculate “non-parametric” POD curves. Such option allows
overcoming the usual hypotheses of the parametric models [22]. Here, we directly calculate the Hit/Miss ratio for
each defect size. Again, such method is relevant here only because, thanks to the metamodel, we can provide at low
cost a tremendous number of data. The “non-parametric” POD can be used in order to assess the hit miss POD
curve. In the present case Hit/Miss and non-parametric POD curves are superimposed demonstrating the validity of
Hit/Miss assumptions.

Figure 12: Simulated POD curve (above) with a Hit-Miss analysis then compared to a non-parametric curve
(below) to estimate the reliability of the simulated POD

As for the previous case, a beam of POD might be calculated to evaluate the sensitivity to the statistical distribution.
In the same way it is also possible to estimate if a better mastering of the inspections parameter could help reaching
a “better POD”. It is just a question of resampling the data which is easy and a kind of real time operation thanks to
metamodels. For instance, below is displayed the POD curve if no variability is assumed on the inspection
conditions (water path, incidence angle) but only on specimen and defect parameters. The POD reaches thena a 90/95
of 1.8mm for this detection threshold (see figure 13).

Figure 13: Obtained POD curve with other hypotheses on the inspection parameters variability (fixed values)

41
CONCLUSION

Simulation tools gathered in the CIVA platform provide an efficient solution to support NDE reliability study. In
particular, the recent introduction of metamodels offers new possibilities such as real-time resampling of the data
which is particularly useful to perform sensitivity analysis (Sobol indices evaluation) or advanced POD analysis
(assessment of statistical models, beam of POD curves, or even non parametric POD curves, etc.). Simulation can
then be used either to help defining the design of Experiment for an experimental campaign, to compute directly the
POD curves or to give insights for inspection method optimization to reach a targeted value of POD.

REFERENCES

(1) F. Foucher, S. Lonné, G. Toullelan, S. Mahaut, S. Chatillon, 2018, An overwiew of validation campaigns of the
CIVA simulation software, ECNDT.
(2) S. Mahaut, S. Chatillon, M. Darmon, N. Leymarie and R. Raillon, 2009, An overview of UT beam propagation
and flaw scattering models in CIVA, QNDE.
(3) M. Darmon, S. Chatillon, 2013, Main Features of a Complete Ultrasonic Measurement Model: Formal Aspects
of Modeling of Both Transducers Radiation and Ultrasonic Flaws Responses, Open Journal of Acoustics,
Vol.3 No.3A, http://file.scirp.org/Html/8-1610079_36873.htm#txtF2.
(4) V. Baronian, A. Lhémery, K. Jezzine, 2010, Hybrid SAFE/FE simulation of inspections of elastic waveguides
containing several local discontinuities, QNDE.
(5) G. Pichenot et al., 2005, Development of a 3D electromagnetic model for eddy current tubing inspection:
Application to steam generator tubing, QNDE.
(6) J.Tabary, P. Hugonnard, A.Schumm, R. Fernandez, 2008, Simulation studies of radiographic inspections with
Civa, WCNDT.
(7) R. Fernandez, S.A. Legoupil, M. Costin, A. Leveque, 2012, CIVA Computed Tomography Modeling, WCNDT.
(8) K. Jezzine et al, 2017, Modeling approaches for the simulation of ultrasonic inspections of anisotropic
composite structures in the CIVA software platform, QNDE.
(9) R. Miorelli et al, Database generation and exploitation for efficient and intensive simulation studies, AIP
Conference Proceedings, Volume 1706 (2016)
(10) USA Department of Defense Handbook, 2009, MIL-HDBK-1823-A, NDE system reliability assessment .
(11) ASTM, 2012, E2862-12: Standard practice for probability of detection analysis for hit/miss data .
(12) ASTM, 2012, E3023-15: Standard practice for probability of detection analysis for â versus a data .
(13) B. Thompson et al, 2009, Recent Advances in Model-Assisted Probability of Detection, European-American
Workshop on reliability in NDE.
(14) Aldrin, J.C., Knopp, J.S., Lindgren, E.A., Jata, K.V., Model-assisted probability of detection evaluation for
eddy current inspection of fastener sites, AIP Proceedings, Volume 1096, 2009, Pages 1784-1791
(15) F. Jenson et al, 2010, Simulation supported POD: methodology and HFET validation case, QNDE
(16) Dominguez, N., Feuillard, V., Jenson, F., Willaume P., 2012, Simulation assisted POD of a Phased Array
Ultrasonic Inspection in Manufacturing, Rev. of Prog. In QNDE, Vol 31 (2012) pages 1765-1772
(17) B. Chapuis et al, 2014, Simulation supported POD curves for automated UT of pipeline girth welds, welding in
the world, V58, 433-441
(18) M. Pavlovic et al, 2016, Reliability Analysis of the Ultrasonic Inspection System for the Inspection of Hollow
Railway Axles, WCNDT.
(19) G. Ribay et al, 2016, Model-based POD study of manual ultrasound inspection and sensitivity analysis using
metamodel, AIP Conf. Proc. 1706 (2016)
(18) N. Dominguez et al, 2014, POD Evaluation using simulation: PAUT case on a complex geometry part, AIP
Conf. Proc. 1581, 2031

42
(20a) B. Chapuis, P. Calmon, F. Jenson, 2016, Best practices for the use of Simulation in POD Curves Estimation,
IIW Collection.
(20b) P. Calmon and al, 2016, The use of simulation in POD curves estimation: An overview of the IIW best
practices proposal, WCNDT
(21) N. Dominguez et al, 2013, A new approach of confidence in POD determination using simulation, QNDE.
(22) Spencer, F.W., Nonparametric Pod Estimation for Hit/miss Data: a Goodness of Fit Comparison for
parametric Models, Review of Quantitative Nondestructive Evaluation, AIP Conference Proceedings, 2011

43
 Flaw Detection Capabilities in Aerospace with Eddy Current Array
Flaw Detection Capabilities in Aerospace with Eddy Current Array Technology
Technology
Jesse Herrin,
Jesse HERRIN, Nicholas
Nicholas Cardillo, Steve
CARDILLO, SteveTimm,
TIMM,and
andTom
TomRohlfing
ROHLFING

Zetec
Zetec
8226 Bracken
8226 Bracken Place
Place SE
SE
Suite 100
Suite 100
Snoqualmie, WA
Snoqualmie, WA 98065
98065
(425) 974-2700;
(425) faxfax
974-2700; (425) 974-2701;
(425) email
974-2701; jherrin@zetec.com
jherrin@zetec.com

ABSTRACT
Eddy current array (ECA) has enhanced the practice of flaw detection in the aerospace industry through expanded
coverage areas, reduced inspection time, improved detectability, repeatability, and data recording. However, because
ECA detectability is inversely proportional to scan speed, the signal-to-noise ratio (SNR) of today’s ECA techniques
limits their ability to detect the very small flaws and loss of material that can occur in many aerospace inspections
such as friction stir welds and multi-layer structures susceptible to corrosion and fatigue cracking. Minimum flaw
size and required data density are the main factors affecting scan speed when using eddy current array. As speed
increases, the SNR decreases, therefore the minimum flaw size becomes larger. Zetec introduces exceptional SNR in
an eddy current array instrument, eliminating the need to sacrifice scan speed for superior flaw detectability.

Keywords: eddy, current, array, EC, ECA, SNR, POD

INTRODUCTION
The introduction of portable ECA instruments was a leap ahead for eddy current (EC) inspections of critical
components in numerous industries. Over the last 5 years the further advancement of ECA electronic design has
resulted in improved SNR of ECA techniques which, in turn, has resulted in improved probability of detection
(POD). However, despite these advancements and because POD is inversely proportional to ECA scan speed,
inspections that require a high POD remain slow. In the case of exceptionally small flaws and loss of material that
can occur in friction stir welding (FSW) [1][2][3] and multi-layer structures susceptible to fatigue cracking and
corrosion, scan speed is compromised to achieve the required POD.

An array probe has a high number of EC coils placed near each other typically in two or three rows that rapidly get
energized across the array to create a full coverage scan width. ECA has enhanced the practice of flaw detection in
the aerospace industry through expanded coverage areas, reduced inspection time, improved detectability,
repeatability, and data recording. ECA probes can be driven in absolute mode as single coils or in reflection (driver-
pickup) in coil pairs. Penetration depth in the test material for eddy current is up to 6.3 mm depending on coil type.

The theoretical maximum scan speed for an ECA probe is the time-slot rate (TSR) divided by the product of the
spatial resolution (SPR) and number of time-slots (TS). A common SPR is 2 samples/mm. The theoretical maximum
scan speed is not practical for ECA applications that demand high signal quality because the processing time of the
EC signal is not long enough to reject noise that is present. Practical sample rate (SR) for ECA applications is much
lower than the theoretical maximum to allow enough processing time to produce lower noise signal. This makes
high SR superfluous for ECA.

The SNR of the array probe data is an inverse function of the SR. The maximum SR will result in the worst SNR.
Slowing down the SR will achieve better SNR by stretching the periods of the TS to average more samples. This
will improve the detectability of small flaws. The SNR will improve by 3 dB each time that the sampling rate is cut

Copyright 2018. This paper is intended for the sole use of registered attendees. No part of this publication or its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.

44
in half. For example, a scan speed of 200 mm/sec will have an SNR that is 3 dB better than a 400 mm/sec scan
speed. Slowing the scan speed down to 100 mm/sec (one-fourth) will produce an SNR that is better by 6 dB.

ECA is ideal for large-area applications such as FSW, where weld seams up to 36 meters need to be scanned with
full coverage to have high POD. Chemically-milled airplane skins are another application where there are many
meters of transitions where cracks can occur. The C-Scan capability of ECA instruments provides a visual
confirmation of full scan coverage, which improves POD. The C-Scan benefits can also be realized in bolt hole
scanning where flaws can occur in different layers. The C-Scan allows to determine exactly which bolt hole layer
the flaw is present.

In ECA applications, the fastest possible scan speed that will maintain the required POD is desired to complete the
inspection in a cost-effective manner compared to other nondestructive testing (NDT) methods. An ideal ECA
solution, consisting of instrument and probe, is one that produces the best SNR to meet these inspections needs.

SURFACE CRACKS IN ALUMINUM FRICTION STIR WELDING FOR ROCKET

COMPONENTS
NASA’s Space Launch System (SLS) utilizes FSW in many of its components due to its ability to produce high-
strength bonds that are virtually free of defects. The weld must be able to withstand extreme conditions: 5.5 million
lb liftoff weight, 0 to 17,500 mph in 8 minutes, 5,000 °F component temperatures, and -400 °F external
temperatures. Rocket scientist Wernher von Braun stated during the Saturn V development in 1966 that “A lifetime
of rocketry has convinced me that welding is one of the most critical aspects of this whole job” [4]. The extreme
conditions make NDT a critical part of design validation.

