Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
DEVELOPMENT OF NEUTRON RADIOSCOPY AT THE
SLOWPOKE-2 FACILITY AT RMC FOR THE INSPECTION OF
CF1 88 FLIGHT CONTROL SURFACES
DÉVELOPEMENT DE LA RADIOSCOPIE NEUTRONIQUE AVEC LE &ACTEUR
SLOWPOKE-2 AU CMR POUR L'INSPECTION DES SURFACES DE CONTROL DE VOL
DU FC188
A Thesis Submitted
To the Faculty o f the Royal Military College of Canada
BY
Thomas Robert Chalovich
In Partial Fulfilrnent of the Requirements for the Degree of
Master of Engineering in Chernical and Materials
May 2000
O This thesis may be used within the Department o f National Defence, but copyright for open publication remains the property of the author.
National tibrary u*m of Canada Bibliothèque nationale du Canada
Acquisitions and Acquisitions et Bibliographie Services services bibliographiques 395 WeUington Street 395. rue Wellington OttawaON KlAON4 OnawaON K 1 A W Canada Canada
The author has granted a non- L'auteur a accordé une licence non exclusive licence aUowing the exclusive permettant à la National Library of Canada to Bibliothèque nationale du Canada de reproduce, loan, distribute or seil reproduire, prêter, distribuer ou copies of this thesis in microform, vendre des copies de cette thèse sous paper or electronic formats. la forme de microfiche/nlm, de
reproduction sur papier ou sur format électronique.
The author retains ownership of the L'auteur conserve la propriété du copynghî in this thesis. Neither the droit d'auteur qui protège cette thèse. thesis nor substantial extracts fiom it Ni la thèse ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent être imprimés reproduced without the author' s ou autrement reproduits sans son permission. autorisation.
ACKNO WLEDGMENTS
1 wisb to express my sincere gratitude and appreciation to Dr. L.G.1 Bennett and Major
W.J. Lewis, for their guidance, advice and assistance during the course of this research. for
without their help this thesis would not have been possible.
1 would also like to thank Ms K.S. Nielson, Director of the SLOWPOKE-2 Facility, for
her assistance and patience throughout this project. 1 am grateful to Mr. Orv Francescone who
instnicted me on the technical and practical aspects of neutron radiography, giving me a solid
foundation for this research. Further thanks goes to Leslie Chalovich, for being my initial editor
and for holding down the "fort" while 1 was pre-occupied with this joumey.
ABSTRACT
Chalovich, Thomas Robert. M. Eng (Chernical and Materials). Royal Military College of Canada. May 2000. Development of Neutron Radioscopy at the SLOWPOKE-2 Facility at RMC for the Inspection of CF1 88 Flight Control Surfaces.
Supervisors: Dr. L.G.1 Bennett Major W.J. Lewis
An ongoing investigation has identified a problem with the flight control surfaces on the
CF1 88 Hornet aircraft in that they suffer from water ingress into the composite layers and the
aluminium honeycomb core causing hydration and wat er entrapment , respectively . Neutron
radiography is being used to indicate the presence of water in these components. However, the
low neutron flux at the image plane at the SLOWPOKE-2 Facility at the Royal Military College
(RMC) causes relatively long exposure tirnes for neutron radiography. As a faster alternative,
neutron radioscopy, which is the combination of charged coupled device (CCD) or analogue
camera technology with a neutron source to produce a digital image on a persona1 computer, has
been investigated. With the advent of high quality and efficient CCD carneras, improved
software enhancement techniques and powerfûl cornputers, it is now possible to combine these
technologies to shorten the exposure times relative to neutron radiography for some applications
that do not require high resolution.
A methodical and intensive study was conducted to detennine the CCD camera's
characteristics and response to the elhination of light leaks, changes to carnera temperature,
varying exposure t h e , different diffuser characteristics and variable neutron flux levels. The
availability of two research reactors with neutron beams (SLOWPOKE-2 at RMC and the
Breazeale Nuclear Reactor at Pemsylvania State University) allowed for testing at different
neutron f lues which further validated the response of the carnera system. Further testing of the
system was carried out to confirm its ability to detect water in a honeycomb test piece. Final
testing required imaging a CF188 rudder with known water ingress to detemine the camera's
ability to image the presence of water in a realistic situation. Finally, a cost comparison between
neutron radiography and neutron radioscopy was made.
Analysis of the neutron radioscopy (camera) system installed at the SLOWPOKE-2
Facility at RMC has shown that it is capable of producing equivalent neutron images of flight
control surfaces at significantly shorter exposure times. In fact, by analysing the digital image as
a histogram of pixel count vs intensity value, a clear indication of the presence of moisture was
made evident. Moreover, comparison to higher neutron intensities confhned that the
SLOWPOKE-2 Facility has the capability to perfom adequate neutron radioscopy. As a result of
this investigation, it is recommended that neutron radioscopy be used as the initial inspection
method for the CF 1 88 flight control surfaces, followed by neutron radiography when warranted.
Chalovich, Thomas Robert. M. Eng (Chernical and Materfals). Collège militaire royal du Canada. May 2000. Dévelopement de la Radioscopie Neutronique avec le Réacteur SLOWPOKE-2 au CMR pour l'Inspection des Surfaces de Control de Vol du FC188
Superviseurs: Dr. L.G.1 Bennett Major W.J. Lewis
Une recherche continue a identifié un problème avec les surfaces de commande de vol
sur l'avion CF 18 8 Hornet c'est-à-dire qu'elles absorbent l'eau dans ses composantes en
composites et le noyau de nid d'abeilles d'aluminium causant de l'hydration et de l'occlusion,
respectivement. La radiographie neutronique est employée pour indiquer la présence d'eau dans
ces composantes. Cependant, le bas flux de neutron au plan d'image en service au SLOWPOKE-
2 du Collège Militaire Royale (CMR) entraîne des temps relativement longs d'exposition pour la
radiographie neutronique. La radioscopie neutronique comprend la combinaison de la
technologie analogique d'appareil-photo avec une source de neutron et le dispositif couplé chargé
(CCD) pour produire une image digitale sur un ordinateur personnel. Avec l'arrivée d'appareils-
photo et de CCD de haute qualité, les techniques améliorées de perfectionnement de logiciel et
les ordinateurs puissants, il est maintenant possible de combiner ces technologies pour raccourcir
le temps d'exposition aux neutrons pour certaines applications qui ne requiert pas Ia de haute
résolution. Une étude méthodique et intensive a été entreprise pour déterminer les
caractéristiques de de l'appareil-photo CCD et la réponse à l'élimination des fuites légères, les
changements de température de l'appareil-photo, le temps variable d'exposition, les différentes
vii
caractéristiques du diffuseur et les niveaux variables du flux de neutron. La disponibilité de deux
réacteurs pour des recherches scientifiques avec des faisceaux neutroniques (SLOWPOKE-2 au
CMR et le Breazeale à l'université de 1' Etat de la Pennsylvanie) a permis des tests à différentes
intensités du flux de réacteur ce qui a validé la réponse du système d'appareil-photo. Davantage
de tests du système ont été effectué pour confirmer la détection d'eau dans un morceau d'essai de
nid d'abeilles. Le test final comprenait l'imagerie d'un gouvernail de direction CF188 avec une
absorption d'eau connue afin d'évaluer la capacité de l'appareil-photo de détecter la présence
d'eau dans une situation réaliste. En conclusion, une comparaison des coûts entre la radiographie
neutronique et la radioscopie neutronique ont été faits. L'analyse du système de radioscopie
neutronique (appareil-photo) installé au seMce SLOWPOKE-2 du CMR a démontré qu'il peut
produire des images satistaisantes de piéces d'avion à des tamps dlexposition beaucoup plus
courts.
... Vlll
TABLE OF CONTENTS
Page
.................................................................................. ABSTRACT iv
.......................................................................... LIST OF FIGURES xi
... ........................................................................... LIST OF TABLES ml
........................................ LIST OF SYMBOLS AND ABBREVIATIONS xiv
................................................................................... CHAPTER 1 1
1 . Introduction ............................................................................... 1 ............................................................... 1.1 Non-Destructive Testing 1
1.2 Radiography .............................................................................. 1 ......................................................... 1.3 Status of Neutron Radioscopy 4
...................................................................... 1.4 Goal of this Thesis 6
................................................................................... CHAPTER 2 8
2 . Principles of Radioscopy ............................................................... 8 2.1 Neutron Sources ........................................................................ -8
.................................................................... 2.2 Neutron Attenuation 9 ........................................................................ 2.3 Imaging Devices 11
................................................................. 2.3.1 Scintillator Screens 11 ....................................................................... 2.4 Camera Assembly 12
2.4.1 Charged Coupled Devices (CCD) Camera ...................................... 13 2.5 Image Formation ........................................................................ 17 2.5.1 Hurnan Perspective ................................................................. 17 2.5.2 Scintillator Screen .................................................................. 18 2.5.3 CCD Camera ........................................................................ -20
2.6 Image Error .............................................................................. 22 . . 2.6.1 Statistical Error ..................................................................... -22 2.6.2 Systematic and Random Error ..................................................... 24
2.7 Digital Image ............................................................................ 26
................................................................................. 2.7.1 Format 27 ................................................................. 2.7.2 Image Enhancement 29
...................................................................... 2.8 System Resolution 38 2.8.1 Total Unsharpness ................................................................... 38 2.8.2 Measuring Total Unsharpness ..................................................... 40
.............................................................. 2.8.3 Resolution ftom MTF 41 2.8.4 Sensitivity Indicator (SI) ........................................................... 45 2.9 Image Represent ation ................................................................... 47
............................................................................ 2.9.1 Histogram -47 2.9.2 Intensity Difference Method ....................................................... 49
.................................................................................. CHAPTER 3 51
3 . Experimental Procedure ............................................................... 51 3.1 Equipment .............................................................................. 51 3.2 Experiments ............................................................................. 54 3.2.1 Installation and Optimisation of the CCD Camera System at RMC ......... 56 3.2.2 Camera Parameters ........................................................... 57 3.2.3 Resolution Measurement ..................................................... 59 3.2.4 Water Ingress Simulation ......................................................... 60 . .....................................................................* 3 2.5 CF 1 8 8 Rudder 64
................................................................ 3.2.6 Comparison to Film -66
................................................................................... CHAPTER 4 67
................................................................. . 4 Results and Discussion 67 ................................. 4.1 Installation and Optimisation of Diffuser at RMC 67
........................................................... 4.1.1 Reduction of Light Leaks 67
.......................................................... 4.1.2 Configuration Parameters 70 ...................................................................... 4.2 Carnera Parameters 72
............................................... 4.2.1 Dark Image - Temperature Change 72 4.2.2 Dark Image - Diffuser Changes ................................................... 74 4.2.3 Unified Images - Neutron Flux Effects .......................................... 76 4.2.4 Dynamic Range .................................................................... -79
................................................................. 4.3 Resolution Calcuiation 80 4.3.1 System Resolution Calculations .................................................. 81 . . . 4.3.2 System Sensitiwty ................................................................ 84 4.4 Water Ingress Simulation ............................................................. 85 4.4.1 WaterIngress ........................................................................ 85 4.4.2 Sïmulated Water .................................................................... -102 4.5 CF188 Rudder ........................................................................... 108 4.6 ComparisonofRadioscopy to Radiography ......................................... 118 4.6.1 Resolution ........................................................................... 118 4.6.2 Time and Cost Study ................................................................ 120
................................................................................... CHAPTER 5 125
................................................................................ 5 . Conclusion 125 ............................................................................ 5.1 Introduction 125
.............................................................. 5.2 Installation o f Hardware 125 ........................................................... 5.3 Charactensation of System 126
.............................................. 5.4 Development of Neutron Radioscopy 128 5.5 S u m m r y ................................................................................ 130
................................................................................... CHAPTER 6 131
......................................................................... 6 . Recomrnendat ions -131
APPENDICES
A Characteristics of the SLOWPOKE-2 and Breazeale Reactors .................... A- 1 ..................................................................... B Image Specincations B-1
....................................... C System Optimisation and Camera Parameters C-1 ............................................. D Error Calculation of Intensity Difference D- 1
......................................................... E Enhanced Images of Test Piece E- 1 ....................................... F Neutron Radioscopy Images of CF 1 88 Rudder F- 1
....................................................... G Cornparison of Nomialised Data G- 1
LIST OF FIGURES
Page
........................................................... Radiography Configuration 2 ............................................................ Radioscop y Configuration 3
............... Mass Attenuation Coefficients as a Function of Atomic Number 10 ................................................................... Single Pixel of CCD 14
..................................................................... CCD Architectures 16 .................................... Radiography and Radioscopy Image Formation 19 . . . . .................................................................... Digitisation Process 21
.............................................................................. Grey Scale -28 ..................................................... Image Enhancement Techniques 29
...................... Example of a Contrast and a Gamma Conversion Function 31 ............................................................... Geometric Unsharpness 39
................................................ 2.10 Ut Measurement for GD Knife.edge 41 ....................................................... 2.1 1 Example of Frequency per mm 42 ..................................................... 2.12 Sample of ERF and LSF Graphs 43
............................................................................. MTF Graph -44 ................................................... Modified ASTM E545-75 Type A 46
................................................................ Example of Histograrn -48 .................................................................... Histogram Example 49
..................................................... Beam Stop and Diffiser at RMC 52 ............................................................ Diffuser used at Penn State 53
................................................. Camera Assembly and Control Box 54 ........................................................... Flow Chart of Experiments 55
............................................ Example of Dynamic Range Histogram 59 .................................................... Construction of a CF1 88 Rudder 60
...................................................... Test Piece with Water at RMC 62 ........................................................... Test Piece with PE at RMC 63
..................................................... Test Piece with PE at Penn State 64 Film Layout on CF1 88 Rudder ....................................................... 65
................................................. Full Intensity Scale of Optimisation 68 .................. Enlarged Histogram Before and M e r Light Leak Reductions 68
................................................ Statistical Parameters of Dark Image 70 ...................................................... Effects of Camera Temperature 73
................... Peak intensity Values of Dark Images for Different Difisers 75 ............... Peak Intensity Values of Unified Images at Various Flux Values 77
xii
............................ 4.7 Dynamic Range of S ystem at Various Power Settings 79 ................................ 4.8 MTF Graph for 4-Minute Exposure taken at RMC 82
4.9 LinePair Gauge ........................................................................ 83 ................... 4.10 Effects of Varying Exposure Time on a Single Ce11 of Water 86
....................... 4.11 Intensity Difference of a Single Ce11 at 0.14 rnL of Water 89 ............................... 4.12 Effects of Varying Exposure Time for Three Cells 91
......................... 4.13 Intensity Difference of Three Cells at 0.16 mL of Water 93 ................................ 4.14 Effects of Camera Gain on a Single Ce11 of Water 95
.................................. 4.15 Effects of Camera Gain on Three Cells of Water 96 ................................................ 4.16 Fout Minute Exposure of Test Piece 98
............................................. 4.17 Results fkom Equalisation of Histogram 99 .............................................................. 4.18 Results f?om Erode Filter 100 .............................................................. 4.19 Results fkom J Flat Filter 101
........................... 4.20 Results of Varying Flux Levels for a Single Cell of PE 103 ............................ 4.2 1 Results of Varying Flux Levels for Three Ceils of PE 105
.............................................. 4.22 Exposure 2 of Rudder ffom Penn State 108 ......................... 4.23 Sequence of Exposure 5 Images taken at RMC (10 kW) 110
..................................... 4.24 Histogram of Exposure 5 nom RMC (10 kW) 111 ................... 4.25 Sequence ofExposure 5 Images taken at Penn State (10 kW) 112
............................... 4.26 Histogram of Exposure 5 nom Penn State (10 kW) 113 .................. 4.27 Sequence of Exposure 5 Images taken at Penn State (100 kW) 114
.............................. 4.28 Histogram of Exposure 5 from Penn State (1 00 kW) 115 4.29 Best Results of Exposure 5 fiom (a) RMC at 10 kW, @) Penn State
............... at 10 kW, and (c) Penn State 100 kW, with image enhancements 116 .............................. 4.30 Norrnalised Graph at Two Different Power Settings 117
................................................................ 4.3 1 Resolution Cornparison 119 .................................................. 4.32 Tirne Study Flow Chart for Rudder 123
LIST OF TABLES
Page
..................................................... 2.1 Designation for Sensitivity Levels 46
........................................................... 4.1 Statistical Data of 30 Images 71
...................................................... 4.2 Configuration Image Parameters 71 .