FSW can be inspected using ultrasound (UT), x-ray, and penetrant testing (PT). However, surface breaking cracks
can be difficult for UT to detect and size. X-ray is time consuming and it can be difficult to get proper access for
detectors. PT requires the handling of chemicals as well as time to prepare the surface and clean.

ECA is a fast and easy technique for detecting the small flaws characteristic of FSW applications and ECA can
detect and size surface cracks in aluminum FSW. The small probe and portable nature of the instrument allows for
fast and easy access to all FSW seams. The scan does not require any surface preparation, nor does it require the
handling of chemicals.

The FSW standard in Figure 1 is aluminum with six notches in two sets of three. The first set from the left (1), (2),
(3) are all axial in orientation. The second set of three from the left (4), (5), (6) are 30° off axial, 45° off axial, and
transverse.

Figure 1. FSW standard six notches: from left to right (1) Axial, (2) Axial, (3) Axial, (4) 30° off Axial, (5) 45°
off Axial, (6) Transverse

45
The scan in Figure 2 is from a MIZ-200 EC instrument with an ECA probe in a reflection (driver-pickup) mode. The
axial channel is set to 400 kHz. The axial notches (1), (2), (3) are all detected in the axial channel, but at different
voltages based on the height of the signals. This can be used with phase angle information to get the depth of the
flaws.

Figure 2. ECA axial 400 kHz notches (1), (2), (3)

The scan in Figure 3 is from a MIZ-200 EC instrument with an ECA probe in a reflection (driver-pickup) mode. The
axial channel on the left is set to 400 kHz and the transverse channel on the right is set to 400 kHz.

Figure 3. ECA axial 400 kHz notches (4), (5), (6), transverse 400 kHz notches (4), (5), (6)

The notches (4), (5), (6) are not all detected in the axial channel. Specifically, notch (6) has a minimal signal. The
notches (4), (5), (6) are all detected in the transverse channel. Notch (6) can be characterized as a crack-like flaw in
the transverse direction since it showed up on the transverse channel, but not on the axial channel. Notches (4) and
(5) show up on both the axial and transverse channels, but the heights of the signal peaks are opposite which
indicates these flaws are also crack-like. In contrast, a volumetric flaw would have the same height in both the axial
and transverse channel. This is the process used to characterize flaws using ECA.

The small size of the cracks in the FSW example show that ECA is a fast and easy method to detect small flaws that
occur in FSW. This resolves the gap in UT detection capability. In addition, ECA does not get a noise or shadow

46
effect from the FSW metal crystal structure, which can hide flaws using UT. The small probe and portable nature of
ECA instruments, such as the Zetec MIZ-21C, allows for fast and easy access to all FSW seems, which is difficult
for x-ray. The ECA scan does not require any surface preparation or handling of chemicals that PT uses.

TRANSITION CRACKS IN CHEMICAL MILLED ALUMINUM AIRPLANE SKIN

Cracks occur at the edge of the transition from thinner (1.07 mm) to thicker (2.08 mm) skin. The ideal NDT
inspection is performed from the external surface to detect flaws on the far side of the airplane skin. The location of
these cracks in the transition can cause them to be hid-den to UT. X-ray is not possible in all locations due to
obstructions. PT does not work from the external surface since the flaws are not surface breaking. This is an ideal
application for ECA.

The chemically-milled airplane skin in Figure 4 is 2.08 mm thick aluminum with three flaws in two 1.07 mm thick
pockets. The first pocket on the left has flaw (1) and (2). The second pocket on the right has flaw (3).

Figure 4. Chem mill airplane skin ext side: (#1) pocket with 2 flaws (1) and (2), (#2) pocket with 1 flaw (3)

The chemically-milled pockets can be seen in Figure 5 which is the reverse side. The three flaws are in the
transitions colored black for visibility of the geometry.

Figure 5. Chem mill airplane skin int side: (#1) pocket with 2 flaws right side, (#2) pocket with 1 flaw left side

47
The scan in Figure 6 is from a MIZ-200 EC instrument with an ECA probe in a reflection (driver-pickup) mode. The
axial channel and transverse channels are set to 50 kHz. The scan was performed from the external surface from
bottom to top across the thickness transitions. The axial channel shows flaw (1) and the transverse channel does not,
which characterizes flaw (1) as crack-like in the axial scan direction. In addition, the thickness changes can be
detected using ECA in both the axial and transverse channel. This improves POD of the layer in which the flaw is
located.

Figure 6. ECA axial 50 kHz, transverse 50 kHz flaw (1)

The scan in Figure 7 is using the same setup with the same results for flaw (2). Additionally, the bolt holes are
visible as multiple red lift off signals near the flaw.

Figure 7. ECA axial 50 kHz, transverse 50 kHz flaw (2)

48
The scan in Figure 8 is using the same setup but the scan direction is from left to right. Flaw (3) shows up in the
axial channel and not the transverse channel because of the 90° scan orientation change of the ECA probe for flaw
(3).

Figure 8. ECA axial 50 kHz, transverse 50 kHz flaw (3)

The capabilities of ECA allow for the detection of crack-like flaws in transitions between different thick materials,
which can be a problem for UT. ECA can detect sub-surface defects which allows for access that x-ray and PT can’t
perform.

FLAWS IN MULTILAYER ALUMINUM AIRPLANE BOLT HOLES

It is not easy to determine the layer in which a flaw occurs in a multilayer bolt hole. This is particularly true when
the flaw is very close to the transition between layers. The current method uses a rotating scanner with a rotary
probe, but the depth of the flaw can require use of a marker or tape on the probe to indicate depth.

The three-layer NRK standard in Figure 9 is 6.35 mm, 3.18 mm, and 7.92 mm thick aluminum with a total thickness
of 17.5 mm. In Figure 10, flaw (6) is in layer 01, two flaws (5) in layer 05, and no flaws in layer 03.

Figure 9. NRK standard three plates: (01) 6.35 mm (.250 in), (03) 3.18 mm (.125 in), (05) 7.92 mm (.312 in)

49
 Figure 10. NRK standard three notches (6) plate 01, (5) plate 05, (5) plate 05

The scan in Figure 11 is from a MIZ-21C EC instrument with rotating scanner and rotary probe in a reflection
(driver-pickup) mode. The first channel is set to 300 kHz frequency. The C-Scan has a lowpass and two highpass
filters.

Figure 11. Bolt hole scan 300 kHz: notches (5), (5), (6) from left to right

The first channel is used to detect the flaws, which can clearly be seen in both the waterfall and C-Scan. The
orientation of the flaws in the bolt hole correlate to the vertical position of the flaw signal on the waterfall, where
from top to bottom is equal to 0° to 360°. The C-Scan also shows position of the flaw. This allows for depth and
angle positioning of the flaws based off an index point.

50
The scan in Figure 12 is using a MIZ-21C EC instrument with rotating scanner and rotary probe in a reflection
(driver-pickup) mode. The second channel is set to 100 kHz frequency, and the C-Scan has a highpass filter.

Figure 12. Bolt hole scan 100 kHz: layers (05), (03), (01) from left to right, and notches (5), (5), (6) from left to
right

The second channel is used to detect the layers, which can clearly be seen in both the waterfall and C-Scan with the
two flaws (5) showing in layer (05), no flaws in layer (03), and flaw (6) in layer (01). The orientation of the flaws in
the bolt hole correlate to the vertical position of the flaw signal on the waterfall, where from top to bottom is equal
to 0° to 360°. The C-Scan also shows position of the flaw. This allows for full characterization of the flaws in the
multilayer bolt hole and yields a higher POD.

The C-scan advancements of ECA applied to bolt hole scanning improves flaw detection capability compared to
traditional rotating scanner inspections. Using an EC instrument with high SNR further improves the POD.

CONCLUSIONS
Based on the results of the case studies presented in this paper, the following conclusions can be drawn regarding
flaw detection in aerospace applications using ECA technology:
 Surface breaking flaws can efficiently be found in FSW using an ECA probe and a handheld ECA
instrument in comparison to existing techniques
 Transition cracks in chemically-milled aluminum skins can be detected using an ECA probe and a handheld
ECA instrument in comparison to existing techniques
 During multilayer aluminum bolt hole inspections, it can be determined in which layer flaws exist by using
C-Scan and dual frequencies
 A high-quality SNR ECA instrument can increase the scan speed ability while maintaining the required
POD

51
REFERENCES
[1] N.T. Kumbhar, G.K. Dey, K. Bhanumurthy, “Friction Stir Welding of Aluminium Alloys”, BARC Newsletter
321, 11-17, (2011).
[2] A. Lamarre, O. Dupuis, M. Moles, “Complete Inspection of Friction Stir Welds in Aluminum using Ultrasonic
and Eddy Current Arrays”, (January 2007).
[3] R. J. Ding, “Friction Stir Welding: Standards & specifications in today’s U.S. manufacturing and fabrication”,
NASA, (2017).
[4] M. Burkey, “A (much) Closer Look at How We Build SLS”, Rocketology: NASA’s Space Launch System,
(June 2016).

52
 Quality
QualityClassification ofAdhesive
Classification of AdhesiveBonds
Bondsin in Composite
Composite Structures
Structures by Single-sided
by Single-sided Air-
Air-coupled
coupled Ultrasonic TestingUsing
Ultrasonic Testing Using Linear
Linear Phased
Phased Array
Array Probes
Probes

Mario Kiel11, Stefan Scheunemann11, Maria Grötzner11, Manuel Lucas22, and Andreas Mück22
Mario Kiel , Stefan Scheunemann , Maria Grötzner , Manuel Lucas , and Andreas Mück
Forschungszentrum Ultraschall [Research Center for Ultrasonics]
1
Forschungszentrum Ultraschall
1
Köthener [Research
Straße 33aCenter for Ultrasonics]
Köthener Straße
06118 Halle, 33a
Germany
06118 Halle, Germany
(+49) 345-44 58 39 15; fax (+49) 345-44 58 39 19; email mario.kiel@fz-u.de.de
(+49) 345-44 58 39 15; fax (+49) 345-44 58 39 19; email mario.kiel@fz-u.de.de
2SONOTEC Ultraschallsensorik Halle
2
SONOTEC Ultraschallsensorik Halle
Nauendorfer Straße 22
Nauendorfer Straße
06112 Halle, Germany
06112 Halle, Germany
(+49) 345-133
(+49) 345-133 17
17 0;
0; fax (+49) 345-133
fax (+49) 345-133 17
17 99;
99; email
email mlucas@sonotecusa.com
mlucas@sonotecusa.com

ABSTRACT
Adhesive bonding is widely used in composite material technology. The mechanical performance and durability of
such composite structures strongly depends on the quality of the bond. Air-coupled ultrasonic testing is a promising
nondestructive method for the characterization of these bonds. A main advantage is the contactless measurement
without the use of oil or water using high-power ultrasonic pulses in transmission mode. The absence of glue acts as
an additional interface within the testing object which strongly reduces the transmitted ultrasound wave. Glue
application errors can be detected easily. Additional characterizations such as size determination are possible as
well. In contrast to common transmission mode testing we present a single-sided pitch-catch approach for testing
adhesive bonds between metal sheets and carbon fiber reinforced plastics (CFRP). New approaches of data analysis
are possible by using broadband transducers. A direct characterization of different types of flaw indications becomes
possible. Broadband transducers can also be used for additional characterizations in transmission mode. Due to its
wavelength dispersive properties the quality of the adhesive bonding can be determined within one single
measurement. This method will be demonstrated by a weak bond test within a CFRP structure. The sound velocity
of Lamb waves strongly depends on the material and on the sample thickness. To transmit these waves efficiently
the angle of incidence (AOI) of the ultrasonic wave has to be adjusted. We present a new generation of ultrasonic
linear phased array probes allowing the AOI to be adjusted electronically in order to achieve maximum signal
intensity.