.......................................... 4.3 Resolution Variables for the NRS at RMC 81
.............................. 4.4 Hole Values fiom Modified ASTM E545-75 Type A 84
............................................... 4.5 Interna1 Component s of CF 1 88 Rudder 109
................................................ 4.6 Resolution Variable for Film at RMC 118
..................... 4.7 Cornparison of Capital Cost of Radioscopy and Radiography 121
........................................................................... 4.8 Cost per Image 122
.............. 4.9 Operational Cost and Inspection of Radiography and Radioscopy 124
LIST OF SYMBOLS AND ABBREVATIONS
AECL
ASTM
ATESS
BNR
CCD
ERF
m
FWHM
HFRNR
IFD
IP
1 Q 1
I(V
LPG
LSF
U A )
MTF
Atomic Energy o f Canada Lirnited
Arnerican Society for Testing and Materials
Aerospace and Teiecornmunicat ions Engineering Support
Squadron
Breazeale Nuclear Reactor
Charged Coupled Device
Edge Response Function
Fast Fourier Transfonn
Full Width Half Maximum
High Frame Tate Neutron Radiography
Image File Directories
Imaging Plates
Image Quality lndicators
Light fiom an object
Line Pair Gauge
Line Spread Function
Energy distribution
Modulation Trans fer Function
NAA
NBT
NCA
NDE
NDT
NRS
PE
Penn State
QE
QETE
RESC
Ri.WC
RMS or 00
SI
SD
T m
TRIGA
Neutron Activation Analysis
Neutron Beam Tube
Neutron Camera Assembly
Non-Destructive Evaluatior
Non-Destnictive Test ing
Neutron Radioscopy System
Po lyet hy lene
Pennsylvania State University
Quantum E fficiency
Quality Engineering and Test Establishment
Radiation Science and Engineering Centre
Royal Military CoUege
Root Mean Square
Sensitivity Indicator
Standard Deviation
Tagged image File Format
Training Research and Isotope production reactor by
General Atomics
Total Unsharpness
Geometric Unsharpness
Inherent Unsharpness
Neutron Flux
CHAPTER 1
1.1 Non-Destructive Testing
Non-destructive testing (NDT) and non-destructive evaluation (NDE) are the
generic tems for material evaluation methods used to assess the quality of a part or piece
of material without impairhg its hture usefihess. Non-destructive testing has become a
highly developed and well-utilized science within the a*ame and power plant
manufacturing industries. Modem aircraft with their complex structures, multiple and
unpredict able load paths, 10 w engineering safety factors and extended li fe times, combine
to produce a requirement for non-destructive inspections. The materials now introduced
into the manufacturing process, which can create lighter and stronger components, have
made the inspection of these components more complex and difficult. There are six
standard NDT techniques: visual inspections, liquid penetrant inspection, magnetic
particle inspection, ultrasonic testing, electromagnet ic eddy current and radiograp hy. It is
the last one, radiography, that is the focus of this thesis.
Radiography is a NDT technique in which an image can be taken of the interna1
characteristics of a part by utilizing either gamma, X-ray or neutron radiation. The layout
of a radiographie procedure is straightfonvard as indicated in Figure 1.1. An object is
placed in the path of a beam of radiation (X-ray, gamma or neutron) and a recording
medium is placed behind the object to be inspected. Some radiation will be absorbed or
scattered by the object, but most radiation will travel through and impinge on the
recording medium, producing a latent image. When the recording medium is processed,
the object will be visible. Intenial or material defects of the object will be indicated on the
recording medium as different densities relative to the background density. For gamma or
X-ray radiography, the source is either a radioisotope gamma source or an X-ray tube
while the recording medium is X or gamma ray sensitive or light-sensitive f i l a For
neutron radiography, the neutron source is a nuclear reactor, a radioisotope or an
accelerator (with a neutron producing target) while the recording medium is a converter
screen used to capture the neutrons and transfer the latent image onto X or gamma ray
sensitive or light-sensitive film The preferred neutron source is a nuclear reactor because
of the large neutron intensities available (which reduces exposure tirne) uniess other
conditions are paramount, such as location or any unique requirement of the object to be
inspected (e.g., intact aircraft at an airfield).'
l maging Plane
I Object
Figure 1.1 - Radiography Configuration
Neutrons interact with the nuclei of atoms rather than with the orbital electrons, as
do X or gamma radiation, and therefore, each elemental nuclide has its own characteristic
neutron cross section, showing no pattern in cornparison to the increasingly uniform
attenuation with miss nurnber for X and gamma radiation (see Section 2.2, Figure 2.3). In
fact, there is a slight reverse or complementary trend towards light elements for neutrons.
The major advantage of neutron radiography is its ability to reveal light elements such as
hydrogen contained in materials such as corrosion products and water.
With neutron radioscopy, the same source is used as with neutron radiography, but
the recording medium is different. It c m be any combination of a scintillation screen, a
photosensitive screen, an image intensifier, an analogue video camera, or a charged
coupled device (CCD) camera. A typical layout for radioscopy with a scintillation screen
and CCD camera is shown in Figure 1 .S.
Neutron Source
Aperture /
Scintillation Scryen
I
Ob ject _.: . ....... _.:.:.
Lens & A CCD Carnera
Figure 1.2 - Radioscopy Configuration
1.3 Status of Neutron Radiosco~v
Neutron radioscopy utilising a CCD camera is a relatively new technology as it
was fia applied to neutron radiography in 1990.' Advancements in CCD cameras and
cooling technology have made the use of CCD cameras both affordable and practical, and
therefore, their use has proliferated in the last five or six years. Although the CCD camera
is a new application in regards to neutron radiography, the development of a CCD system
has been thoroughly in~esti~ated.' As well, the CCD camera performance has been
analysed and the effects of c hanging the camera temperature, the exposure time and the
background radiation levels have been well documented.' Nurnerous investigations into
different material compounds that gWe the scintillation screens their basic characteristics
have been carried out to determine the matenal having the largest target for neutrons, the
best resolution and the bnghtest light output.sv 6 p As well, screens that respond to thermal,
epithermal and fast neutrons have also been investigated to determine their practicality.8
The majority of facilities that have installed a CCD camera system for neutron radioscopy
have done so for research purposes.
One of the largest area of research for neutron radioscopy utilising CCD camera
has been in the field of three-dimensional tomography, as the digital format lends itself
toward this form of compiled imagery. A practical use for tomography bas been the
detection of corrosion in turbine blades of jet engines. 9
It has been suggested that neutron radiography could be used for many aircraft
applications; specifically, it has been suggested that aluminium, honeycomb and
composite structures could be inspected for corrosion, moisture and adhesive defects. l0
Research has indicated that neutron radiography has a good sensitivity to detect corrosion
in aircraft grade aluminium. ' l Neutron radiography has been used to inspect corrosion in
KC-135 and DC-9 wingsl' and water ingress in CF188 flight control surfaces. l3 The
McClellan Air Force Base had the facilities to inspect complete aircrafi (FIA- 18 and F-
1 1 1) using neutron radiography and radioscopy (with Vidicon tube cameras). Although,
McClellan Air Force Base has been closed and the facility to inspect aircraft has been
decommissioned, the TRIGA (Training Research and Isotope production reactor by
General Atomics) reactor facility, at tbis location, with its five beam ports has the
capability to inspect large individual aircraft components.
At present, the practical use of neutron radioscopy using a CCD camera system has
been limited to only a few aircrafi applications. For example, one such application has
been the use of neutron radioscopy to investigate corrosion inside Iap joints of a KC 135
aircraft. '" In the recent past, a concern was raised over the airworthiness of the CF1 88
Homet, as initial indications pointed to possible problems that may exist in some of the
structures made fiom honeycornb construction. The Canadian Forces had aircrafi 188733
cornpletely inspected at McClellan Air Force Base with both real-time radioscopy and
conventional (film) radiography using both X-rays and neutrons.15 The results of the
inspection revealed 93 anomalies that included moisture ingress, ce11 corrosion, damaged
honeycomb core, foreign object material, voids and repaired areas. The problem areas
tended to be associated with components that could be removed fiom the aircrafi relatively
easily. Afier this inspection, the major emphasis then focused on how the components
that indicated potential problems could be inspected off the aircraft in-house rather than by
the very costly method of a complete aircrafi inspection at such a facility.
After an initial investigation to detemine which NDT technique would be suitable,
the Royal Military College W C ) , in conjunction with ATESS and QETE, was tasked to
do a comparison test between through transmission ultrasonic testing and neutron
radiography. While obtaining commissioning information, a small sample of CF 1 88 flight
control surfaces was inspected, which included thirteen rudders, eight ailerons, eight
trailing edge flaps and eight inboard and outboard leading flaps, l6 and the results of th&
test indicated that neutron radiography was supenor in detecting water ingress. The
SLOWPOKE-2 Facility (Safe Low Power c(K)ntical Experirnent) at RMC was
commissioned and then tasked to inspect a total of 10 sets of aircraft flight control
s d a c e s using neutron radiography. As well, due to the low neutron flux at the image
plane in the RMC Facility, a studyl' was completed in which faster neutron radiography
imaging rnethods and a quantitative measure of the water trapped in the honeycomb were
investigated to aid in the inspection of these components.
1.4 Goal of this Thesis
The goal of this thesis was to install, characterise and develop a near reakirne
radioscopy system at the neutron radiography system at RMC in order to expedite the
inspection of CF1 88 aircraft composite flight control surfaces for water ingress.
Installing, characterishg and developing a near real-the radioscopy system
requires a knowledge of the production and behaviour of neutrons, the operation
characteristics of a CCD carnera and scintillation screen, the formation of an image, the
accumulation effects of system error, the application of digital enhancements and the
calculation of system resolution.
M e r these concepts were understood, the camera system had to be tested in a
methodical way to determine how changing the basic parameters of the system affected
the image qualit y. Light leaks, camera temperature, exposure time, diffuser c haract erist ics
and neutron flux levels were al1 varied in different experiments to understand their
interdependency and their effects on the camera system. However, these tests and
experiments did not indicate how the system would perfom in detecting water ingress in a
honeycomb structure.
A honeycomb and composite test piece with varying amounts of water and high-
density polyethylene (PE) was used to discover the system's sensitivity to detect small
amounts of water. This test piece was also used to determine the optimum exposure tirne
and the digital enhancement technique for the system.
Finally, a CF188 nidder with known water ingress problems was imaged to
validate al1 of the previous experiments and to test the performance of the camera system
in a realistic situation. This component was used to provide a cornparison with film-based
neutron radiography.
PRLNCIPLES OF RADIOSCOPY
2.1 Neutron Sources
The three recognised neutron sources available for neutron radiography or
radioscopy are a radioactive source, an accelerator, and a nuclear reactor." These
sources offer a wide range of neutron intensities, operational complexity and special
features. The advantages and disadvantages of the these sources are well documentedg
and, depending on the requirements, it is generally accepted that a nuclear reactor is the
b a t neutron source with respect to neutron flux (4) (defineci as the number of neutrons
passing through a unit area per second).20 Two nuclear reactor facilities were used to
explore the parameters associated with radioscopy throughout this study. The first was
the SLOWPOKE-2 Facility at Royal Military College (RMC) in Kingston, Ontario,
Canada and the second was the Breazeale Nuclear Reactor (BNR) at Pemsylvania State
University (Penn State), Penns ylvania, United States. The characteristics of both
reactors used in this study are descnbed in Appendix A.
2.2 Neutron Attenuation
Attenuation of absorbed neutrons is exponential and is dependent on both the
neutron energy and the target material. The attenuation of neutrons in a beam follows
an exponential law:"
= L~WP)P'
where:
1 = emergent radiation intensity, I,, = incident radiation intensity, p = linear attenuation coefficient (cm-'), p = density, and x = material thickness
A cornparison of the mass attenuation for most of the elements for thermal
neutrons and X-radiation (Figure 2.1) shows that, for neutrons, each nucIide has a
varying attenuation which has no pattern or relationship to density or other physical or
chernical properties. In fact, there is a slight reverse or complernentary trend towards
light elements for neutrons. Neutron attenuation varies drastically between
neighbouring elements, and therefore, a single neutron radiograph can easily show
many different materials.
For most neutron radiography, the energy range used is between 0.0 1 to 0.3 eV
and the neutrons are referred to as thermal neutrons since their average energy is equal
to that of the material in the moderator. "
Figure 2.1 - Mass Attenuation Coefficients as a Function of Atomic Number
For neutrons, it is more wnvenient to have a relationship between the
coefficient and the cross section of the material. The linear attenuation coefficient can
where:
N = number of nuclei per cm3 of attenuating material q = total cross section (cm2) equal to the sum of the absorption (aa) and
scattering (a,) cross sections
The nuclear cross section is a vaiue that indicates the probability that an
interaction will take place between a neutron and one or more nuclei in the target
material. The cross section can be considerd as a measure of an effective target area
with units of square metres or barns (1 O"* m2).24 There are severai types of cross
sections, but the two of principal interest for neutron radiography are the absorption and
scattering cross sections, with the total cross section being their sum. The absorption
and scattering cross sections are the probabilities, respectively, that a neutron is
absorbed or scattered by the nuclei in the target rnaterial.
2.3 Imarring; Devices
For neutron radioscopy, a method is required to attenuate the neutrons and,
through a reaction, produce a sufficient amount of light to detect them with an
electronic or photosemit ive imaging s ystem. There are several methods used for
detecting neutrons to fonn an image, but the three most common rnethods used are
image intemifiers, imaging plates (IP) and scintillation screens of which the latter one
is used in this study. "
Originally, the term scintillation was applied to an optically clear homogeneous
slice of a single crystal of a rnaterial that fluoresces when irradiated by X-rays. The
terrn scintillation now takes on a more generic meaning and refers to al1 materials that
fluoresces whether clear, opaque, homogeneous, non-homogenous or constnicted fkom
a vast composition of elements. A typical screen is composed of a substrate, binding
matrix and phosphorescent rnaterial. The substrate can be constructed of plastic, glass
or aluminium. The substrate of choice for neutron radiography is aluminium as
attenuation to neutrons for this material is minimal, making it vimially transparent and
thus allowing a maximum number of neutrons to collide with the phosphorescent
rnaterial. The binding matrk should also be transparent to neutrons to eliminate further
neutron attenuation. There are several different compositions of phosphorescent
materials that have been tested for resolut ion and int ensity , with evaluat ions indicat k g
that Ti-based scintillation screens showed clear superiority? The mechanism for
producing light fiom neutron interactions is similar for most screens. This mechanism
will be discussed as it specifically applies to a 6~i-based scintillation screen having the
composition of 6 ~ i ~ : ~ n ~ : ~ u (NE 427 @) as used in this study.