Keywords: ultrasonic testing, air-coupled, Lamb wave, adhesive bonds, phased-array probes

INTRODUCTION
Air-coupled ultrasonic testing is promising technique for non-destructive testing of light-weight materials. It is
already widely used. With new digital filtering, faster data acquisition, high performance PCs, etc. many new
applications arise and the number of potential areas of application increases every day. A main aspect is a contact-
free, radiation-free, and fast measurement.
In a standard setup the through transmission is used to gain information of the sample condition and allow to
characterize flaws and inhomogeneities. Transmission measurements need access to both sides. For more
sophisticated structures of the components this access is not possible. Thus, a single-sided approach is necessary.
We present results for single-sided testing on adhesive bonds using Lamb waves. Broad-band transducers allow a
characterization of the bond quality. Additionally, we present new multi-element transducers for phased-array
application to tune the angle of incidence of the incoming ultrasonic wave for Lamb wave excitation.

Copyright 2018. This paper is intended for the sole use of registered attendees. No part of this publication or its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.

53
EXPERIMENTAL SETUP
The data presented here has been performed using a SonoAir air-coupled UT system (SONOTEC). It carries the part
of signal generation, data acquisition, processing and visualization. The ultrasonic transducers are mainly home-
build prototypes from the research center for ultrasonics. They enable a larger band-width compared to conventional
transducers found on the market.

RESULTS
Single-sided Air-coupled Ultrasonic Testing using Lamb Waves
The method used in this paper is the excitation and analysis of the well-known Lamb waves in flat samples. The
measurement setup is shown in Fig. 1. The transducers for transmitting and receiving are arranged in a pitch-catch
configuration with a certain angle (ϕt and ϕr) between its acoustic axis and the surface normal. As depicted in the
cartoon, this angle depends on the sound velocity of excited Lamb wave and is adjusted in the way that the projected
wave length in air fits perfectly to the wave length of the horizontally propagating Lamb wave in the plate.

Figure 1: Setup for air-coupled Lamb wave measurements.

The Lamb wave reemits an ultrasonic wave which is detected by the receiver. Any flaw, discontinuity or change of
physical properties of the plate within the ultrasonic path between transmitter and receiver will influence the signal,
i.e. the signal intensity or the time-of-flight. These changes are analyzed and allow conclusions about the sample
properties.

Testing of Adhesive Bonds

To simulate a typical joint between an aluminum and a CFRP plates, epoxy as adhesive was used. Two different
types of bond were prepared, a normal one and a weak one. The testing setup for single-sided testing is shown in
Fig. 2 a) and b), respectively. Note that both plate materials have different speed of sound of the generated Lamb
wave. This causes a different angle of the transducers with respect to the surface normal. This is not considered in
both images.

(a) (b)
Figure 2: Setup for single-sided air-coupled ultrasonic testing of adhesive bonds between aluminum and
CFRP plates. (a) Testing on the aluminum and (b) on the CFRP side.

For a better understanding the sample was investigated in through transmission first. The results are shown in the C-
scan image in Fig 3. A main advantage of air-coupled UT in transmission is to reveal any type of delamination or

54
loss of contact in joints or composite materials. The additional interfaces lead to a huge intensity drop. This is visible
in the C-scan by a blue color code indicating a low signal intensity. For most regions of the sample it is expected
that no signal observed due to the lack of adhesive (compare Fig. 2). However, only two of four regions reveal a
high UT signal as indicated by solid lines. The other two regions (dashed lines), where the weak bonds were
prepared exhibit almost no signal. This already indicates a delamination-like behavior.
Performing single-sided measurements lead to C-scans as shown in the other two images of Fig. 3. In these
measurements the UT signal is large in the plate regions without adhesive and vice versa. Having this in mind, we
can compare the transmission results with the results obtained on the aluminum side. This leads to an almost perfect
agreement of both measurements. Note that the features in the single-sided measurement appear elongated in the
vertical direction, which is due to the lateral separation of both transducers in this direction (compare Fig. 2). In
contrast to that result the measurement at the CFRP side reveal all four adhesive regions. This leads to different
interesting conclusions. First, the single-sided approach is able to characterize adhesive bonds. Second, the adhesive
is still sticking at the CFRP side. And third, the upper right bond reveals an elongated signature compared to the
transmission C-scan. This indicates that we observe a kissing bond like behavior as well.

Figure 3: C-scan images for transmission and single-sided testing.

Further tests have been performed using so-called broad-band transducers. This is something special, because air-
coupled UT usually uses transducers with an extremely small band width to gain maximum ultrasonic power due to
the intrinsic intensity loss at each interface within the ultrasonic path (transducer – air – sample – air – transducer).
We were able to reach almost 50% band width at a center frequency of 350 kHz. This allows to apply frequency-
coded signals, e.g. a simple frequency sweep. Thus, the ultrasonic pulse carries frequency-dependent information of
the sample. In the post-processing we can extract these information using frequency filters. This is shown in Fig. 4
for the low and high frequency fraction of the signal, respectively. The measurement has been performed on the CFP
side. Both C-scan images are almost similar. Note that the intensity loss of the upper two adhesive regions (dashed
lines) is smaller for the high frequency fraction in comparison to the low frequency image as visible by the lighter
blue color code. To enhance these small differences, the right image shows the ratio of both measurements. It is
normalized at regions without adhesive. A ratio greater than one (green color) indicates the normal bond and a ratio
of less than one (red color) the weak bond.

55
Figure 4: Single-sided testing using broad-band transducers on the CFRP side. Analysis of the high
frequency, low frequency fractions of the signal and the normalized intensity ratio.

Angle Adjustment Using Linear Phased-Array Probes

As already discussed within the scope of Fig. 1 the angle of the transducers with respect to the sample surface is a
crucial parameter for the signal quality. The angle always depends on the speed of sound of the excited Lamb wave
within the plate which mainly depends on the thickness of the plate. But other physical and technical parameters also
influence the speed of sound. Thus, it is recommended to tune the angle of incidence for in-situ signal optimization.
In contrast to a mechanical setting techniques an electronic solution would be able to realize a fast feedback loop.
Our approach is a multi-element transducer in terms of phased array.
The principles of the passed-array technique are nicely summarized in Ref. (1). In order to determine the best
dimensions of the array geometry four main aspects should be realized (Ref. 2):
 The element width lx should be smaller than half of the wavelength: lx < λ/2
 The pitch distance p smaller than 0.67 of the wavelength: p < 0.67 λ
 The length of the elements ly approx. 0.7 to 1 times the width of the whole transducer array: l y~B, or
 Length divided by pitch greater than 10: ly/p > 10
All those considerations are perfectly proofed in conventional UT as long as frequencies above 1 MHz and the speed
of sound is large enough. In air coupled UT the situation is totally opposite. Thus, phased-array applications in air
are not as popular.
For visualization of the obvious problems we did FEM simulation in a one-dimensional model of a linear array with
four elements. To realize different angles of incidence the time delay between the excitation of the transducer
elements have to be adjusted in a proper way. Figure 5 shows the resulting sound pressure fields for different angles
between 5 and 45°, respectively. The gray rectangles indicate the size and the position of the active elements.
In Fig. 5 (a) an angle of 5° is shown. A wave front is formed which nicely fits to the desired angle. In contrast to that
the wave front for larger angles is more and more split into the fractions of the single elements which form ring like
waves. This is far off the intended behavior.

56
 (a) (b)

(c) (d)
Figure 5: FEM simulation of the ultrasonic wave in air caused by a four-element air-coupled transducer. The
time delays are adjusted to obtain a 5°, 10°, 20° and 45° orientation of the wave front, respectively.

For the proof of principal, we build up a prototype transducer with the given geometry of the FEM model. Figure 6
shows the resulting wave. Note that in this image the transducer elements are located at the top. The black line
indicates the tilt of the propagating sound wave. However, the wave front (red and blue lines) is still oriented
horizontally. This is in good agreement with the theoretical result in Fig. 5 (b) where the tilt is mainly visible due to
the trend of the intensity maxima instead of the wave front itself. Such wave front forming is not applicable for
Lamb wave excitation.

Figure 6: Measurement of the sound pressure field using a transducer design discussed above. The time delay
was adjusted to realize a tilt angle of 10°.

57
In a second step we optimized the geometry of the transducer elements. Since the UT electronics has four output
channels we decided to develop a multi-element transducer with 12 elements but excite always three of them
simultaneously. The exact geometry will not be further discussed due reasons of ongoing research. Figure 7 show
the results for the propagating wave front in the FEM simulation. At the beginning we observe a nicely tilted wave
front which will be able to excite a Lamb wave as depicted in Fig. 7 (b) and (c). For long time delays an interference
of circular shaped waves occur. Those interferences have less amplitudes at the plate surface. Thus, they should not
influence the Lamb wave excitation significantly. To build up a transducer using this proposed design is current
work in progress.

(a) (b)

(c) (d)

Figure 7: FEM simulation of the ultrasonic wave in air caused by a new element design using four channels as
well. The images show different time delays.