Thermal neutrons interact with Lithium due its large cross section of 9 IO barns
to produce an alpha particle (4~e") and tritium ( 3 ~ daughter products.'7 The energy
fi-om the alpha particle is deposited into zinc sulfide (ZnS), an efficient phosphor that
produces visible light. The copper elenient acts as a wave length shifier to produce
light in the yellow-green region, which has an average wavelength of 525 na For an
optimised screen thickness, the 6 ~ ~ : ~ n ~ : ~ u screen can have a minimum spatial
resolution of 200 The characteristics of the 6 ~ i ~ : ~ ~ : ~ g screen is identical to
those of the 6 ~ i ~ : ~ n ~ : ~ u screen except that it has a average wavelength of 440 nm
caused by the silver (Ag) element. The 6 ~ i ~ : ~ n ~ : ~ g screen has a light output per
stopped neutron of 8.5 x 104 photonslneutron and a maximum detection efficiency of
15% for a given thi~kness.'~ For comparable screen parameters, it can be assumed that
the 6 ~ i ~ : ~ n ~ : ~ u screen has similar characteristics.
2.4 Camera Assemblv
The electronic medium used for recording an illuminated image fi-om an image
intensifier or scintillation screen is either a Vidicon or a SIT video camera or a CCD
video camera, of which the latter is used in this study.
2.4.1 Charged Coupled Device (CCD) Camera
CCD cameras are the most prominent devices used in neutron radioscopy due
to their high resolution, high quantum efficiency, wide spectnim response, low noise
and linear characteristics. The primary component of the carnera, fiom which the name
of the canera is derived, is the CCD silicon chip. The lem, electronic equipment and
woling unit support the CCD chip. The lem focuses and gathers incident light ont0 the
chip while the electronic equipment transfers the intensity values and location fkom the
CCD chip - either by an analogue or digital signal - to an output device.
The CCD chip is made fkom a matrix of individual collection sites. These sites
are called pixels. A single pixel can range f?om 6 pn to 30pm in size and, when
grouped together into a square array, the format can range from a 5 12 x 5 12 matrix to a
4096 to 4096 mat ri^.^' Pixel and matrix size determines the overall physical size of the
array and the area coverage possible.
Growing a thin layer of silicon dioxide ont0 the base silicon chip and then
depositing a conductive polysilicon gate structure over the oxide foms a collection site.
Applying a positive voltage to the gate creates a depletion zone between the silicon
dioxide and silicon interface, Figure 2.2.
Incoming Light
1 1 1 1 Elecrrical Connection
Figure 2.2 - Singie Pixel of CCD
A depletion zone or space charge zone is an area where charges are rare;
therefore, a charge void has been created in the material. In the depletion zone,
incoming photons are converted into electrons (called photoelectrons) and stored in the
depletion zone, also known as a potential well. Incident light projected on a collection
site for a fixed time allows an accumulation of photoelectrons into the potential we1lO3l
If an image in the f o m of incident light is projected onto a CCD chip, then each
individual potential well will accumulate photoelectrons according to the intensity of
the light hitting each collection site. At the end of the integration time, the image is
transfmed f?om the CCD chip. The process of charged coupling is used to move the
stored charges in the potential wells. The photoelectrons associated with a potential
well are called a charge packet During charge coupling, charge packets at a collection
site are transfmed fiom site to site by rnanipulating the gate potentials. The separation
of individual charge packets is preserved during the hansfer process. Three diEerent
methods of charge coupling are utilised, each haWig advantages and d i~advan ta~es .~~
To remove an image fiom a CCD chip, each charge packet must be transferred
to a serial register. There are three possible CCD architectures, each having a different
configuration for t r a n s f e g the charge packets to the seriai register, as shown in
Figure 2.3. A Full Frame CCD moves a row of charged packets into the serial register
and the packets are moved one at a time to the output amplifier. Downloading al1 the
charge packets to the amplifier takes a considerable amount of tirne, typically up to
twenty seconds. Altemately, a Fnune Transfer CCD chip is divided into two areas.
The first area is the normal configuration of a CCD, while the second area is masked
nom the incident light. The charge packets are transferred to the masked area, thereby
making the CCD chip ready for another exposure. A row of charge packets can then be
loaded into the serial register independent of the actions being taken by the other CCD
area. This configuration is faster at downloading the image, but this increase of speed
cornes at a significant cost. Either the resolution is reduced, due to half of the area
available, or the camera becomes more expensive to purchase, due to the increased size
of the CCD chip. Thirdly, the Interline Transfer CCD has small vertical areas masked
fiom the incident light. The charge packets are t r a n s f d to these masked areas
simultaneously and then loaded by rows into the serial register. This architecture has
fast d o m transfer thes , but the resolution is reduced and the masked areas may
interfere with picture resolution. The CCD camera used in this thesis was configured
with a full fiame transfer CCD chip.
Output Serial Clocks Output Seriai Clocks Output Amplifia AmpUer I
vvv vvv Amplfier
b
Dlrcctlon of Direction of ~ a r a l ~ d shirt ~ a r a l l ~ r shut
RiI1 RPme CCD Frime M u CCD -fer CCD
Figure 23 - CCD Architectures
2.4.1.1 CCD Chip Pdormance
The CCD chip's performance c m be defined in t m s of its dynamic range,
linearity, Quantum Efficiency (QE) and transfer efficiency. The dynarnic range of a
CCD chip is the ratio of the full saturation charge to the system noise and is expressed
in units of electrons. System noise is dependent on the quality of the system and can
vary fiom a few electrons for high performance systems to several thousand for
inexpensive video equipment. The saturation charge is detemiined by the well
capacity, which is related linearly to pixel size.
Linearity is the relationship between the amount of illumination on a single
pixel and the response to that ill~mination.~~ Linearity is lirnited to a working range of
intensities. At low illumination, threshold phenornena reduce linearity significantly
and, at high illumination, saturation of the potential well degrades the response.
Quantum efficiency is the measure of the effectiveness on an imager to produce
electronic charge fiom incidence photons. QE is dehed as the average nurnber of
detected photons divided by the average number of incident photons. CCDs have
efficiency ranges fiom 40% to 60%, compareci to photographie emulsions whose
efficiencies are presently 2% to 4%." The sensitivity of a CCD is limited by the gating
structure, which interferes and absorbs some of the incoming light. One solution to
eliminate this effect and increase the QE for a given CCD is to thin the silicun on the
backside of the CCD chip and have the incident light corne nom this side. A CCD chip
that has been thinned is refmed to as a backside iiluminated CCD. The CCD camera
used in this thesis is a backside illuminated CCD chip for the reasons stated above.
Transfer Efficiency is the rneasure of effectiveness of a CCD to iransfer charge
packets fiom one register to another. Only very small transfer inefficiencies cm be
tolerated due to the large quantities of charge packets that pass through the large sized
arrays. CCD registers have a typical transfer efficiency of 0.999990." h w
temperatures ( las than -1 00°C) and small nunber of charges (less than 1 O00 electrons)
adversely affect the transfer efficiency.
2.5 lmaae Formation
Visible light is electromagnetic radiation that is perceived by humans. Light is
quantified by the spectral energy distribution L(k), a fhction of wavelength (A) that
range fkom 350 to 780 nm. Light received ~ o m an object can be defïned by either the
reflectivity or transmissivity of the object p(h) and the incident energy distribution UA)
and express4 as:
I(V = ~ ( h ) x UA) (2.3)
The range of illumination over which the human eye c m operate is roughly 1 to 10'' or
10 orders of magnitude.
The human visual system is not stnctly the structure of the eye, but a
combination of receptor (eye) and interpreter (brain and visual experiences). Human
perception is sensitive to luminance contrast rather than the absolute luminance.
Bnghtness (also refmed to as apparent brightness) of an object is the perceived
luminance and is dependent on the luminance of the surrounding are~i.'~ Humans are
only capable of discriminating between 64 levels of shades of grey; given large number
of grey shades, a human cm no longer distinguish between the different shades." At
fewer than 16 shades of grey, the contours between the grey shades become a hindrance
in analysing an image. It will be shown in Section 2.7.1 that the CCD camera cm
quanti@ many more levels of grey than is possible with the human eye.
2.5.2 Scintillator Screen
The image formation fkom neutron radioscopy is fundarnentally different from
neutron radiograp h y. For radiography, after the neutrons interact (b y either absorption
or scatter) with an object, they are absorbed by the conversion screen and a conversion
electron fiom the conversion screen produces a latent image on the film. When the
negative fihn is processed, the area associated with the location of the object is less
exposed, translating to a lower film density. The radiographer sees this area as having a
lighter shade of grey than the surrounding area. In radioscopy, neutrons interact with
an object in the same manner with fewer neutrons hitting the scintillating screen.
Therefore the screen is not exposed in this area, but the remaining area of the
scintillating screen will be luminous at a wavelength of approximately 500 to 575 m.
The radiographer will view the area associated with the matter as a darker shade of grey
compared to the surrounding area. Essentially, a positive image, rather than a negative
one, is directly obtauied. Figure 2.4 is a drawing iIlustrating the results of radiography
(using film) and radioscop y (using a scintillating screen).
rp Radiography
S cre en Processed Film
Radioscopy
Scintillating Screen Compiaer Screen
Figure 2.4 - Radiography and Radioscopy Image Formation
2.5.3 CCD Carnera
An image presented to a sensor is made up of a continuous spectnim of
fiequencies and amplitudes of energy. A process interprets these energies so that the
image can be represented in a form that can be used by the sensor. For a CCD carnera,
digitisation is the method of mapping a continuous function into a discrete matrix with
a f i t e number of eiement~.'~ Digitisation of a continuous image wnstitutes an
enonnous loss of information, since the information conceming the grey values is
reduced fkom an infinite to a finite number of points. This Iarge loss of information
relative to the original object is one source of error in the system.
Digitisation of an image includes three steps as shown in Figure 2.5. These
steps include image formation, sampling and limitation to a finite window. Image
formation Unplies that the illumination intensity is collected over a specific area of the
sensor rather than at an exact location. The area is dependent on the size of the sensor.
For a CCD camera, the illumination intensity is integrated over each individual pixel.
Sampling takes the intensity values of the given area and assigns an average value of
intensity to grid points that outline this area of interest. The intensity values associated
with the area are then removed, leaving only grid point values. Sampling occurs in a
CCD when the illumination intensity is reduced to a single point, which is associated
with a single pixel. Sampling is physically done in the CCD when the collected
photons are converte- to an electrical signal. In th-, the image is still infinite in
size, as the original image is a continuous fûnction. The last process is called "limiting
to a finite window", and allows the image to be represented as a 2D matrix. In practical
ternis, the CCD array acts as the f i t e window and the pixel size lùnits the resolution
of the image.
Image Formation :
Sampling :
Limitation J
A Quant ization . ,'
Figure 2.5 - Digitisation Process
Once the image is digitised, the pixels are related to continuous grey scale
values. Due to the limitations of a cornputer, this continuous grey scale must be
mapped ont0 a lirnited nurnber of discrete grey values. This process is called
quantization. Quantization is a mapping process, but tends to be confusecl by users as it
is often used when referring to the bit size of an image. Although there are several
quantization scales, only two will be of interest to this process. The fint is an eight bit
image, which has 2* = 256 grey scale, and sixteen bit image which has 216 = 65536 grey
scale. The latter is prirnarily used as this large grey scale produces b e r intensity
levels, which produce intensity values that closer represent the original image intensity
values.
2.6
not
Image Error
Any measurement process introduces a probability that the value generated may
represent the actual item being measured. Two important classes of error exist,
statistical and systematic =or, into which al1 errors associated with neutron radioscopy
can be divided.
2.6.1 Statistical Error
Statistical error descnbes the scattering of measured values if the same
measurement is repeated. A measurement of the width of the distribution of a function
gives the statistical error while the centroid gives the mean measured value. An
accumulation of data indicates probability and confidence levels.
In neutron radioscopy, the measured quantity at a certain point in the image
plane is the luminescence. Originally, the observed radiance is charactensed by a
probability density function, p@, which is the probability of observing a grey value
(g). Quantization transfonns the radiance values to a discrete value and therefore,
statistically, the data must not be handled as a continuous but as a discrete fiinction, &.
A measured quantity governed by a random process, as is the case for observing grey
values, g, is referred to as a random variable.
Two basic parameters describing a random variable are the mean and variance
value. The mean value for discrete finite values is defïned as:
where p = mean grey value, and g, = grey value
Since a finite number of measurements are taken, the determination of the mean
remains an estimate with uncertainty, as it depends on the form of the probability
density function, in other words, the type of random process. Many processes with
contïnuous random variables (before quantization) can be adequately described by the
nomal or Gaussian probability distribution.
The variance 2 is a measure of the extent to which the measured values deviate
fiom the mean value. The variance for discrete values with probabilities that are al1
equal is define. as: 39
Given a normal distribution, there is an expectation that a single point will lie in
the confidence interval. A confidence interval is a part or whole value of a variance
that is added and subtracted to the mean value. Examples of the three most used
confidence intervals are:
p + l c + 68.27%
+ 20 + 95.45% C L - f 30 + 99.73%
It is only the middle confidence interval that is used throughout this report. The percent
value is referred to as the confidence level and implies that a person is confident by X
percent that a random value will fa11 within a given range. For example, using two
variances either side of the rnean, a point will have a value between the upper variance
and the lower variance 95.45 % of the time. The level of significance is the probability
that a given solution to a problem is wrong. The level of significance is directly related
to testing a statistical hypothesis and, d e r extensive tests of the hypothesis, a
probability is calculated indicating the potential failure of that hypothesis. A table
indicating the varied values for confidence levels and level of significance will indicate
the number of times the hypothesis must be tested4'
2.6.2 Systematic and Random Error
Systematic error occurs when the mean value of a measurement is beyond
statistical error margins of a true value. A precise but inaccurate measurement is
encomtered when the statistical error is low, but the systematic emor is hi&. Items that
cause significant systematic mors for radioscopy are electronic noise, dark current,
thermal noise, blooming and radiation.
Electronic noise associated with a CCD camera include noise fkom the
amplifier, the CCD (fixed pattern noise) and f?om charge transfer. Al1 of these types of
noise tend to be very small and are consistent for a given CCD camera. An amplifier
converts the photoelectrons to voltage values after leaving the register. The noise
associated with the amplifier tends to be very small and not significant. Fixed pattern
noise is due to the mors associated with CCD chip fabrication as the QE and charge
collection volume for each collection site varies. Unifoxm illumination of a CCD chip
causes site to site non-uniformity in collecting charges. Consistent and carehl
manufacturing will make variations in the site response small, thus making the error
negligible. The error associated with the charge transferring is a small residual charge
left in a site when the charge packets are moved to the register. Current CCD's have
achieved charge transfer efficiencies greater than 0.99999 and therefore it is reasonable
to assume that al1 charges are transferred and that this noise source is negligible.
Dark current is a leakage of charge into a site due to thermal agitation of the
crystalline lattice of the image dete~tor.~' If the dark current is not rernoved by reading
the element, it will accumulate and evenîdly saturate the pixel. Dark current is
dependent on the integration time and temperature of the CCD. At room temperature, a
standard CCD will saturate in only a few seconds. Dark cwent halves in value for
about every 6OC &op in temperature and are negligible at -1 00°C. " The magnitude of
this error is limited by the hardware specifications. This noise will be depicted as a
broad intensity spectnim fiom one pixel count to several hundred per site. The limiting
factor of dark current is that the quantity of dark current varies between sites, therefore
producing a non-unifonn image.
Thermal noise is a random component of the signal and is dependent on the
mean value of the dark ~urrent.'~ If a single site or pixel is measured over a single
exposure taken in total darkness and compared to a mean value determined from a large
number of identical exposures, the single value will differ slightly fiom the mean value.
The standard deviation or Root Mean Square (RMS) of total electrons accumulated for
a given site is represented as:
where a, = RMS, and Nt = number of total thexmal charges created
Bloorning is an observable effect due to light saturation of a single or multiple
sites. A site subjected to excessively strong illumination can accumulate photoelectron
charges that exceed the well capacity and leak into adjacent sites. The characteristic
image of over saturation is a bright spot with a halo effect surrounding this spot. In
addition, the number of charges accumulated in a saturated well can be such that its
contents cannot be emptied in one or more tramfers. A trail starting at the saturated
point appears in the direction of the transfer of rows and is the signature of blooming.
Cosmic and background radiation also introducw noise to an image. The site is
not a filter and therefore cannot distinguish between visible light and other fonns of
energy. The CCD is susceptible to cosmic radiation over long exposure times (100-500
seconds) at low temperatUres." A CCD camera operating in a nuclear reactor
environment will be subjected to higher background noise due to the presence of
gamma and neutron radiation. Radiation affects individual sites and can increase the
photoelectron amount significantly. Gamma radiation is visible on an image as one to
several pixels that have been saturated and therefore have a high value of intensity.
Neutron radiation will produce a line of saîurated pixels in an image. Noise due to
radiation is difficult to reduce given the relatively high levels of radiation encountered
at the camera location and, therefore, it is necessary to place the CCD camera in a
properly shielded location near the bearn.
2.7 Digital Image
With the advent of fast and large capacity cornputers and the development of
scanners and CCD carneras, the concept of digital images has become common. A
digital image diffas significantly fiom conventional film, in that a digital image is an
image constructecl £iom nurnbers and can be manipulateci in real time to adjust a poor
image or reveal latent information.
2.7.1 Format
AAer the camera produces the image, it can be foxmatted into several different
data types or classes. The data type or class indicates the number of bits used to
represent the pixel intensity value in an image and is referred to as its pixel depth or bits
per pixel. The term "bit" stands for binary digit and is the fundamental placeholder in a
binary system of numbers. A combination of bits equates to a number in the base 1 0
system. Eight bits equal one byte and the byte is the foundational unit when r e f e g
to digitd cornputers. The classes for image formats include:
Byte or Grey Scale 8 - 1 byte per pixel, 2* = 256, values in the image will range fiom O to 255;
Short integer - 2 bytes per pixel, values range fiom -32768,32768;
Unsigned 16 or Grey scale 16 - 2 bytes per pixel, values range fiom O to 65535;
Long Integer or Grey Scale 32 - 4 bytes per pixel, values fiom -232, z3*- 1 ;
Colour 16 - 2 bytes per pixel, with 5 bits assigned to the intensity of each of these colours, red, green & blue.
Colour 24 or RGB 24 - 4 bytes per pixel, with 8 bits assigned to the intensity of each of these colours, red, green, bIue.
There are several formats in which the image may be saved to a amputer
storage medium. A requirement exists so that no data are lost when saved to a file;
therefore, the file format must not have any built-in data compression. To be
transportable to different types of software, a recognised standard format is required. A
file format called Tagged Image File Format (TIFF) is one such format meeting al1 the
requirements. A TIFF file format is comprised of three basic components:
1. An image file header which indicates the order and version of TIFF file and the location of the first IFD;
2. The image's pixel information which indicates location and intensity value; and
3. The Image File Directones (IFD) which are a collection of descriptors that indicate the characteristics of the image.
There are more details signifiant to the TIFF file format including further
explanations on the three components making up this file fomat.''
The image is represented as a 3D matrix compnsed of values that indicate the x
and y location and the intensity level for every pixel at the location. The intensity level
depends on the data type or class. An image format of grey scale 16 produces intensity
values that correspond fiom black, havkg a value of zero, to white, having a value of
65535, as indicated in Figure 2.6.
Intensity Values
Black White
Figure 2.6 - Grey Scale
The location of the pixels on the image correlate to the location of the pixels
fiom the CCD camera. Therefore the image size is identical to the nurnber of pixels in
the CCD chip; a CCD with 5 12 x 5 12 pixels will produce an image that has 5 12 x 5 12
pixels and be approximately 512 kilobytes in size and will require half a megabyte of
storage space.
2.7.2 Image Enhancernent
The goal of image enhancement is not to increase inherent information in the
data, but to accentuate certain features for subsequent analysis or image display. Three
basic methods or classifications associated with dancing an image are m o d i m g the
intensity level, applying a spatial filter and manipulating the image fiequency (as
indicated in Figure 2.7). No universal grouping or classification has been officially
recognised, as other persons have chosen to categorise these methods differently.46
Image Enhancement I
htensity Level Spatial Filters Frequency Filtering
t
Brightness
Gamma Correction High P a s Fiat Fielding
Thrrsholding m i 0 Flat
Histograms J Flat
Figure 2.7 - Image Enhancernent Techniques
2.7.2.1 Intensitv Level
M o d i m g the intensity can enhance the overall appearance of an image by
changing the brightness, the contrast, the Gamma correction factor, the thresholding
values or the histognim shape (refmed to as histogram equalisation). uitensity is
related to the amount of light within a given image exposure and the distinction
between the shades of dark and light grey.47
Brightness (also referred to as apparent brightness) of an object is the perceived
luminance and is dependent on the luminance of the mounding area4* Luminance is
a measurable quantity that is not dependent or related to the object's surrounding
luminance. Mathematically, increasing the brightness of an image requires that each
pixel be universally increased to a higher value. Increasing the brightness requires that
a constant numerical value be added to the value of each pixel - the greater the value,
the larger the increase in brightness. Lowering the brightness requires that a constant
value be subtracted fiom each pixel value. In mathematical terms, if a given pixel
f?om an image is defined as Il, then increasing the brightness of that pixel is as follows:
I I f = II + Constant (2-7)
Contrast indicates the degree of difference between the brightest and darkest
components of an image. An image has poor contrast if the pixel values are within a
small range; conversely, an image with good contrast will have a large range associated
with the pixel's values. Changing the contrast of an image requires that each pixel
value be scaied by a contrast value, which redistributes the intensity values over a wider
or narrower range. Changing the wntrast is accomplished by multiplying each pixel
value by a constant, which by definition is a simple linear conversion. The formula for
changing the cuntrast of a single pixel (II) is as follows:
1,' = Il x Constant (2-8)
This method can also expressed as a h c t i o n and graphed as indicated in Figure 2.8.
Contrast and Gamma F unctbns
(Black) O 8192 16384 24576 32768 40960 49152 57344 65536 (White)
Original lnterrsity Value
Figure 2.8 - Example of a Contrast and a Gamma Conversion Function
Applying a Gamma correction factor is a non-linear method of changing the
contrast. The Gamma correction factor is used to enhance the contrast in either the
darkest or lightest area of the image. This is accomplished by applying pre-detennined
shaped non-linear curves to the existing pixel values. These curves skew the
conversion of the pixel values so that a dark image cari be lightened to enhance the
cuntrast in the darker areas or a light image can be darkened to enhance the contrast in a
lighter area as shown in Figure 2.8. The characteristics of the curves can be linear,
logarithmic, exponential or several other predetermined shapes.
Thresholding is a technique used to segment an image by its intensity values in
order to extract important features fiom that image. Thresholding requires the user to
indicate the upper and lower bounds of the intensity values of interest, while the
rernaining intensity values are ignored. Thresholding can be used to separate out the
intensities associated with a specific feature such as neutron absorption by the water.
Histogram equalisation is used to alter an image with poor visual characteristics,
by altering the associated histogram of that image. A nomal histograrn will either span
the full range (fiom O to 65k) or a large range of intensities. When an image has a
compressed range of intensities, histogram equalisation can be utilised to expand the
histogram completely by applying one of the many available equalisation functions.
Some types of equalisation functions include linear, bell-shaped (a greater
concentration of intensities are Iocated at the centre of the histogram), logarithmic,
exponential or a best-fit function. The best-fit function stretches the histogram to
achieve the best possible contrast di~tribution.~' The steps utilised for histogram
equalisation are as f01lows:~~
a. Compute the histogram of the image,
b. Integrate the histogram to form the cumulative distribution,
c. Scale the cumulative distribution to a mapping, which has values between O
and 65536, and
d. Apply the resultant mapping to the image.
2.7.2.2 Svatial Filters
Filtering operations reduce or increase the rate of change that occurs in the
intensity transitions within an image. Areas in which there are sudden or rapid changes
in intensity appear as hard edges in an image; convesely, areas where there are gradual
changes produce sof€ edges. Spatial filters detect and modiw the rate of change of
intensities throughout an image. A filter can increase the intensity differences on sofi
edges to make a feature appear sharper, or d u c e the rate of change of the intensity on
a hard edge in order to smooth and soften a feature.
Spatial filtering techniques are divided into three categories. The h t two
filtering techniques are called convolution (linear filters) and non-convolution (non-
linear) filters. The final filtering technique does not fit into either of these categories
and is classified under the heading of Mathematical in Figure 2.7. Convolution and
non-convolution filters alter the original image by examining and processing the data
around small regions, called pixel "neighbourhoods." A neighbourhood is a square
region of image pixels, typically 3x3,5x5, or 7x7 in size.
2.7.2.2.1 Convolution Filters
There are several different convolution filters available to the user; al1 of these
types of filters multiply the values within a neighbourhood by a matrix of filtering
coefficients (integer values) which is referred :O a s a kemel. Kernels are the same size
as the neighbourhood. The multiplieci results (neighbour pixel value times kernel
value) are summed and divided by the sum of the kemel values. The different b d s of
kernels and their size will produce different results to an image. An example of the
convolution method is as follows:
A 3 x 3 neighbourhood of an image is displayed below:
A sample kernel applied to the complete image could have the values:
The new intensity value located at the centre of the matrix would be:
The values in the kemel dictate the type of convolution filter. There are
different types of Convolution filters available and their application is dictated by the
difficulties associated with the image.
2.7.2.2.2 Non-Convolution Filters
There are several different non-convolution (non-linear) filters available to the
user. Non-convolution filters use pixel neighbourhoods and kernels, but unlike the
convolution filters, no predetemined values are used in the kernel; rather the intensity
values around the pixel of interest in the image are used in the kernel. Either the
statistical methods or mathematical formulas are applied to the kernel and
neighbouiliood values so that the pixel of interest may be modified. There are several
different kinds of non-convolution filtm available, but only Median and Erode filters
are applicable in this study.
Applying a Median filter will smooth an image by mod img pixel values that
Vary significantly fkom their surroundings and is accomplished by replacing the centre
pixel in a neighbourhood with the median value of the neighbourhood. Although
median filtering will soften an image, it generally preserves the edges in an image.
This filter is particularly effective at removing random, high-impulse noise from an
image (e.g., spots or points that vary significantly fkom the background).
The Erosion filter is a morphological filter that changes the shape of objects in
an image by reduculg (eroding) the boundaries of bright objects, and enlarging the
boundaries of dark abjects.'' The Erosion filter is used to d u c e or elirninate small,
bright objects in the background of an image.
2.7.2.2.3 Mathematical Techniques
The Mathematical filter method does not fit into either dassification
(convolution or non-convolution filters), but is a f o m of spatial filter. This spatial
filter mathematically combines either a single image or several images to a master
image. Each pixel value in the master image is altered by a pixel value fiom another
image at the same location.
Multiplying the master image by itself several times is a simple mathematical
technique, and has the effect of changing the intensity of an overly dark image to a
lighter image depending on the multiplication factor. This filter amplifies the original
image's intensiv, the effect is similar to increasing the gain of the camera, without the
introduction of electronic noise usually associated with large gain increases. This
technique also amplifies the intensity of the original noise in the image.
The next four mathematical techniques are similar and use reference images
(Flat fielding, Ratio flat, J flat and Optical density). An image of total darkness
(referred to as a dark image), an image of interest (referred to as the chta image) and
an image with a constant value of brightness (referred to as a unified image) are
mathematicaiiy mixed by a specified ratio. A dark image is an image where there was
no neutron interaction with the scintillation screen. The unified image is an image of
the scintillation screen that has been illuminated by neutron interaction without an
object present between the screen and the neutron source and therefore the scintillation
screen would be illuminated unifonnly, if the neutron beam were equaily distributed
over the area of the screen. Image ratios are used in a variety of applications to correct
or calibrate image data.
Flat fielding is generally used to correct for defects in an imaging detector array
and for non-uniform lighting. Such non-uniformity is really a "multiplicative noise"
problem and therefore requins a division operation to correct for this type of defect.
This correction method is more accurate than simply subtracting the background. Flat
fielding is calculated £iom the f o l l o ~ i n ~ : ' ~
Flat fielding = Constant * ((Data image - Dark image)Nniform image) (2.10)
Another variation of Flat fielding called Ratio Bat has also been utilised with some
success. The large différence is that the constant is replaced with a mean dark image.
The formula for Ratio flat is as follows:
Ratio 8 at = mean (dark image) * (Data image - Dark image)/Dark image (2.1 1 )
A simpla Flat fielding c m be utilised by eliminating the bnght image and
multiplying the result with a constant value. This method will be referred to as J flat
and calculateci by:
J flat = Constant * Data image/Dark image
Optical density calculations are more rnathematically intensive and produce
more accurate flat fielding. A base 10 logarithm is performed on the data after the ratio
of images has been taken. The formula for this method is as follows:
Optical density = Log 1 0 * Dark image/(Data image - Dark image) (2.13)
2.7.2.3 Frwuency Filtering
Image interference in the form of a reguiar pattern across an image can be
difficult to remove using spatial filters. The best method for removing periodic or
coherent noise is by converting the image f?om the spatial domain to the fkequency
domain and then editing out the noise and converting the image back to the original
foxmat. Coherent noise is a stable, reproducible, well-defined pattern that is not time
dependent? Although there are several types of fkequency filtering methods available,
the best known is Fast Fourier Transforms (FFT). Neutron radioscopy does not
produce an image having the characteristics that require the application of FFT and
therefore furtha discussion of this method is not required.
2.8 Svstem Resolution
Spatial resolution is generally quantified in terms of the smallest separation at
which two points can be disthguished as separate entities. The fiml resolution is an
accumulated resolution fkom d l components of the system. Some components will
have greater effects in reducing the totr! remlution. !f pixel size is the limiting factor,
the resolution is related to the srnallest item that can be resolved and not necessarily the
smallest feature. A feature considerably smaller than a single pixel can affect the
pixel's contrast relative to the surrounding pixels and is known as "partial volume
effect."" Direct analysis of a test sample only determines the total unsharpness of the
system and not necessarily the resolution of the system. System sensitivity is
quantified in terms of the system's ability to detect the smallest standard discontinuity
in any given sensitivity indicator (SI) that is obsenmble in an image.''
2.8.1 Total Unsharpness
Radioscopie images are essentially a "shadow picture" where several factors
contribute to the blurred detail. The total unsharpness of a system is defined as:
U: = cl: + U: (2.14)
where:
Ut = total unsharpness U, = geometric unsharpness Ui = inherent unsharpness
Geometric unsharpness is related to the physical characteristics of the neutron
facility and the object to image plane relationship. Geometric unsharpness is the
bliirring of the image at an object's edge. In some cases, geomeûic unsharpness will be
the limiting factor of a system's resolution. Geometric parameters required for
unsharpness calculaiions are illustrateci in Figure 2.9. The geometnc unsharpness has
unie of length and is calculated as follows:
If (L >> t), then geometric unsharpness can be written as:
where:
Ug = geometric unsharpness, D = neutron source size or collimator aperture size, t = object thiclcness or object to image distance, and L = source to image distance.