REFERENCES
(1) Schmerr, Lester W.: Fundamentals of Ultrasonic Phased Arrays. Springer International
Publishing, 2015. http://dx.doi.org/10.1007/978-3-319-07272-2.
http://dx.doi.org/10.1007/978-3-319-07272-2
(2) Olympus: Introduction to Phased Array Ultrasonic Technology Applications /
Olympus NDT. Version: 2007. http://www.olympus-ims.com/de/books/pa/pa-intro/. 2007. – research report

58
 Camera Encoded Phased Array for Semi-Automated Inspection of Complex
Camera Encoded Phased Array for Semi-Automated
Composite Components Inspection of Complex
Composite Components
Borja Lopez
Borja Lopez
Innerspec Technologies, Inc.
Innerspec
2940 Perrowville Technologies,
Road, Forest, VAInc.
24551, USA
2940 Perrowville Road, Forest, VA 24551
blopez@innerspec.com
email blopez@innerspec.com

ABSTRACT
This paper introduces a new wireless solution that permits performing accurate and traceable ultrasonic scans of
components with complex geometries using a hand-held scanner. The system integrates an array of 3D cameras that
track the position of the hand of the inspector with a high-performance PAUT instrument to provide accurate, high-
resolution C-Scans on any component. This paper provides results of hand-held scans on complex composite parts,
and explores how the solution compares with traditional semi-automatic and automatic systems in terms of setup,
ease-of-use, performance, productivity, and cost.

Keywords: wireless encoding, 3D camera encoding, Phased Array of composites, semi-automated inspection.

INTRODUCTION
A basic premise in project management is called the Triple Constraint or Iron Triangle of Time-Quality-Cost.
Experienced project managers like to remind their supervisors or clients that they can choose fast, better, or cheap,
but only two at a time.

A common solution to push the boundaries of this triad is by increasing the level of automation in a process. A
higher degree of automation is typically associated with more output, better quality, and, despite the initial higher
cost, a cheaper alternative for large-volume production. For NDT in manufacturing, this has meant the introduction
of in-line and off-line integrated inspection systems with increasing degrees of sophistication. These systems can be
run by line operators without NDT expertise, and still deliver permanent records of inspection to demonstrate
compliance to a customer or regulator. In these manufacturing environments, manual inspections are increasingly
relegated to MRO (Maintenance, Repair, Overhaul) processes that are characterized by low volumes and complex or
difficult to access components. The findings and quality of records on a manual inspection will be ultimately
dependent on the quality, training, and experience of the NDT practitioner.

In these manufacturing environments, the challenge arises when quality inspections and a permanent record is
demanded, but the complexity of the component or limited production volumes do not justify the investment in an
automated inspection system. This problem is compounded when there is limited access to qualified and
experienced NDT inspectors to perform the work and certify the results.

The solution many times involves semi-automated scanners, crawlers or similar devices that are mounted on the part
for complete or partial scanning. These systems are typically limited to simple-geometry parts since they do not
have the mechanical flexibility of robots, and the ability to provide accurate 3D encoding to match the inspection
results with the actual geometry of the component. Also, setup, inspection, and interpretation still require trained
NDT personnel.

This paper introduces an innovative solution to perform advanced ultrasonic inspections on complex parts with very
limited setup costs and investment. The system integrates an advanced phased array ultrasonic system (PAUT) with
an array of infrared (IR) vision cameras to provide three-dimensional (3D) spatial location of the probe on manual

Copyright 2018. This paper is intended for the sole use of registered attendees. No part of this publication or its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.

59
inspections. The system provides a complete C-Scan of the part with sub-millimeter positioning accuracy on very
large parts, completely eliminates the need for mechanical encoders, and can be used on parts of any shape.

This encoding technique can also be used to simplify an upgrade from conventional UT to advanced PAUT on an
existing inspection cell by eliminating the need to integrate the ultrasonic instrument with the line robot or gantry.

The results presented are from actual systems used for inspection of Carbon Fiber Reinforced Polymer (CFRP) used
in aerospace. The same technique can be used for inspection of forgings, or other large components of any shape.

DESCRIPTION OF EQUIPMENT
Figure 1 shows a block diagram of the main components of the inspection system.

Figure 1. Block diagram of the UT inspection system based on infra-red visual tracking.

The system includes:

• Data Acquisition System (DAS). The PAUT instrument is based on Tecnatom’s SONIA platform and
InspectView software.
• UT Assembly. Includes a 32 or 64 element linear Phased-Array probe, custom wedges for mechanical contact
to the part surface, a bubbler for ultrasonic coupling in local immersion, a device for ergonomic handling of
the assembly, and the IR markers for spatial location of the complete assembly (small spheres reflecting light
in the infrared spectrum).
• Infrared Vision System. This system contains several infrared cameras (from 3 to 20), depending on the
size of the part inspected. These cameras are placed in the cell so at least three cameras have always a direct
line of vision to the UT Assembly. The special triangulation algorithms processed in real-time by a central
processing unit (CPU) generate the 3D positioning coordinates of the UT Assembly, which are transferred to
the DAS to generate the C-Scan of the part.
• Water Management. The bubbler holding the transducer is continuously supplied by a small water flow for
proper ultrasonic coupling. A tray collects the overflow which is filtered and pumped back to the UT
Assembly.
• Part Fixture. Holds the part and provides mechanical stability during the inspection process. The Part Fixture
also includes IR markers for correct mapping of the UT Assembly on the part.

The inspection technique is a linear Phased-Array UT Pulse-Echo (PAUT-PE) with custom transducer and wedges to
adapt to any of the geometries that are found in production parts (inner/outer radius, ramp, basins, etc.). The sensors
can run at different frequencies (typically 3.5 or 5 MHz) to meet the inspection needs.

Figure 2 shows an example of a standard implementation of the solution in an industrial environment. In this
configuration, the system has eight IR cameras installed at the top of a metallic structure to ensure that the UT

60
Assembly is followed by at least three cameras at all times. The structure also permits wiring and piping for the UT
Assembly and holds the DAS and screens used by the inspectors to monitor the results.

Figure 2. Industrial implementation of a inspection cell based on IR visual tracking.

INTEGRATION CHALLENGES AND SOLUTIONS

UT inspection and IR 3D scanning are known technologies that are frequently used independently. This
section covers the integration challenges and the custom solutions developed for this innovative application.

• Ultrasonic Coupling. There are several coupling methods to perform PAUT-PE inspections, in manual or
semi-automated inspections. A probe bubbler with custom wedges was the method selected because of its
reliability and productivity. Use of dry-coupling techniques is currently being investigated but has not yet
been implemented due to severe limitations for inspection of curved shapes.
• Ergonomics. Considering that the inspection is performed manually, the ergonomics of the Part Fixture and
UT Assembly are critical to provide a safe and productive work environment. Each part needs to be properly
analyzed prior to installation to set up the right height of the fixture and monitors, and permit comfortable
scanning of all surfaces.

Figure 3. Detail of ergonomic implementation of the IR markers on the UT module.

• Human-machine Interaction. Together with ergonomics, proper design of the human-machine interface is
critical for safety, reliability, and productivity. The CAMUS 3D interface has been optimized to provide real-

61
 time and easy-to-interpret C-Scans with an accurate depiction of the location of the probe. The system
includes controls so the operator can easily pause and resume the inspection without losing data, and re-scan
and overwrite areas of the part without compromising the integrity of the results. The equipment also permits
merging the information from several operators working on the same part, or even having several inspections
on different parts in the same inspection cell.
• Latency Control. One of the key integration components with IR tracking systems is the time required by
the IR system to process the spatial coordinates and transfer the information to the UT instrument, which can
take tens of milliseconds. Even though manual inspection speeds are typically low (less than 150 mm/sec)
this latency can cause hysteresis phenomena that will distort the C-Scan results. To avoid this problem,
especially when the IR tracking system is mounted on faster robotic lines, the technique requires software
controls designed specifically for integration between the IR and UT systems.
• Resolution, Accuracy, and Repeatability. UT inspection of parts with complex geometries typically
requires resolution, accuracy, and repeatability between 0.1mm and 1mm. These requirements involve good
mechanical characterizations of the component, the Part Fixture, the UT Assembly as well as fast and
accurate calibrations of the IR cameras with three-point correction tools. Fast calibration is also important to
quickly resume work if there is an undesired movement on the part, cameras, or Part Fixture.
• Lighting. IR tracking systems can be affected by direct or reflected infrared light from natural or artificial
sources causing location errors. To avoid these errors, IR markers need to be set in a pattern that is easily
tracked by the vision system. This is also important during the design of the inspection cell so the cameras
are not affected by light sources or reflections from outdoor windows, mirrors, lamps, etc.
• 3D Analysis and Rendering. Since CAMUS 3D is primarily used to scan complex geometries, it is
paramount that the results are also analyzed and provide interpretation in easy-to-interpret 3D renderings.
InspectView software provides tools to incorporate CAD data and overlay the inspection results on the
imported file. In addition to this, all UT results are analyzed based on the actual volumetric configuration of
the part for proper sizing and interpretation of the results. The inspection data can be presented in 3D and 2D
formats for analysis by the inspectors based on different code requirements and personal preferences. Figure
4 shows one of the available tools for 3D rendering and analysis.

Figure 4. Software tool to adapt complex geometries for 3D inspection environment.

• Combined Part Records. InspectView software for CAMUS 3D is designed so all the files from inspections
of one part can be merged into one single Part Record. This is especially relevant when the part has been
inspected at different times using probes and wedges adapted to the geometry of different sections. A typical
example is found in the inspection of parts combining flat surfaces and radii (either inner or outer) that are

62
 common in aerospace. In these cases, having different scans combined into one Part Record facilitates final
analysis and record keeping.

RESULTS
Figures 5 and 6 show some examples of C-Scan records produced with a CAMUS 3D system. Data from CFRP
reference blocks with artificial defects are shown. Images show two magnitudes: signal amplitude and sound
path. Both images are of high quality and homogeneity.

Figure 5. Comparative view of C-Scan records, both in amplitude and sound path. Obtained with the
CAMUS 3D inspection system on a CFRP reference block (pulse-echo technique, 5 MHz)

Figure 6. Comparative view of C-Scan records with amplitude signal. Left: data acquired with
automated inspection system (water tank). Right: data acquired with semi-automated inspection
system (Pulse-echo technique, 3.5 MHz)

Depending on the flexibility of the part, the rigidity of the holding fixture, and the UT technique, amplitude
records may reveal slight variations due to the changes in pressure applied by the inspector. These changes
are usually below 1dB, well within the tolerances for aerospace standards.

It is important to note that all the defects have near perfect resolution and there is no hysteresis observed
even at speeds of 500 mm/s.

63
FASTER, BETTER, CHEAPER
Manual and semi-automated inspection systems are usually associated with low productivity when compared
to automated solutions, and typically relegated to the inspection of small parts. One of the main advantages of
combining visual tracking with manual PAUT techniques is the ability to inspect very large parts with
complete reliability and at fast inspection speeds. A CAMUS 3D inspection cell can accommodate parts that
are up to 30m long, and have several inspectors scanning the part at the same time.

The inspection process is similar to painting with a broad brush. The monitor provides immediate feedback of
the location of the brush and the results of the “painting” on the part. The real-time scan delivers
instantaneous information on the quality of the results, and whether an area has been properly covered. This
feedback permits the operator to retrace their steps, increase or reduce the speed, or change angle and
pressure depending on shape and geometry, surface conditions, and other inspection variables. The CAMUS
3D system has been measured to provide an average inspection speed on flat areas of approx. 200 mm/s,
with productivity of 20 m2/h. These values are typically in between those achieved with conventional semi-
automated and fully automated solutions.

Another very important factor contributing to the high productivity of this equipment is new-part setup and
part-change times. By eliminating robot programming, new parts can be configured in a matter of hours,
while parts can be moved in and out of the fixture and be ready to inspect in minutes. Also, based on practical
experience, CAMUS 3D has a clear advantage over fully automated systems when inspecting large and less-
rigid parts where slight variations in shape during manufacturing, or flexing of the part when on the fixture
can cause position variations that robots cannot handle, but a human inspector can easily accommodate.

From a cost point of view, a full CAMUS 3D inspection cell will typically cost 60-80% less than an equivalent
robotic line and it requires considerably less space, maintenance, and safety considerations. This equipment
can be delivered in 2-3 months and installed in a couple of days compared to the standard 12 months
required for a fully automated line.

CONCLUSIONS
This paper shows how IR visual tracking technology has been integrated with a high-end PAUT instrument
and software to provide wireless 3D encoding during manual inspections. Depending on the part and
production volumes, this semi-automated tool can provide equivalent or better quality and productivity than
a fully automated solution at a fraction of the cost. This encoding technique can also be used to upgrade a
conventional UT inspection cell to PAUT without having to manipulate or reprogram the inspection robots or
gantries.

Future work includes adding other scanners and techniques (e.g. EMAT), and improving the man-machine
interaction using augmented reality and similar techniques.

REFERENCES & ACKNOWLEDGMENTS

This paper has been adapted for ASNT from an original version delivered at ECNDT by Mr. Fernando Ojeda
(Tecnatom). The work is supported by several national and international patents, and has been possible
thanks to a technology transfer agreement with FADA-CATEC, the effort of many people in Tecnatom and
Innerspec, and to the contribution and feedback of aerospace component manufacturers, such as Corse
Composites Aéronautiques (CCA), Turkish Aerospace Industries (TAI), Aernnova Illescas, Componentes
Aeronáuticos SA (COASA), Airbus, and Lockheed Martin.

64
 Implementation Implementation
of Special PA UTofTechniques
Special PAfor
UTManufacturing
Techniques Inspection of
Friction Stir Welds
for Manufacturing Inspection of Friction Stir Welds
Michael
Michael Monette,
MONETTE 1 DanyDEVOS
, Dany Devos, 1and
, andDaniel
DanielRichard
RICHARD1
1Zetec
Zetec
875 Boulevard
875 Boulevard Charest
Charest Ouest,
Ouest, Bureau 100
Bureau 100
Québec, QC G1N 2C9, Canada
Québec, QC G1N 2C9, Canada
(418) 266-3020; faxfax
(418)-266-3020; (418) 263-3742;
(418) email
263-3742; mmonette@zetec.com
mmonette@zetec.com

ABSTRACT
Appropriate and efficient non-destructive inspection of friction stir welds is a major challenge in the aerospace
industry. Select advanced UT phased array techniques have been considered in view of replacing conventional UT
and radiographic testing for friction stir welds of aerospace components.

This paper will address how special UT examination techniques are implemented using phased array technology, to
improve on the generic techniques.

One of the main challenge is the time consuming multiple probe configurations required to inspect the areas of
interest.

A dedicated probe design is used to implement multiple simultaneous inspection configurations. Combining
inspection configurations optimizes the inspection process by reducing requirement for the examiner manipulations
while enhancing volumetric coverage.

Theoretical considerations on probe design will be presented, as well as the practical implementation of the focal
laws using the UltraVision® phased array calculator.

The performance of the special technique will be illustrated with ultrasonic data from practical trials on
representative specimens with artificial and realistic flaws. It will be shown how the UltraVision software supports
comprehensive visualization of the data generated by the special techniques, and how the advanced tools allow for
efficient data analysis.

Keywords: phased array, weld inspection, friction stir weld

INTRODUCTION

Friction stir welding (FSW) is a solid state joining process frequently used in the aerospace industry. Depending on
the manufacturing parameters (such as rotation and translation speed), the FSW process sometimes creates
discontinuities within the weld, thus a non-destructive testing method must be applied to ascertain their integrity.
The most common inspection method is based on phased array UT, however the currently implemented technique
has several limitations in terms of industrialization as well as in terms of performances and reliability.

In an effort to facilitate and streamline the FWS inspection technique, Zetec has designed a dedicated optimized
probe as the preliminary step of a new industrial inspection solution development. The next paragraphs will describe
how this innovative dedicated optimized probe overcomes the current technique challenges while improving upon
discontinuity detection and sizing. Inspection of relevant samples will be used to demonstrate the capabilities of the
new solution.

Copyright 2018. This paper is intended for the sole use of registered attendees. No part of this publication or its contents may be copied, uploaded to the internet, or stored in any shared retrieval system.

65
PHASED ARRAY INSPECTION CONFIGURATION – STATE OF THE ART

The most common practice in the industry is to inspect FSW using multiple probe positions and several incident
angles to be able to cover the entirety of a weld. A butt joint weld is usually inspected using two different refracted
angle orientations, a +45-degree shear wave and a symmetrical -45-degree shear wave (+20-degree and -20-degree
incident wave angles in immersion). In case of a lap joint weld, the same two beam orientations (+45SW and -
45SW) are used in conjunction with a 0-degree longitudinal wave.

By using one 10MHz 64 elements probe, such a scanning plan requires three passes with manipulations between
them. It is important to note that those manipulations can be tedious, but also need to be performed with the utmost
care since the orientation adjustment of a standard commercially available probe is of first importance to generate
the proper angle within the material. This angle adjustment is critical to achieve the necessary inspection
performance.

The main reason these manipulations are necessary is the limited steering capability of the probe whose design
parameters are typically not well adapted for FSW inspection. As a consequence, the standard probe must be re-
oriented each time a different refracted angle is required (see Figure 1) unless a multiple probes inspection setup is
implemented. This multiple operation inspection procedure significantly complicates the calibration process.

Figure 1 Probe alignment for 45-degree refracted shear wave in aluminum FSW plate

Once the probe orientation is set, one single UT channel configuration reaches an acquisition rate of
approximatively 60 Hz, meaning an inspection speed of 60 mm/s with a 1 mm displacement increment per second.
An additional characteristic of this inspection configuration is the limited detection capability for non-axially
oriented indications.

OPTIMIZED INSPECTION SOLUTION

Standard 10 MHz 64 elements probe performance evaluation

As a first step and to assess the inspection configuration and associated detection performance, a standard 10 MHz
64 elements probe has been used to scan a FSW weld (FSW plate-01) with artificial indications which simulates
realistic FSW discontinuities. This stands for a performance baseline to demonstrate the benefits and improvements
the new optimized design offers compared to standard commercially available solutions.

66
The standard probe has a law-to-law resolution of 0.5mm, which yields to a one pass coverage of 30 mm. It must be
oriented at an angle of 20 degrees with respect to the surface of the specimen in order to generate 45-degree shear
waves within the material (as shown in Figure 1). This orientation is critical since a misalignment of the probe will
not generate the expected beam and lead to erroneous results.

The inspected plate is 6.6 mm thick and has three types of typical FWS flaws. As shown in Figure 2, the specimen
has side-drilled holes (SDH), flat bottom holes (FBH) and notches, each at different depths or orientations. The
diversity of the flaws is representative of the different defects susceptible of occurring in standard FSW.

Figure 2 FSW plate-01 indications map

Table 1 details specific characteristics of each of the indications.

67
Table 1 Indications of FWS plate-01
Depth Length Width/Diam
ID Reflector Side/Location Orientation
Thickness (mm) In (mm) In (mm)
01 T/4 (1.7)
02 SDH Volumetric Along Weld T/2 (3.3) 0.25 (6.35) Ø1/64 (0.4)
03 3/4T (5.0)
04 Ø1/64 (0.4)
Weld joint
05 FBH 45° T/2 (3.3) N/A Ø3/64 (1.2)
plane
06 Ø5/64 (2.0)
07 0.05T (0.33)
08 Notch Root ID Along Weld 0.1T (0.66) 0.25 (6.35) 0.02 (0.5)
09 0.15T (1.0)
10 0.1T (0.66)
Notch Crown OD Along Weld 0.25 (6.35) 0.02 (0.5)
11 0.15T (1.0)
12 Notch Root ID 22.5°
13 Notch Root ID 45.0° 0.1T (0.66) 0.25 (6.35) 0.02 (0.5)
14 Notch Root ID 90.0°
15 T/4 (1.7)
16 SDH Volumetric Along Weld T/2 (3.3) 0.25 (6.35) Ø3/64 (1.2)
17 3/4T (5.0)

The scan result is presented in Figure 3. As can be seen, the standard method allows the detection of all axial
indications using the 45-degree shear wave channel, whereas non-axially oriented flaws are not detected at all.

Figure 3 FSW plate-01 scan result with a standard 10 MHz 64 elements probe

68
Optimized Probe Design

Zetec has developed a dedicated 128 elements probe to meet the objectives of simplifying and optimizing the FSW
inspection process.

The concept is to have a probe capable of generating plus and minus 45-degree shear wave inside the material as
well as 0-degree longitudinal wave with a single probe orientation, as displayed in Figure 4.

Figure 4 Multiple beam configuration using a single probe position

To do so, the probe parameters, such as element sizes, pitch and center frequency, have been optimized to enhance
drastically the steering capability compared to the standard probe, while keeping an acceptable sensitivity. This new
design also has the direct advantage to provide a much finer resolution between focal laws which lead to a better
sizing capability.

PERFORMANCE DEMONSTRATION OF THE PROPOSED SOLUTION

In an effort to demonstrate the efficiency of the proposed solution, inspections of two representative samples are
performed with the new design. The first sample, already described in the previous paragraph, is scanned with the
dedicated probe to assess the performances compared to the standard probe results. A second sample, with more
challenging flaws, is inspected solely by the new method because most the manufactured indications would not be
detected by the standard probe.