Image Plane
Neutron Object A us t
4 - Ue t
Figure 2.9 - Geometric Unsharpness
uiherent unsharpness is the product of al1 the imaging components contributing
to the unsharpness of an image. Inherent unsharpness cannot be rneasured directlx
rather the total is rneasured for an image and then the contribution of the geometric
unsharpness is subtracted.
2.8.2 Measuring Total Unsharpness
An image taken of a gauge made fiom a neutron absorbing material with
v-g size holes c m be analysed to detennine the smallest hole visible. The smallest
size hole will roughly indicate the resolution or more accurately the system's
unsharpness. Similarly, a gauge with varying width slots can also be used in the same
rnanner.
Another method of deterrnining the total systern unsharpness (-Ut) can be
measured directly fiom the Edge Response Function (EW) graph as shown in Figure
2-10? An ERF graph is produced by taking an image of a very thin (25 pm thick)
gadoiinium (Gd) or cadmium (Cd) sheet placed with the largest area perpendicular to
the neutron beam. The next step is to take a line density reading across the image of the
Gd or Cd sheet and the background. The line density values were measured utilising a
function in the image software for this research. Plotting the normalised densities
against distance produces a graph with a shape sirnilar to Figure 2.10. A curve fit of the
data will produce an ERF, which corresponds to the line in Figure 2.10.
0.0 0.4 0.8 1.2 1.6 2.0 2.4 Distance from ecige (mm)--->
Figure 2.10 - measurement from Gd Me-edge
2.8.3 Resolution fiom MTF
A Modulation Transfer Function (MW) is a more general method compareci to
the SI or other foms of Image Quality Indicators (IQI) and has been used for assessing
the spatial resolution of an optical imaging device." MTF provides a fast, reliable and
repmducible means of indicating the resolution of a system and is capable of
monitoring any changes in the optical system. The MTF of a system is the ratio of the
magnitude of the system's output amplitude to the magnitude of the system's input
amplitude as a function of spatial fiequency.58 The measurement of the MTF can be
made by imaging either ao object with altemathg O to 100% contrasting bars, a thin
open slit between 100% contrast material or a knife edge of 100% contrast material. A
knife-edge is the most convenient object to comtruct and image for neutron radioscopy
and was used to calculate the resolution of the CCD system at RMC. The resuit fiom
an MTF analysis is a graph indicating MTF values against fkquency (cycles per mm).
The resolution of the system is indicated by the fiequency value, which is the number
of line pairs in a given distance. A line pair is two equally thick lines that are highly
contrastai. The larger the fkquency number, the thinner the cycle is and the h e r the
resolution of the system. An example of a frequency with 2 cycles is shown in Figure
2.11. A system with a resolution of 2 cycledmm (Figure 2.1 1) would have a single
cycle width of 0.5 mm and a resolution of 0.25 mm (half of a single cycle).
Two Cycles per mm
Figure 2.11 - Example of Frequency per mm
The first step in producing an MTF graph is to create an ERF for the system as
previously describecl, which will produce a graph similar to Figure 2.12.
Differentiating the ERF will produce another fùnction called a Line Spread Function
(LSF) and, when the LSF is plotteci, it would look sirnilar to Figure 2.12.
Graph of ERF Graph of LSF derviveci fmm ERF
Figure 2.12 - Samples of ERF and LSF Graphs
Mathematicall y, an MTF is the magnitude of the optical tramfer function, which
is a nomalised Fourier transform of the LSF. Calculation of the MTF fiom the
nomalised transforrn of the LSF is as follows:
is the Fourier transforrn of the magnitude of the
system's output amplitude (LSF), and
.Cao
ILSF(x)d ir the to the Fourier transfomi of the magnitude of the - aD
system's input amplitude
An approxirnate solution to this equation, which is used to calculate the system
resolution at RMC (Section 4.3) is: 59
where p = measured pixels per mm, f = fiequency per mm, and Ut = total unsharpness of the system
Varying the value of the fiequency, f, the MTF values are cornputed and are
plotîed on a graph a s shown in Figure 2.13. A value of 10% for a MTF is normally
chosen and the fiequency is then detennined from the graph. A value of 10Y0
modulation indicates a 10% contrast difference between each half cycle.
A direct method of calculating the MTF of a system can be accomplished by
taking an image of a line pair gauge. A line pair gauge is constnicted so that several
fiequencies per mm are simulated with material having a 100% contrast- The image of
the line gauge is read and, when the altemathg contrasting material cannot be
distinguished, then the spacing associated with that group is the MTF value. A line pair
gauge was used at Penn State to determine the MTF for CCD camera system.
Figure 2.13 - MTF graph
2.8 -4 Sensitivity Indicator (SI)
No recognised standard resolution gauge or Sensitivity Indicator (SI) is
available for neutron radioscopy, but several methods have been adopted fiom neutron
radiography for measuring the total unsharpness of a system.
A modified Type A SI, fkom the American Society for Testing and Materials
(ASTM) 545-75 standard can be used to determine the smallest discontinuity or
porosity that the system can image. The goal of this A S T M sensitivity indicator is to
provide visual and quantitative data that could indicate the size of defect and the
maximum thickness of material for a given hole size. To adapt this ASTM standard to
radioscopy, the step wedge with the holes was thickened and the holes were enlarged,
Figure 2.14. A table indicaihg the R value fkom the ASTM and the new hole sizes and
wedge thiclmess are presented in Table 2.1. The sensitivity level reported is the largest
consecutive value of R that is visible in the image of the SI. An R value was calculated
for the CCD system at RMC and at Penn State using the Type A SI.
Figure 2.14 - Modified ASTM E545-75 Type A
Table 2.1 - Designation for Sensitivity
1 o f R 1 Discontinuity 1 Absorber Step 1
Levels 1 Value 1 Size o f Hole Thicbess of
2.9 Image Rmresentation
Depending on the method used to analyse a digital image, there may be a
requirement to present the image in other formats. In general, a digital image can be
analysed by a subjective or comparative method. The subjective method requires an
expert to analyse the image and give an opinion on the contents. The results fiom a
subjective analysis are not reproducible as the opinion is based on a person's
experience. The subjective method will not be discussed M e r . The comparative
method requires that the information fkom an image be presented as a histogram grapha
or a graph containing intensity difference values.
2.9.1 Histogram
A digital image is constmcted fkom a 3D matrix comprishg of values indicating
the x and y location and intensity level for every pixel location in an image. A
histogram is a graph that compiles the intensity data from an image into a particula.
format. By convention, the y-axis corresponds to the arnount of pixels per intensity
interval and the x-axis is the intensity scale, f?om O to a maximum value. The x-axis is
divided into sequentially grouped intensity intervals. An intensity interval can have a
range of values fiom one to a maximum of 256. An exarnple of a histogram is shown
with an explanation of its characteristics in Figure 2.15. Due to the large range of pixel
counts encountered, fkom 1 to 262 144, at times only a section of the histogram will be
s h o w e.g., the intensity may range fkom O to 4096 (black end of the scale), where the
intensity value 4096 is the cut-off point. There are still data beyond the cut-off point,
but due to a very low pixel count, anywhere £kom 0.02% (524 pixels) to 6% (1 5729
pixels) of the total number of pixels, it is deemed not significant enough to be included
in the histogram.
Histogram Graph
This point has 1000 pixels with an intensity value of 51 2
lntensity Scale
Figure 2,15 - Example of a Histograrn
A histogram is a powerful tool, as inîrhsic information can be identified more
easily and made more readable than by visually analysing an image. Subtle variations
in intensities are eliminated when presented as an image. Limitations to the image can
be due to both the cornputer monitor, which is capable of showing 256 levels of
intensity, and the human visual system which is sensitive to only 64 levels of intensity.
Histograms can represent the complete intensity scale. Unfortunately, a histogram can
only indicate quantitative information and can simultaneousIy represent two different
images as s h o w in Figure 2.16.~' Both images (Figure 2.16-a and Figure 2.16-c)
produce the same histogram, yet look totally different; therefore, the picture associated
with a histograrn shouid be reviewed to understand Mly the characteristics of the
histogram. The nurnber of pixels that correspond to the black intensity are the same for
each picture. This situation produces the left peak on the histograrn (Figure 2. la).
The number of white intensity pixels are also the sanie for both pictures; hence, the Ieft
peak in Figure 2.16b. If there were only two true intensities, black and white, there
would be only two lines at each end of the scale. Since these are not true black and
white images, the peaks are located more centrally and not at the ends. The peaks have
a width which indicate intensity variations around a centrally dominant intensity value.
Histogram
Intensity
Figure 2.16 - Histogram Example
2.9.2 Intensity Difference Method
For comparative analysis, the image must not be digitaily enhanced as simple
enhancements such as changing the contrast, brightness and gamma values oui improve
a poor image or degrade a good image.
Utilising the mean square criteria, which produces an average or s u m of squares
of the error between two images, is also not applicable.62 The mean square criterion
requires that al1 images be analysed agauist a standard or master image. No standard
image is available for comparison and, therefore, this technique is not applicable.
Normalising the image data allows comparison of images without the previous
disadvantages. Normalising, or what will be referred to as intensity diEerence, is
accomplished by comparing the average intensity of the background of an image
against the average intensity of a specific area or region of interest, i.e.,
intensity Difference (%) = [(avmge background intensity - average area of interest intensity)/ average
background intensityJgl 00 (2.19)
Utilising this method allows comparison between areas of interest on the same
image or similar areas of interest on other images. Intensity difference is a valid
method for comparing images, as the human eye is sensitive to perceived luminance
contrast rather than the absolute luminance valued3 In other words, the eye is more
sensitive in spotting an item if there is a large difference in contrast between two items,
in this case, the background and the area of interest. The difference in contrast between
two areas is the fundamental principle behind neutron attenuation; therefore, intensity
difference is an ideal mathematical method for comparing areas of interest or images.64
CHAPTER 3
EXPERIMENTAL PROCEDURE
The first neutron source used for neutron radioscopy for this thesis was the
Neutron Radioscopy System (NRS) installed in the SLOWPOKE-2 Facility at RMC. The
research reactor in this Facility is capable of producing a maximum thermal power of 20
kW (kilowatts), and is regulated by the neutron flux measured at the reactor core. For
neutron radiographylradioscopy, the reactor presently operates at half power for a nominal
neutron flux outside the reactor core of 5.0 x 10' n/cm2 S. The flux f?om the core produces
2 to 3 x 1 o4 dan2 s at the image plane.65 The Neutron Beam Tube (NBT) is circular in
cross section, tilted 8.S0 fiom vertical and has a measured LAI of approximately 1 0 0 . ~ ~
The second neutron source used by the author was the Breazeale Nuclear Reactor at
(BNR) Pemsylvania State University (referred to as Penn State). This facility is capable
of producing a maximum power of 1 MW. For the experiments, the reactor was set to 1,
10, and 100 kW which equate to a neutron flux level at the image plane of 3.0 x 104
n/cm2s (calculated), 3.0 x 10' n/cm2 s (calculated), and 3 .O x 106 n/crn2s (measured),6'
respectively. Flux values for 1 and 10 kW were calculated assurning a linear relationship
between power setting and measured flux value. The beam tube has a round shape, is
horizontal and has an UD ratio of 155.
The section of the NRS at RMC called the rniddle beam stop has a reenterent shape
that not only contains and stops the neutron beam, but also acts as a difhser for the CCD
camera system as shown in Figure 3.1. An aluminium and rubber beflows structure
comects the scintillation screen to the diffuser making a light tight enclosure. For the
experiment at Penn State, the scintillation screen and CCD camera were properly oriented
by a portable aluminium rectaogular and trapezoidai enclosure that was dimensionally
identical to the diffuser at RMC, Figure 3.2.
Scinl Screen
Beamstop
CC0 Carnera 8 Zoom Lens
Diffuser
Bottam Beamstop
Figure 3.1 - Beam Stop and Diffuser at RMC
Top View
Neutrons - /
Scintillation Screen
Camera
Figure 3.2 - Diffuser Used at Penn State
Except for the diffuser, the same camera systern was used at each site. The
equipment consisted of a 431.8 mm (17 in.) x 431.8 mm (17 in) aluminium backed
6 ~ i ~ : ~ n ~ : ~ u scintillating screen, an Apogee AP7 CCD camera with a SITe502 CCD
sensor, a front surface mkor with adjusting mechanism, a Vivitar AF-SC 28-70 mm, F3.5
zoom lens, a red laser source (used for focusing the lens), an Apogee video capture board
and a 200 Mhz personal cornputer running Windows 98. The SITe502 CCD is a back lit,
full h n e CCD chip with an anay size of 5 12 x 5 12 with a pixel size of 24p. The pixels
have a maximum well capacity of 296000 electrons and a dark curent rate of 0.8
electrons/pixeVsecond at -33°C. The CCD camera, cooling unit, laser, &or and mirror
adjuster are housed in a single unit as indicated in Figure 3.3. The CCD canera is
thennoelectrically cooled to an absolute temperature of -50°C, which corresponds to a
-30°C working temperature. Image acquisition was accomplished with ''Image Pro Plus"
software and the image analysis was completed with "IP lab Spectnim -Scientific Imaging
Software for Macintosh". All image analysis was done on an Apple Macintosh Cornputer,
Mode1 8500, as recomrnended by the software vendor.
Figure 3 3 - Camera Assembly and Control Box
There were six different sets of experiments used to analyse the potential
capability of the CCD camera system. The first three sets of experirnents detexmined the
characteristics of the system when no object was present in the neutron beam. The next
two sets of experiments indicated the performance of the camera when a component was
in the neutron beam. The last experiment compared neutron radiography with neutron
radioscopy. A flowchart of the major headings and sub headings of the various
experiments is shown in Figure 3.4.
Initially, software restrictions forced the CCD camera to be cooled to -15°C rather
thm -30°C and reduced the camera resolution to 5 1 1 x 5 1 1 vice 5 12 x 5 12. Throughout
al1 the experiments, the exposures were increased by one-minute intervals except for the
first one, which was only increased by 30 seconds. In Appendix B are indicated camera
temperatures, camera resolution and qmtity of images taken for each experimental
section. The normal operaing temperature for the camera was -30°C with a resolution of
512 x 512.
Installation and Optimisation of System at RMC Reduction of Ligbt Leaks
Configuration Parameters
1
Camera Parmeters Dark Image - Temperanne Change Dark Image - Diffuser Changes
Unified Images - Neutron Flux Effects Dynarnic Range
1 Resolution
Systern Resolution Calcuiations System Sensitivity il
r 7
Water Ingress Simuhtion
Water inpress Water Ingress - Single Cells Waîer Ingress - Three Cells Wata Ingress - Camera Gain
Water Ingress - Digital Enhancement
Simuiated Water Simuiated Water - Single CeIl Simuiated Water - Three Cells
l Comparison of Radioscopy to Rndiography Resolution
C m I Figure 3.4 - Flow Chart of Experiments
3 -2.1 Installation and Optimisation of the CCD Camera S ystem at RMC
3.2.1.1 Reduction of Li& Leaks
The CCD camera, bellows and scintillation screen were installed, and the camera
was visually lined up to the scintillation screen. As part of the optimisation processes, the
insides of the bellows and screen assembly were painted flat black to reduce light
reflections. Silicon sealant was applied to any crevasses or seams, which were by-
products of the manufacruring process. Opaque duct tape was applied to areas that could
not be pemianently sealed due to the disassembly and removal requirements of the
bellows. Not al1 fonns of duct tape are opaque and testing for opaqueness was mandatory
before use.
The reactor remained off and five images of various exposure times (referred to as
dark images) were taken before any changes where made to the bellows and screen
assembly. Light leaks in the system were eliminated by trial and error, until no M e r
improvement could be obtained. Another series of images were taken to compare its
histogram against the histogram of the original images in order to quanti@ the
improvement.
3.2.1.2 Confimation Parameters
Due to the configuration of the camera system, the bellows and scintillating screen
are removable as the scintillation screen occupies the location required for the vacuum
cassette used in radiography. To veri@ that light leaks have not been induced due to the
bellows installation, a method for detennining the quality of the system is required so that
consistent performance fkom the camera systern is achieved. After optimisation, 30 dark
images of a 2-minute exposure were taken to determine statistically the lowest pixel count
value of the centroid and the largest FWHM intensity values of the curve. These two
values indicate the minimum acceptable performance of the camera system. When
beginning to use the CCD camera, an image was taken with a 2-minute exposute and its
peak value and FWHM intensity values were compared against the statistical values
calculated in this section. An image not meeting these criteria uidicated a light leak or a
problem with the CCD carnera. A 2-minute exposure was chosen so that a reasonably
reproducible image could be produced in a relatively short exposure time.
3.2.2 Camera Parameters
The following experiments detennined the characteristics of the camera and
scintillation screen. Altering the diffuser, changing the temperature of the camera, varying
the exposure tirne and varying the flux level al1 affect the performance of the camera
system and the quality of the image. The above parameters were altered when the system
was not exposed to neutrons (called a dark image) and when the system was exposed to
neutrons (called a unified image). A unified image is a standard term used in digital
enhancement techniques and will be used throughout this work to indicate when no object
is placed between the neutron bearn and the scintillation screen so that the neutrons
uniformly illuminate the screen. If intensity values fiom a dark image and a unified image
are combined, the resdting histogram will show the dynamic range of the system.
3.2.2.1 Dark Image
With the reactor off, the temperature, di&er and exposure times were varied so
that the performance of the camera could be andysed. The temperature of the çamera was
altered fiom -1 5°C to -30°C, for varied exposure times.
For the experiments involving the different diffusers, the camera temperature and
resolution were held constant at the standard values, while the type of d i h e r and the
exposure times varied. The first series of exposures did not involve the diffûser. The
camera was placed in a photographic dark room and the lem covered. These exposures
represented the absolute darkest Mage that the carnera could record. The second series of
exposures were taken with the portable diffber used at Penn State reactor, while the last
series of exposures were taken with the fked diffuser at RMC.
3.2.2.2 Unified Image
Varyhg the neutron flux and the exposure time indicated the response of the
scintillation screen and the camera. With RMC's reactor at half power, a series of images
were taken with varied exposure times. At Penn State, images were taken when the
reactor was set at 1, 10 and 100 kW.
3.2.2.3 Dvnamic Range
The dynamic range was calcuiated by combining the data fkom the histograms of
the dark image and the unified image which produces an envelope indicating the size of
the dynamic intensity range for a given exposure, as shown in Figure 3 .S. The range is
calculated by subtracting the hvo peak centroid values, the k t peak centroid value from
the dark image and the second peak centroid value fiom the unified image. This span is
the dynamic intensity range of the system for a given neutron flux and exposure the
(Figure 3.5). When a component is placed in the neutron beam, there will be a range of
pixel intensities that range fiom a minimum value that corresponds to a dark image to a
maximum value that corresponds to a uninecl image.
Eumple of Dynuiilc Range C.lwht&n
LHS - Peak œmid 80000
mOaO
60000 - C
s - 2 p-
30000
2MXK) of udfied image I
lOOaO
O l a4 i2M r i 3 6 1782 aW8 23M 2560 2816 3072 3328 3584 3840 4096
Figure 3.5 - Example of Dynamic Range Histogram
3.2 -3 Resolution Measurernents
The system resolution was analysed for varying neutron flux and exposure times.
At the NRS at RMC, the MTF and a modified ASTM 545-75 Type A indicator were used
to determine resolution and sensitivity. Images of the two indicators were taken at various
exposure times varying fiom 30 seconds to 10 minutes. At Penn State, a Line Pair Gauge
(LPG) and a modified ASTM 545-75 type A indicator were imaged with the reactor set at
10 and 100 kW.
3.2.4 Water Ingress Simulation
3.2.4. i Test Piece
A s m d section of a CF188 rudder, measuring 14 cm x 14 cm was used as a test
piece for a series of experiments. The CF1 88 rudder is a composite construction that
consists of two outer skuis of carbon graphitdepoxy made fkom several plies of carbon
graphite material (1- to 3-mm thick) attached to an duminium honeycornb core by a FM-
300 adhe~ive,~* as shown in Figure 3.6. The top skin was removed so that the honeycomb
could be accessed. Afier the cells of the honeycomb were filled with distilled water, the
top was held in place with clear tape. The test piece was not as thick as the original rudder
as some honeycomb had been removed when the top skin was cut fiom the test piece. It is
estimated that the removal of the top surface reduced the height of the honeycomb by
approxirnately 0.5 cm ( 1 3% reduction).
Figure 3.6 - Construction of a CF188 Rudder
3.2.4.2 Procedure
For a single exposure tirne, a specific area of interest was extracted fiom the
original image. Areas associated with a set of single cells and a set of three cells were
analped. Thresholding was applied to the new image and statistical data such as mean,
minimum and maximum and standard deviation were calculated for each ce11 or cells
containhg water or High Density Polyethylene (PE). The intensity difference was then
calculated for each ceIl or cells containing water or PE.
The intensity difference for a representative volume of water was calculated for
either one or three cells for exposure tirnes that varied f h n 0.5 to 10 minutes. These
intensity difference values were then plotted against the exposure time. The amount of
error for these graphs was calculated fiom the standard deviation values for each ce11 or
cells. The optimum exposure t h e for the system was determined from the graphs for the
intensity difference of the representative volume of water versus the exposure time.
3.2.4.3 Water Inmess
All images of water ingress were taken at the NRS at RMC and accurately
simulate ingress into a flight control surface. The results fiom this experiment were used
to detemine the optimum exposure tune, minimum sensitivity, best camera gain setting
and best digital enhancement method. The effects of camera gain were not explored in
previous sections, as ampl img a "dark image" or "unified image" does not indicate the
effects that the gain will have on an image of an object. To simulate water ingress in a
fligbt control, distilled water was inserted into the test piece in the pattern and quantity
indicated in Figure 3.7. The test piece was imaged at various exposure times.
Further images of the test piece (Figure 3.7) were taken using a three-minute
exposure and changing the gain fkom O to 16, by increments of four. From past
experhents, increasing the gain incrementally by a small nurnber (a factor of one or two)
did not rnake a noticeable diffaence in the results. A factor of four is easily divided into
the maximum gain value and clearly showed the trend of the data for increased gain.
Utilizing the optimum exposure time and gain setthg for the test piece, the
exposures were subjected to a large number of enhancement techniques. The goal was to
create the best picture from a given exposure. The best image would include as many of
the features and as much detail as is lmown from the original object The details fiom the
original object were well understood, being a test piece with known characteristics that
were either implanted or were naniral for this structure. The enhanced images were
judged subjectively to indicate the best enhancement techniques.
Section of CF-18 Rudder PerpendiculYtaSlirhCeofRudder
Figure 3.7 - Test Piece with Water at RMC
3.2.4.4 Simulated Water
To compare the effects of changing the neutron flux on the test piece, water wuld
not be used as the test piece sat horizontally at RMC and vatically at Penn State. High
Density PolyethyIene (PE) was substituted for water. Al1 images taken at RMC had PE
and water inserted into the honeycomb of the test piece in the pattern and quantity
indicated in Figure 3.8. The results fkom the water were not used in this experirnent, but
the water was added to the test piece to take advantage of the
more data. The test piece with PE and water Iayout
for various exposure times.
Section of CF-18 Rudder Perpendicular to the Su- of R d e r
opportunity for collecting
in Figure 3 -8 was irnaged
Wter (mL) in a single ceil
.O 28.0 24.0 22.0 20.0 18.0 10.0 14.0 120 10.0 8.0 6.0 4.0 2.0
t
PE (mm) u-d- ,-,-: ,- in a single ceIl
'-- 18.0 20.0 10.0 14.0 120 10.0 Q.0 8.0 7.0 0.0 5.0 4.0 3.0 2.0
Aluminum Honeycom b
Upper and lower safaces mrnoved for danty
Figure 3.8 - Test Piece with PE at
AU images taken at Penn State had PE inserted into the honeycomb of the test
piece in the pattern and quantity indicated in Figure 3.9. With the reactor set at 10 and
100 kW, various exposures were taken at each power setthg.
Section of CF-18 Rudder Perpendiwlar to aie Surface of Rudâer
Upper and lowef surfaces removed for ciarity
Figure 3.9 - Test Piece with PE at Penn State
3.2.5 CF188 Rudder
A CF188 Rudder, senal
3 1 1 -NRT (revision 1) utilising
NRS is capable of mapping al1
radioscop y
8 positions
0241, was inspected as per technique MC- 188-
instead of radiography techniques. The RMC
of the rudder, as s h o w in Figure 3.10. At the
Penn State reactor, the same images were taken as per the technique, less exposures 1, 7
and 8. The beam tube at Penn State produced a round beam that did not cover the
complete area of the scuitillating screen, as shown in Figure 3.10 for exposure 5.
Therefore, taking images 1, 7 and 8 was not feasible. Image 1 is of the root of the rudder,
containing a large interna1 aluminium structure. Image 3, 7 and 8 are of the trailing edge,
located at the inboard end of the aircraft, and contain small amounts of honeycomb
structure. At1 three images are of less significance for water ingress inspection.
CF-18 Rudder
~ a d i o ~ r a ~ h k Film Positions 17in. xl4in.
Figure 3.10 - Fîim Layout on CF188 Rudder
3.2.6 Comparkon to Film
3.2.6.1 Resolution
The resolution of an image using radioscopy, SR film (the slowest, highest
resolution) and CX film (the fastest, lowest resolution) was derived fiom the MTF
technique and the modified ASTM E454-75 Type A SI.
3.2.6.2 Time and Cost Study
The cost of doing either technique can be divided into capitd costs and cost per
exposure. To simplifi calculations, al1 costs were assumed to occur initially and therefore
"the value of money" was not used. The concept of " t h e value of money" implies that
money has a value that varies depending on when it is received or disbursed. Capital cost
did not inchde items that are considerd part of the Facility, such as the reactor, the bearn
stops, the positioning systern and the radiation meters (installed, not portable). These
items are cornmon and essential for both neutron techniques and therefore were not
included in the cost study.
A CF 188 rudder was neutron radiographed at RMC as per technique MC- 188-3 1 1-
NRT (revision 1) and the time to complete the activities was recorded. These timings
were compared to the time requird to comp1ete neutron radioscopy using the sarne
procedure.
CHAPTER 4
RESULTS AND DISCUSSION
4.1 Installation and Optimisation of Diffiiser at RMC
4.1.1 Reduction of Light Leaks
Cornparing the histognims of the systern before and after optimising the RMC
diffuser (as described in Section 3.2.1.1) produced simila. trends for any given exposure.
A graph of a two-minute exposure illustrating the effects of improving the system is
s h o w in Figure 4.1. Two mmplete sets of exposure data fYom 30 seconds to four
minutes, collected before and after optimisation are, in Appendix C, Figures C. 1 and C.2.
The graph in Figure 4.1 shows the complete intensity grey scale fkom O to 65536. The
identical data are plotted in Figure 4.2 with the upper intensity scale value limited to
4096, to reveal the shape of the two curves.
Figure 4.1 - Full Intensity Scale of Optimisation
-- - -- -
Figure 4.2 - Enlarged Histogram Before and After Light Leak Reductions
The goal in optimising the system was to eliminate al1 light leaks and hence
obtain the darkest image possible. A dark image would produce a large thin peak with
an intensity value close to zero (black) i.e., to the left of the graph. Figure 4.1
dramatically shows that both peaks are situated to the far left of the graph with the peak
after optimisation taller and namower.
Without optimisation, the number of pixels at the peak centroid is relatively srnall
at 79,635, which was ody 30.5% of the total number of available pixels. The total
possible pixel count fiom a 5 1 1 x 5 1 1 array is 26 1 12 1 pixels. As well, the peak centroid
had an intensity value of 1799. Moreover, the non-optimised cuve is skewed to the right
of the peak centroid as shown in Figure 4.2. Full Width of the curve at Half Maximum
(FWHM)~' is a measurement of curve width. The non-optimised curve has a F W
intensity value of 656. The shape indicates that a large number of intensities are being
detected.
The peak centmid of the optimised cuve has a relatively large pixel count
(2 1 1,626) or 8 1 .O% of total pixels available which is a 62.4% increase in pixel count
relative to the non-optimised curve. The intensity value of the peak for the optimised
c w e is shifted left relative to the peak of the non-optimised curve &om 1799 to 1285
which is a 28.6% irnprovement in peak intensity value. Moreover, the optimised curve is
evenly distributed around the peak cenîroid. The optimised curve is wnsiderably thinner
than the non-optimised curve, with a FWHM intensity value of 280, a 57.3 % decrease in
cuve width.
The above analyses indicate that the work completed for optimisation has
improved the relative darkness of the system. This improvement is indicated by an
increase in the height of the peak centroid, shifhg of the peak centroid intensity value to
the left (darker intensity values), and narrowing of the curve (FWHM value reduced).
4.1.2 Configuration Parameters
Thirty dark images with a 2-minute exposure were taken for statistical purposes to
ensure consistency between radioscopy sessions (Section 3.2.1 2). M y the results from
the thuty images are graphed in Figure 4.3, as the original data are not shown for clarity.
The graph shows the minimum acceptable values corresponding to two standard
deviations (SD) from the average cuve (Section 2.6. l), indicating the envelope for an
expected configuration image.
Figure 4 3 - Statisticai Parameters of Dark Image
The statistical data fkom the thirty images are shown in Table 4.1. The tolerance
values were calcdated for one standard deviation. The pixel count was the only variable
that had a significant amount of error, which was calculated to be approximately 6 %.
This variation indicates the statistical nature and magnitude of the error associated with
the system. Taking thirty images produces a 90% probability that a new image will
produce sirnilar results with a 75% confidence leve170. Peak and FWHM values
indicating the limits for a configuration image are shown in Table 4.2. Future dark
images having peak and FWHM values falling within the statistical parameters indicated
in Table 4.2 would be considered valid and meet the required standard of darkness.
Table 4.1 - Statistical Data of 30 Images 1 Peak Values I LHS -FWHM I RHS -FWHM 1
Average intensity value
Table 4.2 - Configuration Image Parameters
Average Pixel count
Peak Value - Lower 2 Standard Deviaîion values
Pixel Count 1 Intensity
>=985 19
Average Intensity d u e
F m - Outer 2 Standard Deviation values
Pixel Count 1 LHS - 1 RHS -
Average Intensity value
Average Pixel count
Value
<= 1205
Average Pixel count
>=55315
Intensity Value
>=Il81
Intensity Value
<=1219
4.2 Camera Parameters
4.2.1 Dark Image - Temperature Change
The effects of camera temperature were analysed by cornparhg the data taken at
two camera temperatures, - 1 SOC and -30°C (Section 3.2.2.1). Comparing the two
temperatures produced similar trends for any given exposure; thus a graph of a 2-minute
exposure, for example, can be used to illustrate the effects of lowering the camera
temperature, Figure 4.4. Note that the intensity interval was changed fiom 256 (Figure
4.2) to 16, which produced a graph with greater detail. Therefore, the size and values
associated with the -1 5°C c w e in Figure 4.4 do not correspond to the previous curves at
this temperature (Figure 4.2). Two complete sets of exposure data collected at -1 5°C are
shown in Annex C, Figure C.3 and at -30°C are shown in Annex C, Figure C.4.