FSW Plate-01 Scan

For both specimens, the optimized probe scans are performed in a similar manner. The inspections were realized
using a single probe configuration, simultaneously generating -45-degree shear wave, +45-degree shear wave and 0-
degree longitudinal wave. The probe is mounted on a guided probe head to ensure perfect orientation between the
probe and the surface of the specimen. The scanning setup is displayed in Figure 5.

69
 Figure 5 Optimized probe inspection setup

Figure 6 presents a side-by-side comparison of the scan results obtained by using the standard and the optimized
probes.

Standard probe Optimized probe

Figure 6 FSW plate-01 45SW linear scan

The optimized solution demonstrates much greater performances illustrated by more accurate length sizing and a
reduced noise level. The method also demonstrates an improved lateral beam spread and a greater tolerance to
surface condition. Another strong point of the proposed technique is the reduced presence of undesirable artifacts in
the signal. All those factor in for a better signal to noise ratio, confirming the dedicated design’s improved
performance.

The same specimen is also inspected with a 0-degree longitudinal wave. Figure 7 demonstrates how the optimized
design can resolve the skewed notches while the standard probe fails to create clear indications. The enhancements
of the dedicated design previously stated for the 45-degree are still observed with the 0-degree channel. That being
said, the 0.05T (0.33 mm) notch cannot be detected by either method. However, it is possible to use the backwall
signal to detect the shadow created by the indication. Figure 8 shows how the backwall signal can be used to
emphasize the shadow of all indications, including the skewed and the 0.33 mm notches.

70
 Standard probe Optimized probe
Figure 7 FSW plate-01 0LW linear scan

Figure 8 FSW plate-01 0LW optimized probe backwall

FSW Plate-02 Scan

The second plate inspected is also 6.6 mm thick. It only contains ID notches which are all 4.00 mm long and 0.5 mm
wide. Indications #18 to #21 are located at T/4 (1.65 mm) and successively oriented at 0, 22.5, 45 and 90 degrees.
Similarly, notches #22 to #25 are located at T/2 (3.30 mm) at the same angles as the four previous indications.
Finally, indications #26 to #29 are 3.00 mm from the inner surface, at the same succession of angles, but with a
hemispheric surface. Figure 9 displays a plan of the FSW plate-02 specimen.

On this more challenging specimen, the focus is given on the ability to detect the skewed indications with different
height and top surface characteristics (flat and hemispheric).

Figure 9 FSW plate-02 indications

This specimen is only inspected with the optimized solution since it has already been determined that the standard
probe is unable to resolve non-axially oriented flaws. Figure 10 displays the 0-degree longitudinal channel results.

71
All 12 notches are easily identified with the appropriate orientations. An unexpected flaw is also detected at the right
side of the specimen (circled in red). Its presence inside the testing plate has been confirmed with a different testing
technique.

Figure 10 FSW plate-02 0LW linear scan

Sizing Capability

Multiple techniques can be used for flaw sizing. In this case, three techniques are used depending on the
situation. When using the 45-degree shear wave channels, tip diffraction can be measured on large indications. If the
indication is smaller than 1.0 mm, the after-corner mirror tip can be measured. When present on the 0-degree, the
echo from the top of an indication is used to measure flaw height.

Notches #18 and #22 from FSW plate-02 are good examples of large indications. Both are centerline root ID notches
oriented along the weld. Indication #18 is 1.65 mm high and indication #22 is 3.30 mm. Figure 11 shows that, using
tip diffraction, these indications are respectively sized at 1.51 and 3.21 mm.

ID #18 Reported Height 1.51 mm ID #22 Reported Height 3.21 mm

Figure 11 Tip diffraction height evaluation of large indications using 45-degree shear wave channel

As examples of smaller indications, indications #07 to #09 in FSW plate-01 are used. These are root ID notches
oriented along the weld with respective depths of 0.33, 0.66 and 1.00 mm. By using the mirror tip height evaluation
method on the 45-degree shear wave channel, the reported heights are 0.35, 0.44 and 0.85 mm, as shown in Figure
12. In Figure 13, the same flaws are sized using the top echo on the 0-degree longitudinal channel. This technique
yields heights of 0.26, 0.51 and 0.90 mm respectively.

72
ID #07 Reported Height 0.35 mm ID #08 Reported Height 0.44 mm ID #09 Reported Height 0.85 mm

Figure 12 Mirror tip height evaluation of small indications using 45-degree shear wave channel

ID #07 Reported Height 0.26 mm ID #08 Reported Height 0.51 mm ID #09 Reported Height 0.9 mm

Figure 13 Top echo height evaluation of small indications using 0-degree longitudinal wave channel

BENEFITS AND REMAINING LIMITATION

Benefits

All required refracted angles (+45 and -45-degree shear wave as well as 0-degree longitudinal wave) are generated
with a single probe positioned at a normal incidence with respect to the inspected specimen surface. This alone
simplifies tremendously the probe installation and alignment method since all that is now required is to install it in
the 0-degree scanner head. With such a design, operator manipulations are limited to a minimum and successive
inspections can be performed without interruption. In addition to eliminating multiple positioning manipulations, the
single probe proposition has the added advantage of simplifying the calibration procedure. By ensuring perfect
alignment between the three channels, the solution allows for single pass calibration when using the appropriate
calibration block and the UltraVision 3 calibration tool.

The optimized probe design brings a lot of improvements over the standard commonly implemented technique. By
modifying element size, pitch and center frequency, the new design improves the law-to-law resolution which
permits a higher beam resolution along the index axis. The probe also uses more elements than the original
configuration, thus improving the total coverage for a single scan pass. An additional effect of optimizing the probe
is that the new design is now able to detect non-axially oriented indications. Moreover, sizing and detection
capability for originally detected flaws is significantly improved.

73
Additionally, this inspection method is perfectly scalable to respect the industry requirement for fast and efficient
inspections. For a complete setup, including three channels at ±45-degree shear wave and 0-degree longitudinal
wave at maximum linear resolution, the system reaches speeds over 30 mm/s. While this may appear less than what
the previous solution achieves, multiple parameters factor in. The new solution has three channels, which is the
equivalent to three passes of the old technique. The linear resolution is greater, meaning more data is acquired. In
fact, if the resolution is modified to match the 0.5mm of the 64 elements probe, the speed is then doubled, also
matching the probe’s speed. Moreover, the new system eliminates a significant total time by reducing the amount of
complicated manipulations required by the standard technique. If faster inspection speeds are required, 2 probes can
be used on two units in parallel to reach up to 50 mm/s at maximum resolution or 3 pairs can be used for up to 100
mm/s.

The solution can be used with a simple 16/128 acquisition unit like the Topaz16. To get maximum inspection rate
performance, parallel firing capable equipment like the QuartZ are to be considered for an industrial application.

Remaining Limitation

Despite all the advantages the dedicated probe design brings, one major downside has to be mentioned. Upon scan
analysis, it is visible that when the probe generates a 45-degree beam, a symmetrical rear propagating lobe is also
created with a -18 dB amplitude compared to the main beam.
This low-energy lobe duplicates the echo of every flaws along the active aperture of the probe. Figure 14 illustrates
this phenomenon by mean of a beam simulation with two different location of the active aperture, as it is
implemented during a linear scan. Without a careful data analysis, this parasitic phenomenon could increase the
false alarm rate but it is possible to clearly identify these artifact indications as they are always associated with a
high Signal-to-Noise Ratio direct detection. Moreover, the “ghost” echo offsets are proportional to the
corresponding indication depth.

Figure 14 Beam simulation and top view of the optimized probe design on FSW plate-01

74
CONCLUSIONS & PERSPECTIVES

Using inspections of representative sample specimens, it has been shown that:

 A single probe optimized design can meet the inspection needs for FSW
 The proposed design improves on detection capability by allowing detection of skewed flaws and accurate
sizing compared to the common implemented solution
 The one probe solution improves on efficiency with faster total inspection speed and elimination of extra
probe manipulations
 The solution is scalable and can be used on multiple acquisition units depending on the required
performance

The next step in the development of the proposed solution is to provide an adapted localized immersion apparatus to
allow more portability and flexibility to this solution.

Another interesting perspective is to use the optimized probe in combination with the Topaz64 Total Focusing
Method (TFM) to assess if this brings additional benefits to the solution and specifically to determine if this could
further improve detection and sizing capabilities.

75
ABSTRACTS

76
 Commercial, Accelerator Driven
Neutron Radiography Systems
Brad Bloomquist, Michael J. Taylor, Katie Rittenhouse,
and Evan Sengbusch
Phoenix Nuclear Labs
2555 Industrial Dr., Monona, WI
email brad.bloomquist@phoenixwi.com
Faster, Easier, and Securer Inspection
than Using Mercury Vapour Lamps in
Fluorescent Magnetic and Penetrant
Testing UV LED Lamps by Using UV
LED Lamps Fulfilling the Newest
Aerospace Standards
Marc Breit

ABSTRACT SECU-CHEK GmbH

Innovative neutron imaging systems have been designed An der Faehre 9
and constructed by Phoenix to nondestructively investigate Kleinblittersdorf, Germany
components when conventional X-ray imaging or other email marc.breit@secu-chek.de
nondestructive inspection techniques will not suffice. This
method of radiography is commonly applied to comple-
ment X-ray inspection on specimen where a high density ABSTRACT
exterior structure prohibits interrogation of low density The LED-Technology is strongly substituting the conven-
interior components by the high energy X-rays (e.g. metal tional bulb based UV source (Mercury-Vapor, Xenon and
cladding surrounding low density powder or fuse). Phoenix Metal-Halide) in the fluorescent NDT methods.
has developed a high flux, compact, non-reactor based While general industry standards for UV LED lamps, like
neutron source which enables this inspection technique at ASTM E-3022, do not ensure the qualified and established
customer manufacturing sites in reasonable time periods. A process security and performance the newest Aerospace
first-generation electronic neutron generator coupled with Standards noticeable improve the fluorescent inspection
activation films for neutron radiography has been actively NDT processes MPI and FPI while drastically reducing the
used by the United States Army since 2013 to inspect muni- tiring of the inspector by increasing the visual ergonomics.
tions and other critical defense and aerospace components The presentation will give an overview of the newest
in a research setting. A second-generation system has been requirements and their influence to the inspection practice. It
designed, constructed and is currently undergoing test- further introduce new possibilities for enhanced interpreta-
ing and optimization at Phoenix with anticipated installa- tion of fluorescent indications and shows how the standard
tion in a low volume munitions production line in 2018. compliance can be simplified by integrated feature in UV
This system has an increased neutron output, utilizing a LED lamps.
gaseous deuterium target, which yields up to 3x10^11 DD
n/s, generating a higher neutron flux at the imaging plane
of nearly an order of magnitude over the first-generation
system. A third-generation system is in the design phases
with anticipated installation in a high-volume munitions
production line. Phoenix’s high neutron yield systems are
~1000x stronger than other neutron generators on the market
which dramatically reduces interrogation time and maintains
high spatial resolution and low geometric unsharpness. A
description of the neutron generator and imaging system,
including the beamline, target and neutron detector, is given
in this presentation. Sample images will be discussed using
both analog and digital platforms. State of the art neutron
moderators, collimators and imaging detector components
are also discussed in the context of specimen throughput and
image quality optimization.