Figure 4.4 - Effects of Camera Temperature
At the camera temperature of -l5OC, the peak centroid pixel count was relatively
small at 22,8 15 +1369 (6%) which was only 8.7% of the total nurnber of pixels available.
The intensity value associated at the peak centroid was 1397. The -lS°C graph is almost
symmetrical around its peak value and has a FWHM intensity value of 156. n i e shape of
this curve indicates that a large number of intensities are being detected.
Lowering the camera temperature to -30°C produced a graph with a peak centroid
pixel count of 108,170 k6490 (6%) pixels which was 58.6% of the total pixels available
and was an increase in value of approximately four times that of the -15°C curve. The
location of the peak centroid had an intensity value of 1205, which was a 13.7%
Unprovernent in peak centroid intensity value. The "FWHM" for the -30°C curve has an
intensity value width of 34, producing a 78.2% decrease in curve width at the FWElM
value. The analysis has shown that, by reducing the temperature of the CCD camera, the
exposures becarne darker due to the increased peak centroid pixel value, the FWHM was
reduced (narrower curve) and the peak centroid intensity was shifted towards the darker
scale. The results obtained by l o w e ~ g the camera temperature were similar to the
reçults obtained by optimising the carnera system (elirninating light leaks). Reducing the
temperature of the camera reduces the quantity of dark current, which in turn reduces the
noise associated with the image.
A set of 11 images was taken without a diffuser and then a set of images was
taken using the tùted diffuser at RMC and the portable diffuser at Penn State (Section
3.2.2.1). The histograms associated without a di-, fixed diffuser and portable
diffuser are shown in Appendix C, Figures C S to C.7, respectively. Although the camera
had a slightly lower resolution of 5 1 1 x 5 1 1 for the images taken at RMC, the cornparison
is still valid. This slightly lower resolution will not affect the results when comparing
them to the other two histograms. The peak intensity value for each image fiom al1 three
histograms was plotted on a graph of intensity and exposure time in Figure 4.5. The
results fiom this graph were used to determine the effects of the diffuser on the camera
system.
Comparlron benneen DWu+ers Udng Peak htenslty Values 30 Second to 10 Utnute Exposure
@ 3 4 C . 5 1 2 ~ 5 1 2 ~ * r a u 1 I y i n t e n e i o l 1 6
Figure 4.5 - Peak Intensity Values of Dark Images for Different Dinusers
In general, as the exposure time increases for any given diffuser, the intensity
value increases, going fiom dark to lighter. Also the images without a diffuser produced
the darkest values for a given exposure time, followed by the portable diffuser used at
Penn State and the fixed diffiser used at RMC. The error associated with quantization of
intensity values is very low and, when combined with a relatively large intensity interval
of 16, the e m r becomes insignificant. It is assumed therefore, that the error associated
with the data in Figure 4.5 is less than 1%.
The reasons that dl three curves increase in intensity value, as exposure time
increased, is due to the image being integrated over tirne and the accumulation of dark
current. The images without the diffuser behaved as expected, having the darkest images
of al1 three diffusers, as no surfaces for the light to reflect are available, which would
cause an increase in image intensity. Producing an image with an intensity value of zero
is not obtainable due to system noise and dark current. The lowest the camera system can
achieve is a peak intensity value of 1 140 (Section 4.1.1). The scintillation screen, which
is part of the diffuser and is perpendicular to the camera's field of vision, causes some
reflection of light; hence, the increase in intensity values. The bellows assembly attached
to the RMC diffuser most likely contributes to the still higher intensity values. The
bellows assembly has different absorption and reflective properties than the sides of the
diffuser and the probability of a very small light leak in the bellows assembly is higher,
causing the highest peak intensity values.
4.2.3 Unified Image - Neutron Flux Effects
A number of images were taken with the two diffusers without an object in place
and at four different reactor power settings (Section 3.2.2.2). At RMC, the images were
taken at half power (10 kW) and at Penn State the images were taken at 1, 10 and 100
kW. The RMC power settings and the 1 kW settuigs were sirnilar in flux magnitude at
the image plane so that the screen output could be compared. These series of images
were taken to investigate the response of the saeen to higher neutron interaction and to
determine how varying the Iength of the exposure affects the image at these higher flux
values. n i e exposure times were 0.5, 1, 2, 3, 4 and 5 minutes. The histogram of the
unified images fkom RMC (10 kW) and Penn State at 1, 10 and 100 kW are shown in
Appendix C, Figures C.8 to C. 1 1, respectively. The peak intensity values for the
different exposure times are graphed in Figure 4.6. The results f?om this graph were used
to determine the effects of neutron and screen interaction on the camera system.
Peak hlensity Yakes from Unitïed hrrges for OHkrent Ffux VaIues at Vulous Expouire Thes
@ r )C . 512x 512fusaUim imemityintmaîoi 16
- - - - - - -- --
Figure 4.6 - Peak Intensity Vaiues of Unified Images at Various Fiux Values
As the exposure time increases, the intensity of the peaks ïncreases, hence the
positive slope of the lines (Figure 4.6). This trend is similar to the trend associated with
the varying exposure &es of the dark images (Section 4.2.2). Three variables contribute
to this effect, dark current, screen illumination and integration of the image, with
integration being the dominant effect among the three. As well, as power increases (Le.,
flux increases), the peak intensity values increase, as indicated by the increase in the
slope of the line for each power level. Note the large increase in the slope for the 100 kW
value for Penn State. The exception being that the line associated with the data fiom
RMC (at a flux of 2 to 3 x 104 n/cm2 s) was above the Penn State 1 kW values (at a flux
of 3.0 x lo4 n/cm2 s). There are three possibilities f ~ r this discrepancy. For Penn State, it
was assumed that the flux was linear fiom the 100 kW value. As well, the power setting
at Penn State can vary by 5% and still be within normal operating parameters. For RMC,
the flux could be greater than previously measwed.
Figure 4.6 also shows that the half power setting at RMC (10 kW) and the 1kW
power level produced very similar results and therefore the characteristics of the RMC
system were reproducible.
Reviewing tfie images at the highest power setting at Penn State, the image
became visible when the peak intensity value was greater than approximately 10,000.
Therefore, the images at the lOOkW power setting above one-minute exposure would be
visible without image enhancements. Note that, at the 100 kW power setting with a 5-
minute exposure, the image becornes saîurated. At the power setting of 10 kW, images
exposed for approximately 7 to 8 minutes would start to becorne visible without image
enhancement. Extrapolating the data fiom the RMC NRS, it would take approximately
500 minutes to reach saturation and 69 minutes for the intensity value to be greater than
10,000. Therefore, image enhancanent would be required for any reasonable exposure
time at RMC.
The dynamic range was calculated (Section 3.2.2.3) and graphed for different
exposures times taken at RMC at 10 kW (Appendix C, Figure C.12) and at Penn State at
1kW (Appendix C, Figure C.13), at 10 kW (Appendix C, Figure C.14) and at 100 kW
(Appendix C, Figure C. 15). Since the trend was sirnilar for different exposure times for a
given power setting, a representative graph showing only the values of the dynamic range
was fompiled for a 3-minute exposure time fiom each power seîting (Figure 4.7).
Dynamlc Rang. m m Peak Centroid vakwas tor Duk and U n W lmrges
Reactot Power Setthgr
Figure 4.7 - Dynamic Range of System at Various Power Settings
The dynamic range of the system increaz;ed dramatically when the flux was
increased. The dynamic range at 1 kW was extremely nmow with an htensity value of
337. The dynamic range at RMC had an intensity value of 963. Even though the flux
levels for both these systems was assumecl to be similar, the dynarnic range at Penn State
at 1 kW was 65% smaller, indicating that the calculated flux value for RMC at 10 kW
(assuming linear relationship) might be approximately 8 x 104 n/cm2 s instead of the
rneasured value of 2 to 3 x 104 n/crn2 S. The dynamic range for 10 and 100 kW were
3,584 and 36,196, respectively. As previously stated, intensity values above 10,000 are
visible without image enhancements and therefore it is expected that a 3 min exposure at
100 kW would contain not only a visible image but have a full range of intensities.
Having a large dynamic range indicates the potential to record small intensity differences
between two objects. Alternately, having a small dynamic range wouid mean that the
intensity difference between two objects could not be identified.
4.3 Resolution Calculation
The system resolution was calculated using the MTF method and a Iine pair gauge
(LPG). A modified ASTM E545-75 Type A was used to detemiine the sensitivity of the
system (Section 3.2.3).
4.3.1 S ystem Resolution Cdculations
At RMC, the image plane was extended to the fully down position and al1
exposures of a knife-edge gadolinium foil were taken. At this position the pixel size was
calculated to be 0.73 rnm/pixel or inversely 1.37 pixels1mm. The analytical software
recorded the densities across the Gd foil and an Edge Response Function (ERF) graph
was created f?om these densities. The Ut (total unsharpness) was measured fkom the ERF
graphs corresponding to a 0.5, 4 and 10-minute exposure. The Ut values and calculated
cycle/mm values with absolute and relative mors for the three exposure times are shown
in Table 4.3.
Table 4 3 - Resolution Variables for the NRS at RMC
. 10 minute 1 1.37 1 1.50 + 0.50 (33%) 1 0.44 (+ 0.2 11 -0. 12j
Exposure time 30 second 4 minute
Using the pixeVmrn and Ut values for the 4 minute exposure tirne, a MTF graph
was calculated and is shown in Figure 4.8.
pixeVrnm 1.37 1.37
ut (mm) 1.35 L- 0.15 (1 1%) 1.40 4 0.60 (42%)
Cycles1rnm at MTF Values of 10 % 0.50 (+ 0.041 - 0.07) 0.47 (+ 0.341 -0.14)
Figure 4.8 - MTF Graph for 4-Minute Exposure taken at RMC
The resolution does not seem to be time dependent, as the resolution values
measured in Table 4.3 are within the calculated error. At the 30-second exposure time,
the intensity values between the background and gadolinium foi1 did not Vary
significantly. This lack of intensity range makes this exposure time unacceptable. The
10-minute exposure time had a large intensity variation with considerable system noise,
which produceci the largest Ut and smallest resolution measurernent. At the optimum
exposure time of 4 minutes, a system resolution c m be derived from Figure 4.8 that has a
range of 0.8 1 to 0.33 cycles/mm with an average resolution of 0.47 cycles/mm at a 10%
MTF. These values translate into a range that varies f?om 1.23 mm/cycle to 3.33
mdcycle with an average of value of 2.13 mm/cycle. Dividing these values in half, the
system resolution would range fÏom 0.61 mm to 1.66 mm with an average system
resolution of 1 .O6 mm.
System resolution was also measured at Penn State using a LPG. The images of
the line pair gauge were taken at two diEerent power settings and for varying exposure
times. A typical image of the LPG taken at 100 kW for a 5-minute exposure is shown in
Figure 4.9.
Figure 4.9 - Line Pair Gauge
The system resolution value is directly read from the image, and the resolution
value is determined by the smallest set of altemating dark and light lines. The resolution
value fÏom the LPG was neither power setting nor exposure time dependent. The system
resolution was measured at 0.656 Ip/mm or 1.5 m d i p as seen in Figure 4.9. The unit
line pair (lp) is identical to the unit cycle used in the MTF method.
At Penn State, the system resolution was higher (0.656 Iplmm) than the average
system resolution at RMC (0.47 cycles/~nm). The LPG seemed insensitive to changes in
power setting and exposure time. As well, the MTF method was also insensitive to
different exposure times.
Exarnining the images of the ASTM E54S-75 Type A standard taken at RMC and
at Penn State for various exposure tirnes produced the following results as indicated in
Table 4.4. Note that the power level at Penn State (1 kW) comparable to RMC (1 O kW)
was not taken.
Table 4.4 - Hole Values from Modified ASTM E545-75 Tme A
Exposure Time
I 10 I 7 I 11 Note: Vdue from RMC (10 kW) are not M y comparable to Aues from Penn State (10 kW).
~ 6 b e r of Holes Observed at RMC 1 Penn State 1 Penn State
I
The holes visible from the RMC images produced an averaged value of 6 out of
-
(minu tes) 1 (10kw) 1 (10 kW)
16 compared to the images taken at the two higher power Ievels at Penn State, both of
(100 kW)
which had a consistent value of 1 1 out of 16. The flux intensity is directly proportional to
the amount of holes visible in the sensitivity ùidicator. Using the ASTM E545-75
standard," images taken at RMC had a R6 rating and Penn State images had a RI1
rating.
4.4 Water Ingiress Simulation
Without an object in the neutron beam, the camera's charactexistics did not
indicate the system's performance. Using the NRS at RMC, an experiment was carried
out using distilled water and a honeycomb test piece, which simulated water ingress into
a fiight control surface. These two items were used to determine the camera's
capabilities at a low flux level. To determine and compare the effects of different flux
levels on the camera system, an experiment using PE (hi@ density Polyethylene) in the
same test piece was used at both RMC and Penn State.
4.4.1 Water Ingress
Al1 water ingress images were taken at RMC with the reactor at half power, 10
kW. To analyse the best exposure tirne, two tests were made: the k t for single celIs of
water and the second for three adjacent cells of water. In boîh tests, the amount of water
in each ce11 or set of cells was varied. The results fiom the experiments were used to
determine the optimum exposure t h e , minimum sensitivity, optimum camera gain
setting and best digital enhancement method obtainable at RMC.
4.4.1.1 Water Ingres - Single Ce11
Filling single cells with incremental voIumes of distilled water simulated the most
strenuous situation to be encomtered during radioscopy of a flight control surface, due to
the smali quantity of water and small surface area being presented. Varying the mount
of water also gave a good indication of the sensitivity of the radioscopy system for
detecting water ingress in a single cell. The intensity difference denned in Section 2.9.2
was calculated for each ceIl for a given exposure time and then graphed in Figure 4.10. If
the camera system is sensitive enough to record small variations in neutron absorption
and scatter, then the intensity difference should hcrease with increased volumes of water
for a given exposure time. As well, integrating an image over time should cause the
intensity difference to increase as the exposure time increases.
-- - ro mn
-9 min a min 7 m n
6 mm - 5 mn -4 min 4 - 3 min -2 min -1 min 4 -30 a
O i O 0.02 0.04 0.06 0.08 0.1 û.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3
Volume of Water in Ceil (ml)
Figure 4.10 - Effects of Varying Exposure Tirne on a Single Ceïi of Water
Error Calculations For clarity, no error bars are shown in Figure 4.10. Since the
cell intensity is an average value for the cell, the error for a ce11 would be the value + two
standard deviations. Since thresholding (Section 2.7.2.1) was used to detennine the
average intensity values, the error associated with each ceil is relatively small. For mal1
volumes of water, the error is approximately 1% and for large volumes of water,
approximately 15%. As part of the standard thresholding practice, thresholding was not
applied to the background area and therefore system noise increased the error associated
with the background intensities. The error associated with background intensities can be
as high as 60%. Using standard =or calculations produces a relative error of
approximately 1 1% to 5 1 % of the intensity difference fiom the mallest volume and
shortest exposure t h e to the largest volume and longest exposures. For example, at a 5-
minute exposure for 0.04 ml of water, a relative error of 36% was calculated; therefore,
the intensity difference value with errors wodd be 19% f 6% (36%). An error calculation
of this example is included in Appendix D.