We have performed extensive laboratory testing to evaluate

our technique. Test samples were aluminum sheets contain-
ing linear friction stir butt welds. Sheet thickness ranged
from 3 mm to 10 mm. The defects of interest for this study
were elongated voids. We tested samples with these defects
and were able to detect voids below 0.5 mm in cross-section
at process speeds.

77
 Predicting Bond Performance of
Composites Through Nondestructive
Testing
Brooke Campbell and Giles Dillingham
BTG Labs
5129 Kieley Pl., Cincinnati, OH
email bcampbell@btglabs.com
ABSTRACT
In aircraft manufacturing, the successful creation of adhe-
sively bonded primary composite structures hinges on being
able to control bond surfaces and predict bond reliability.
Surface preparation is a critical aspect prior to bonding,
requiring a non-destructive evaluation to verify and charac-
terize acceptable composite surfaces.
Traditional methods for evaluating composite surfaces, such
as dyne inks and water break tests, can be subjective and
destructive or otherwise inappropriate for a manufacturing
environment. An alternative non-destructive method for
determining surface energy is achieved via water contact
angle measurements.
X-Ray Photoelectron Spectroscopy (XPS) is a surface
sensitive technique for characterizing the top 10 angstroms
of a surface. Water contact angle measurements also probes
this uppermost layer. Quantitative results, as determined by
XPS, directly correlate with contact angle measurements
and therefore also with surface energy. XPS verifies the
use of water contact angle measurements to predict bond
performance.
This presentation will discuss the use of water contact
angle measurements to reliably and objectively characterize
composite surfaces with regard to surface preparation and
predict adhesion in bond performance.
78
Improvement to a Novel Ultrasonic
Spectroscopy Technique and Its
Applications to Bond-Line Analysis
Janelle Chambers
Southern Research Institute
757 Tom Martin Dr., Birmingham, AL
email jchambers@southernresearch.org
ABSTRACT
Advancement in composite joining methods to replace
fastening methods and further reduce structure weight and
manufacturing complexity is hindered by the lack of NDT
methods to validate the integrity of bonds throughout ser-
vice life. A new method is proposed that uses a combination
of swept-frequency spectrum signatures, resonance behavior
and time responses to evaluate adhesive layers in material
structures. The ultrasonic spectroscopy technique provides
a customizable range of frequencies with a robust signal-
to-noise ratio (SNR) up to 25 MHz with ranges as broad as
10 MHz. Using single-sided pitch-catch mode, the method
measures the changing resonance behaviors of the structure
as well as the bondline interface reflections as compared to
the backface. Examples are given for comprehensive bond-
line analysis.
The sound velocity of Lamb waves strongly depends on
the material and on the sample thickness. To transmit these
waves efficiently the angle of incidence (AOI) of the ultra-
sonic wave has to be adjusted. We present a new generation
of ultrasonic linear phased array probes allowing the AOI
to be adjusted electronically in order to achieve maximum
signal intensity.
Coherent Adaptive Focusing Technology
for the Inspection of Variable Geometry
Composite Material
Andre Lamarre and Etienne Grondin
Olympus America Inc.
3415 rue Pierre Ardouin
Quebec, Canada
email andre.lamarre@olympus-ossa.com
ABSTRACT
The aviation industry has seen above normal growth in
recent years, owing in part to lower oil prices contribut-
ing to millions of dollars in savings for aircraft operators.
As a result of this growth, production rates for new air-
planes have increased, and new aircraft programs are being
launched. Consequently, aviation component manufacturers
are facing new challenges, including a rise in production
rates, higher probability of detection (POD) requirements
due to the critical nature of the parts being manufactured,
a lack of skilled operators, and parts with increasingly
complex geometry. To respond accordingly, ultrasonic
phased array (PA) instruments have evolved, enabling the
implementation of advanced acquisition strategies, such as
adaptive focusing. Coherent adaptive focusing simplifies the
inspection of variable radiuses, variable opening angles, and
twisted components, and it also compensates for probe mis-
alignment through innovative signal-processing algorithms.
This paper presents an overview of coherent adaptive focus-
ing technology with the goal of helping NDT integrators and
composite material manufacturers address system perfor-
mance, production output, and quality control issues.
79
Eddy Current Array for Aircraft Engine
Component Inspection
Andre Lamarre
Olympus America Inc
3415 rue Pierre Ardouin
Quebec, Canada
email andre.lamarre@olympus-ossa.com
ABSTRACT
Components of aircraft engines are submitted to very high
stresses and temperatures during operation. Consequently, to
assure the safety of the public, the integrity of these com-
ponents must be ensured with reliable inspection methods.
While the use of remote visual and dye penetrant inspection
is widespread, the aircraft engine manufacturing and main-
tenance industries now benefit from recent developments in
magnetic methods, namely eddy current array.
Eddy current testing (ECT) has long been considered the
technique of choice to detect and size cracks on the surface
of aircraft engine components. Unfortunately, inspections
using single-coil eddy current probes can be slow with
results that vary from operator to operator. In these and
other respects, eddy current array (ECA) technology has
advanced the technique significantly. The use of arrays helps
to rapidly achieve full coverage of inspected zones while
maintaining a high resolution. Portable eddy current elec-
tronics enable the use of large arrays and the image result-
ing from the software’s data processing helps the operator
perform a reliable analysis. Probe holders and mechanical
supports are continuously being developed to address the
different applications of ECA for engines.
Today, eddy current arrays are used by major aircraft engine
manufacturers and maintenance companies for the assess-
ment of engine components, such as dovetails, blade attach-
ments, and turbine discs. This presentation will review how
eddy current array technology contributes to public safety.
 Recent Advances in Computed
Radiography Plate Grain Pattern
Compensation and Frame Averaging
Techniques with Related Technologies
Sean Larkin, Timothy J. Holmes, Melena Abijaoude,
Robert Schurz, Zhen Ji, and Joseph Goldfeld
1034 S. Brentwood Blvd. Ste. 525, St. Louis, MO
email sean.larkin@lickenbrocktech.com
ABSTRACT
X-ray sensitive plates used in Computed Radiography (CR)
are known to have sensitivity variations across the entire
plate. These variations, known as CR plate grain pattern,
can be easily observed in images acquired from CR plates.
A technique similar to Digital Radiography gain calibration
has been developed to compensate for the grain pattern in
the images. This technique results in improvements in the
contrast and noise characteristics of the CR images. Another
technique, known as frame averaging, has been developed
that uses multiple exposures of the same specimen to reduce
noise levels within images. These two techniques can be
used in combination to produce images with improved
image quality. Other related software advances such as a
novel Contrast to Noise Ratio (CNR) measurement tech-
nique and X-ray spot size calculation will be demonstrated.
80
Structural Health Monitoring of
Aerospace Composite Structures
Through Integrated Carbon Nanotube
Sensors
Hao Liu
University of Delaware
23 Duke St. Apt. D, Newark, DE
email lhjoe@udel.edu
ABSTRACT
Composites have been the most important materials used
in the construction of aerospace structures since their first
appearance. Numerous aircraft parts such as spoilers,
landing gear doors, wings, fairings and flight controls are
made from fiber reinforced composite materials due to their
high weight-specific mechanical properties including high
strength, low weight and high corrosion resistance. New
generation large aircrafts are designed with many laminated
structures, requiring an in-depth knowledge of structure
design and materials manufacturing. There are multiple
failure modes ranging from micro- to macro-levels that may
occur due to external loadings, including but not limited to
matrix cracks, fiber-matrix de-bonding, delamination and
fiber breakage. In this research, carbon nanotube sensors
are integrated for damage detection purpose of laminate
composite structures. The structural health of the composite
panels subjected to different mechanical tests was monitored
and the development of local damage was recorded and
analyzed. Conventional damage detection techniques have
been utilized simultaneously to help evaluate the carbon
nanotube sensors’ performance. The results have shown that
carbon nanotube sensors have a high detection sensitivity
and reliability for detecting and localizing damage occurring
in laminate composite structures.
 Detection of Defects in RAM-Covered
CFRP Structures by Active Microwave
Thermography
Ali Mirala, Kristen Donnell, and Mohammad Tayeb
Ghasr
A Novel CT Reconstruction Algorithm
for Compensation of Scatter and Beam
Hardening
Miguel Pineros, Sean Larkin, Timothy J. Holmes, Joseph
Goldfeld, Robert Schurz, and Melena Abijaoude

Missouri University of Science & Technology 1034 S. Brentwood Blvd. Ste. 525, St. Louis, MO
214 Engineering Research Laboratory email miguel.pineros@lickenbrocktech.com
500 W. 16th St., Rolla, MO
email smq3b@mst.edu ABSTRACT
An iterative algorithm to reconstruct Computed Tomography
ABSTRACT (CT) images has been developed for fan-beam and cone-
Aerospace carbon fiber reinforced polymer (CFRP) struc- beam systems. The algorithm, known as Fast Expectation
tures may sustain impact damage resulting in delamination Maximization Ordered Subsets (FEMOS), was initially
which reduce the integrity of the structure. Often these designed for non-destructive testing of aerospace system
structures are covered by thick coatings that prevent visual parts and improves resolution and overall image quality. The
identification of damage presence. An example of thick compensation of scatter and beam hardening is embedded
coatings are radio-frequency absorbing materials (RAM) within the algorithm, and is calculated in each iteration from
which are widely used to reduce electromagnetic inter- comparison with raw CT scanner data. The algorithm can
ference and scattering from reflective (i.e., conductive) also be utilized to perform a fast data correction, which will
surfaces utilized in aerospace and military applications. It enable other CT image reconstruction techniques used in
is important to nondestructively assess such structures for the industry to remove scatter and beam hardening. Recent
their structural health (i.e., defect detection). To this end, developments with this technique have provided significant
active microwave thermography (AMT) is considered as a speed improvements, approaching execution times similar
viable solution. AMT is based on microwave and thermo- to the Feldkamp-Davis algorithm. Results show significant
graphic nondestructive testing (NDT) principles, and utilizes beam hardening and scatter removal in tests performed with
a microwave-based thermal excitation. The resulting surface metallic additive manufacturing samples and phantoms.
thermal profile of the structure or material under inspection
is measured with a thermal camera. Defects present in the
CFRP structure may affect the heat diffusion and appear as
indications in the resulting thermal image. As it relates to
structures with a RF/microwave absorbing surface (such as
RAM), when RAMs are illuminated by microwave energy,
this energy is absorbed. Therefore, the RAM can serve as
a thermographic heat source to the underlying structure.
In this way, subsurface defects can be detected and evalu-
ated. The utility of this approach has been demonstrated
in [1], where a thin RAM was placed on the surface of an
omnidirectional CFRP laminate containing flat-bottom holes
(FBH). It has been shown that the presence of the defect
was easily detected when the RAM was present. This work
will extend this proof-of-concept work by investigating
the effect of size and depth of FBH in a CFRP structure.
Furthermore, the effect of microwave illumination parame-
ters (power and frequency, amongst others), and constituent
material parameters (e.g., RAM absorption properties and
thickness and CFRP thickness) will be considered.
This work has been partially supported by the National
Science Foundation Electrical, Communication, and Cyber
Systems (ECCS) Award 1609470, “A Multi-Physics-Based
Approach to Active Microwave Thermography”.

81
 Electromagnetic Characterization of NDE of Aerospace Materials and
Advanced Composites by Voxel-Based Composites using Acoustography
Inverse Methods
Jaswinder S. Sandhu and Charles Pergantis
Harold A. Sabbagh, Elias H. Sabbagh, Kim Murphy, and
Russell A. Wincheski Santec Systems Inc.
2924 Malmo Dr., Arlington Heights, IL
email j-sandhu@santecsystems.com
Victor Technologies LLC
PO Box 7706, Bloomington, IN
email has@sabbagh.com ABSTRACT
Acoustography is an alternative to traditional point-by-point
ultrasonic inspection. A large-area ultrasound detector,
ABSTRACT
called AO sensor, is used to form instant x-ray like ultra-
The nondestructive evaluation of advanced composites, such
sound images of test parts to reveal anomalies. This paper
as carbon-fiber reinforced polymers (cfrp), by electromag-
will discuss the latest advances in this NDE method and
netic means is well established. What is needed to advance
its effectiveness to provide a simple, practical ultrasonic
the state of the art are sophisticated inversion algorithms
inspection of both flat and complex-shaped composites;
that allow layup and impact damage to be determined in
without the need for complicated and expensive motion
localized regions, which means that the more traditional
control equipment such as a robotic or multi-axes gantry
methods of model-based inversion must be replaced by
system. Several examples will be provided to demonstrate
voxel-based methods. Thus, one will be able to better distin-
the effectiveness of Acoustography for the inspection of
guish such things as delaminations from fiber-breakage due
complex composite parts, such as brackets, L-shapes,
to impact damage, or other parameters that characterize the
U-shapes, etc.
mechanical state of the cfrp. We describe one such voxel-
based inverse method, namely set-theoretic inversion, and
show how it, together with arrays of spiral-coil inductors,
can be used to design inspection protocols for cfrps.
Magnetic Rubber Testing for Structural
Health Monitoring of Military Aircraft
Khalid Sulman Shabbir

Pakistan Aeronautical Complex Kamra

F 82 Gulshan Ravi
Lahore, PAKISTAN
email khalidsulman@yahoo.com

ABSTRACT
Magnetic Rubber Testing is used for life extension of FT-5
Aircraft. Detection of cracks on steel structure of FT-5
Aircraft was done through MRT. The programme success-
fully accomplished in life extension of FT-5 fleet.

82
 Thermographic Characterization of
Weak Defects in Cured and Uncured
Composites
Steven M. Shepard and Maria Frendberg Beemer
Thermal Wave Imaging Inc.
845 Livernois St., Ferndale, MI
email sshepard@thermalwave.com
ABSTRACT
Active thermography is widely used to detect defects such
as delamination or trapped water in a honeycomb structure.
The presence of a void causes a region of anomalous cool-
ing at the surface, which appears in the IR image sequence
presented to the inspector. Defects that can be detected this
way are generally spatially finite (they are smaller than the
camera field of view, so that they can be compared to a flaw
free surrounding areas), have large diameter to depth ratio
and their thermal conductivities are significantly different
than the host material. However, “weak” flaw conditions
such as porosity or the presence of Foreign Object Debris
(FOD) do not meet some or all these criteria, and thus, do
not significantly interrupt the flow of injected heat from
the surface. Although they cannot be directly detected in
the image sequence as hot or cold spots, they do cause a
delay in the transit of heat from the front to back wall of the
laminate, which can be measured using the Thermographic
Signal Reconstruction (TSR) method. Weak flaw detection
can be accomplished using several excitation schemes and
on cured and uncured composite parts, as well as in core
Al, fiberglass or foam core sandwich structures. In many
cases, FOD detection can be performed in near-real time.
For porosity detection, the transit time measurement is
well-correlated to ultrasonic attenuation and CT porosity
measurements.
83
Real-Time Measurement of Aerospace
Fastener Loading Using Laser
Ultrasonics
Max Wiedmann, Bradley Bobbs, and Marvin Klein
Optech Ventures
2520 W. 237th St., Torrance, CA
email mwiedmann@intopsys.com
ABSTRACT
We have developed a cost-effective technique for real time
measurement of the tensile stress or load applied to an
aerospace fastener during tightening. Currently the load on a
fastener is estimated from the torque applied with an instru-
mented wrench. This approach is not accurate, as much of
the applied torque is used to overcome friction in the threads
and under the nut being tightened. A more accurate means
of determining the stress is to measure the time of flight of
an ultrasonic wave that travels the length of the fastener
and returns to the starting point. As the fastener is loaded,
the transit time increases, due to the positive applied strain
and also to the decrease in the sound velocity associated
with the acousto-elastic effect. Our approach for measure-
ment of the time-of-flight is to use laser-based ultrasonics.
This technique is noncontact and the measurement can be
performed in real time during the tightening process. In
practice, a fiber-delivered probe would fit over the head or
tail of the fastener and would read out the load while the
fastener is being loaded. The standalone probe could also be
used during maintenance to validate that the load on the fas-
tener is within specification. We will describe experiments
on real fasteners which indicate that the strongest arrivals
arise from waves making multiple reflections along the side
wall of the fastener. Under load, the transit time for each
arrival increase linearly with the load. We will show how
the transit time can be used to measure the load. Finally, we
will describe a miniature probe that we have developed to
demonstrate the practical feasibility of our approach.
 In-Line Inspection of Additive
Manufactured Parts Using Laser
Ultrasonics
Max Wiedmann, Bradley Bobbs, and Marvin Klein
Optech Ventures
2520 W 237th St., Torrance, CA
email mwiedmann@intopsys.com
ABSTRACT
The adoption of additive manufacturing (AM) for rapid
production of complex parts has greatly increased in recent
years. However, there are no reliable inspection techniques
for qualifying these parts. Reliable qualification is espe-
cially important for safety-critical aerospace applications.
Currently, finished AM parts are often inspected by post-
process x-ray CT scanning. If any critical defects are found,
the entire part must be scrapped. If defects could be detected
during the build process, corrective action could be taken or
the part could be scrapped while incomplete.
Production CT Inspection in Aerospace
Jay P. Wood
GE
216 Palmetto Ave., Pacifica, CA
email jaypwood@gmail.com
ABSTRACT
Provide an overview of CT principles and how they relate
to existing Digital Radiographic principles in terms of
resolution and contrast - the keys to a quality image. The
push towards introducing production level CT inspection
performed at speeds that provide a value driven solution
for various market segments. Discuss current Practical CT
limitations and the technology improvements that minimize
them. Close with a couple of short videos that highlight CT
technology and how it is applied to Additive manufacturing.

In this presentation, we describe a technique for inspecting

the workpiece on a layer-by-layer basis, utilizing laser ultra-
sonic testing (LUT). The LUT beams are scanned across the
current layer, immediately after solidification, acquiring data
at regular intervals to generate a B-scan that is interpreted
with automated signal and image processing algorithms to
determine the position and depth of any detected defects.
By detecting defects during production, LUT obviates the
need for post-process inspection. LUT is ideal for in-line
measurements, as there is no sensor in contact with the
rough surface of the workpiece. LUT measurements can be
performed at elevated temperatures and at process speeds.

We have performed extensive measurements on titanium

and Inconel alloy parts in our laboratory. The samples
consisted of interrupted builds, with as-deposited surfaces.
Artificial defects were produced in these samples under their
build surfaces by electron discharge machining. We describe
our inspection method, present our laboratory data, describe
our signal processing methods and introduce our concepts
for integration of our sensor onto production machines.

84
 AUTHOR INDEX
Abijaoude, Melena . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80, 81 Lopez, Borja. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

Arnaud, Alexiane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Liu, Hao . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

Beemer, Maria Frendberg. . . . . . . . . . . . . . . . . . . . . . . . . . 83 Lucas, Manuel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Bellistri, Domenico . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Mirala, Ali . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

Bobbs, Bradley . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22, 83, 84 Monette, Michael . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Bloomquist, Brad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 Mück, Andreas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Breit, Marc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 Murer, David . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Calmon, Pierre. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Murphy, Kim. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Campbell, Brooke . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Pergantis, Charles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Cardillo, Nicholas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Pineros, Miguel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

Collins, Chase. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Proudhon, Henry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Chambers, Janelle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Remacha, Clément . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Devos, Dany. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Revol, Vincent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2

Dillingham, Giles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Richard, Daniel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Donnell, Kristen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 Rittenhouse, Katie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Fernandez, Roman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Rohlfing, Tom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

Foucher, Fabrice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Romero, Edward . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2

Ghasr, Mohammad Tayeb. . . . . . . . . . . . . . . . . . . . . . . . . . 81 Sabbagh, Elias H. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Goldfeld, Joseph. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80, 81 Sabbagh, Harold A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Grondin, Etienne. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 Sandhu, Jaswinder S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Grötzner, Maria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Scheunemann, Stefa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Haines, Harvey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Schurz, Robert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80, 81

Herrin, Jesse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Sengbusch, Evan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Holmes, Timothy J. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80, 82 Shabbir, Khalid Sulman . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Ji, Zhen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 Shepard, Steven M. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

Just, David . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Taylor, Michael J. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Kiel, Mario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Timm, Steve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

Klein, Marvin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22, 83, 84 Vinyard, Jeff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Lamarre, Andre. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 Wiedmann, Max . . . . . . . . . . . . . . . . . . . . . . . . . . . 22, 83, 84

Larkin, Sean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80, 81 Wincheski, Russell A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Leberre, Stéphane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Wood, Jay P. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

85