G r a ~ h Features In general, as the exposure tirne and the volume of water
increases so does the intensity difference. The intensity difference curve for a given
exposure is positively sloped, indicating a correlation between the volume of water in a
ce11 and the intensity difference. The shortest exposure time has the lowest intensity
difference values while the longest exposure time has the largest intensity difference
values.
The ce11 that contained 0.02 rnL of water produced poor results with exposures
greater than 4 minutes. There was such a small quantity of water in the ce11 that it did not
cover the ce11 completely, making it difficult to threshold this area properly. For a given
exposure time the data values tend to oscillate between the measured volumes of water.
The oscillation is well within the error band and is the result of thresholding and data
pmcessing and does not seern to be related to any physical phenomena related to the
system.
There seems to be two bands of data - the 30 second to 2 minute exposures and 3
to 10 minute exposures. This second band tends to be much wider in intensity difference
values. Analysing the raw data of the intensities the for 2-minute and 3-minute exposure
times, the intensity of the water in the cells increased linearly, but the difference in the
empty ce11 background intensities increased dramatically. The empty ce11 measurement is
the intensity value fkom the scintillation screen after absorbing and scattering neutrons
through the test piece's skins, adhesive and honeycomb. During short exposure times
(less than 3 minutes), the test piece is acting a s a filter and lowering the light intensity,
causing the first band of data, until more neutrons are available to pass through and be
detected, causing the second band of data.
From Figure 4.10, because of the sunilar behaviour of al1 volumes of water, any
volume would give sùnilar results therefore, the intensity difference value for a chosen,
and possibly representative volume of water (0.14 mL) was plotted against exposure time
and the results are s h o w in Figure 4.1 1. The mor for each ce11 was calculated fiom two
standard deviations as previously describeci. The error value ranged from 12% for a 30-
second exposure to 47% for the 10-minute exposure as shown in Figure 4.1 1. A sample
calculation of the error is included in Appendix D. Figure 4.1 1 will be used to detennine
the optimal exposure time.
intensity Drtferecrce kr a Single Cell of W m r atû.l4rnLOemsdtanDataib<enarFWC
Figure 4.11 - Intensity Dinerence of a Single Ceii at 0.14 mL of Water
In general, the intensity difference increases as the exposure tirne increases. An
increase in intensity difference of about 20% occurs fiom 0.5 to 4-6 minutes of exposure
tirne, while about 5% of intensity d i f fmce occurs nom 4-6 to 10 minutes of exposure
time. From the MTF resolution calculations, a 10% difference in intensity is the
minimum threshold for human vision and therefore, the optimum exposure time for a
single ce11 of water would be 4 or 5 minutes. The camera system has the sensitivity and
capability of responding to a difference of 0.02 mL of distilled water and capable of
responding to a minimum volume of 0.02 mL in a single cell.
4.4.1.2 Water Ingress - Three Cells
Three cells containing distilled water simulate a typical water ingress situation in
a flight control surface. Varying the amount of water for each set of three cells gave a
good indication of the response of the radioscopy systern for large volumes of water. The
intensity difference was calculated for each group of cells for a given exposure tirne and
then graphed in Figure 4.12. It is expected that, as the exposure time and water volume
increase, so will the intensity difference value.
3 Celh of W8?ar vs htsnsity Dtffenncs al VuylnO Expouire Times -iakriatwcatwPower
Dy-
O am a- aos am 0.1 a12 a i 4 ais ais 0 2 ozz 428 428 0 3 0.34 o s Volume of W i b r (mL)
- 10 min -9 mtn
-8 mm -7 mm 6 min -5 min -4 mm -3 min
-2 mm - 1 min
-30 s
Figure 4.12 - Effect of Varying the Exposure Time for Three Cells
The characteristics of Figure 4.12 are simila. to the single ce11 graph, Figure 4.10.
Using standard error calculations produced a relative error of approximately 4%
(corresponding to 0.02 rnL of water at a 30-second exposure) to 40% (corresponding to
0.34 mL at 10 minute exposure) of the intensity difference. The relative error was
smaller for three cells of water compareci to the relative error for the single cells. The
increased surface area of the three cells makes the detection of water easier and produces
more accurate resuits, thereby causing a reduction in relative error.
The cells that contained 0.02 mL of water produced good results with all
exposures times. The oscillation between volumes of water was also reduced, another
benefit fiom the increased area of the cells.
As in pnor experiments, there were two bands of data - the 30 second to 1 minute
and the 2-10 minute exposures. This second band was much wider in its intemity
différence values. Sunilar to the single ce11 of water, the difference in background
intensities increased dramaticall y between the one-minute and two-minute exposure tirne,
while the intensity of the cells increased linearl y.
As the exposure tirne and volume of water increased, the intensity difierence
value increased. The intensity difference curves for a given exposure are positively
sloped, indicating a correlation between the amount of water in a ce11 and the intensity
diffaence. In general, the intensity difference values are lower for the three cells
compard to the intensity difference values for a single ce11 with the same exposure time.
It was easier to identiw several cells of water due to the increased surface area, even with
slightly lower intensity difference values. It is assurned that the effects of neutron
scattering caused the lowering of the intensity difference.
The intensity difference value for a representative volume of water (0.16 mL) was
plotted against exposure time and the results are shown in Figure 4.13. The emor for each
ce11 was calculated fiom two standard deviations as previously described. The error
ranged fkorn 7% for a 30-second exposure to 38% for the 10-minute exposure. Figure
4.13 will be used to determine the optimal exposure t h e for three cells of water.
htnsity Diffemtnce for 3 CeUs at0.16mLDaiadIrom DotPtîknat FIMC
Figure 4.13 - intensity DWerence of Three Cells at 0.16 mL of Water
The data has been presented in the same manner as in Figure 4.11. In general, the
intensity difference value increases as the exposure time increases. An increase in
intensity difference of about 20% occurs fiom 0.5 to 5 minutes of exposure tirne, while
about 5% of intensity difference occur fkom 6 to 10 minutes of exposure time. The error
associated with the last four exposures (7 - 10 minutes) is rather large due to the
increased noise of the image background, which is caused by the integration of the image.
The three longest exposures produced the largest intensity difference values but they also
produced the largest amount of image noise, which might cancel the benefits of the
increased contrast. If a 10% reduction in intensity difference were assumed, as this
amount of contrast is not detectable, then the optimum exposure for three cells of water
would be either 4 or 5 minutes. Since three ceils increase the contrasted area on the
image, it is easier to accept less than optimum intensity difference values in order to keep
the exposure time to a minimum.
4.4.1 -3 Water hgress - Camera Gain
To determine the effects of altering the digital gain of the camera, a . analysis was
conducted in the same manner as the tests descnbed in the Sections 4.4.1.1 and 4.4.1.2.
The gain was increased fiom a value of 1 to 16 by increments of four. From past
experiments, increasing the gain incrernentally by a srnall nurnber (a factor of one or two)
did not make a noticeable difference in the final results. Both single cells of water and
three cells of water were analysed as described in the previous sections. To determine the
effect of a gain increase, a single exposure tirne of 3 minutes was selected.
4.4.1.3.1 Water ingress - Camera Gain - Single Cells
Single cells in the test piece were incrementally filled with distilled water and
imaged at different camera gain settings. The intensity difference was calculated for each
ce11 for a given exposure time and then graphed (Figure 4.14).
O J
O 0.05 O. 1 0.15 0 2 0.2s 0.3
Volume of Wabr In Cell (mL)
-4xGain 8 x Gain -m- 12 x Gain +Or Gain -16 x Gain
Figure 4.14 - Effects of Camera Gain on a Singie Celi of Water
The relative error for the above data ranges fiom 21% for the srnaIl volumes of
water and increases to 35% for the larger volumes of water. Al1 curves, except for the
largest gain of 16, were within each other's error band and therefore statistically, they are
identical. A large decrease in intensity difference was recordecl when the gain was set to
16. This decrease in intensity ciifference was produced fiom the increase in analogue
noise, which caused saturation of the exposure, making it difficuit to threshold a single
ce11 of water.
When the large relative error is taken into account, the benefits associated with
increasing the gain for a single ce11 are negligible and increasing the gain to maximum
limits had a detrimental effect.
4.4.1.3.2 Water hgress Camera Gain - Three Cells
Three cells of water in the test piece were imaged at different camera gain
settings. The intensity difference was calculated for each ce11 for a given exposure tirne
and then graphed (Figure 4.15).
O L O 0.05 O. 1 0.15 02 0.25 0.3 0.35 0.4
Volume of Waîer In Ceüs (mL)
Figure 4.15 -Effects of Camera Gain on Three Celis of Water
The relative ermr for zero gain to a factor of 12 ranged fiom 18% to 26%. The
relative error for the Mages at a gain of 16 ranged fiom 45% to 57%. This large error is
causeci by image saturation. For three cells, the effect of increasing the gain up to a
setting of 12 is negligible, producing neither benefit nor detriment to the intensity
difference for three cells. This conclusion is based on the close grouping of the curves in
Figure 4.15.
Overall, there is no visible benefit fiom increasing the digital gain of the carnera
as the results have shown that the intensity difference values for both the single and three
cells of water are negligible when the relative error is considered.
4.4.1.4 Water I n ~ e s s - Digital Enhancernent
Previous analysis (Section 4.4.1.1 ) has shown that the optimum exposure time for
the test piece was 4 to 6 minutes; therefore, that image of the test piece will be digitally
enhanced using several methods and then the results will be subjectively analysed. Since
digital enhancernent alters the original information, analysing histogram or intensity
clifferences proved futile in deteminhg the best image. The method used in detennining
the optimum image after enhancements will be a subjective cornparison of hown
features from the test piece to identifiable objects in the image. The image will be
analysed for overall image clarity, amount of background noise (random white spots),
presence of the honeycomb pattern, presence of water, and variations in intensity due to
volume changes in water. The following types or methods of enhancernent techniques
were anal ysed:
Contrast, gamma and brigbtness, Equalisation of histogram, Erode filter, J Flat filter, Median filter, Mu1 tipl ying images, Ratio flat, and Optical density method.
A11 eight enhancement techniques were applied and the results are presented in
Appendix E. Due to the grey scale limitations of the printer and size of image, the
images in Appendix E have poorer resolution on paper than on a computer monitor. Only
the original image of the test piece and the best-enhanced images will be discussed. The
best enhancement methods producing the best images when compared to the original
image of a 4-minute exposure (Figure 4.16) were the equalisation of histogram (Figure
4.17), the Erode filter (Figure 4.18) and the J Flat filter (Figure 4.19).
a) 4 Minute Exposure b) 4 Minute Exposure w/Contrast Changes
Figure 4.16 - Four-minute Exposure of Test Piece
The original image is black as shown in Figure 4.16a and, with contrast
manipulation (Figure 4.16b) the test piece becomes visible. The minimum water volume
visible is 0.06 mL for the single cells and 0.04 mL for the group of three cells. There is a
considenble arnount of noise in the image, making fine details difficult to see.
a) Equalisation of Histogram b) Equalisation of Histograrn W/ Contrast Changes
Figure 4.17 - Results from Equalisation of Histogram
When the original image is enhanced by the equalisation of the histogram
technique, the image has a significant arnount of noise (in the f o m of dark grey areas)
(Figure 4.17a) but, when contrast is applied (Figure 4.17b), the noise is reduced and the
honeycomb pattern is apparent. The minimum volume of water visible is 0.04 mL for the
single cell and 0.02 mL for the three cells. A significant amount of contrast enhancement
was required to reduce the noise and this contnst washed out the background of the
image leaving only the outline of the test piece.
a) Erode Filter b) Erode Filter w/Contrast
Figure 4.18 - Results from Erode Filter
Applying the erode filter does not lighten the image as seen in Figure 4.18a and
therefore contrasting was applied to produce a visible image in Figure 4.18b. Al1
volumes of water were detectable and the noise is reduced considerably. The outline of
the test piece is recognisable, but the honeycomb pattern is not visible. This is the best
enhancement technique for detecting water in the structure. as it is capable of showing
the various volumes of water.
a) J Flat Filter b) J Flat Filter w/Contrast
Figure 4.19 - Results from J Flat Filter
Utilising the J Flat Filter produces a visible image without applying contrast
enhancements as seen in Figure 4.19a. Both images tend to show a large amount of
noise. In Figure 4.19b, the minimum amount of water visible for a single ceil is 0.04 mL
and for the three cells is 0.03 mL.
The equalisation and erode filter are the easiest of the three enhancement
techniques to implement as they are usually integrated in a software package while the J
flat filter requires that another image be taken and added to the original. Of the three
techniques, the erode filter is the best enhancement technique for water detection.
The results from the above experiments were used to determine that:
a the optimum exposure time for the test piece containing water is 4-6 minutes;
b. the minimum volume of water detectable is 0.02 mL;
c. the minimum detectable incremental volume of water is 0.02 mi,;
d. the optimum carnera gain setting is zero; and
e. the best digital enhancement method for water detection is the erode filter.
4.4.2 Simulated Water
To determine the effects of different flux levels, the same test piece was used but
PE replaced the water. Images were taken at RMC at half reactor power (1 0 kW), and at
Pem State at 10 kW and 100kW. The images were processed in the same manner as in
Section 4.4.1 which produced graphs with intensity difference values for one ce11 and
three cells of PE. Only the data for the 10 mm PE will be analysed to determine optimum
exposure time and effecr of neutron flux changes. Similar to the other sections, single
cells and grouped three cells will be analysed.
4.4.2.1 Simulated Water - Single Cell
The results fiom calculating the intensity difference for a representative height of
PE (10 mm) are shown in Figure 4.20. The calculation of the relative error for Figure
4.20 was calculated in the manner as described in Appendix D. Inserting the relative
error onto the graph made it difficult to read and, therefore, the average relative mor is
indicated in the legend.
rtne (min)
Figure 4.20 - Results of Varying Flux Levels for a Shgle Ceii of PE
In general, al1 three curves increase with exposure time. The curve nom the RMC
data increases constantly until becoming asymptotic at the 9-minute exposure time. The
beginning of the RMC curve increased about 37% for the first 5 minutes and then raised
o d y 13% for the remaining 5 minutes, making a total intensity difference increase of
approximately 50%. Accounting for image noise and the lack of ability to differentiate
between 10% contras4 an optimum exposure time would be 6 minutes.
The c w e fiom the Penn State data (10 kW) increased rapidly and became
asymptotic at a 5-minute exposure time with the dope of the cuve increasing again at the
10-minute exposure time. The beginning of the Penn State (10 kW) curve increased
about 45% for the £kt 5 minutes and then raised only 1% or 2% for the remainùig 4
minutes (the 1s t data point is ignored) making a total intensity difference increase of
approximately 47%. Accounting for image noise and the 10% contrast factor, an
optimum exposure thne would be 3 minutes.
The cuve fiom the Penn State data (100 kW) increased rapidly and becornes
asymptotic at the 4-minute exposure tirne. Applying the same logic as before, (noise and
10% factor) an optimum exposure tirne would be a 2-minute exposure.
Both lower fiux values (RMC and Penn State (10 kW)) produced the largest
ïnmease in intensity difference over the 5 minute range. The data fiom W C produced a
37% increase in intensity difference while the data fiom Penn State (10 kW) produced a
45% difference over 5 minutes, cornpared to data fkom Penn State (100 kW) which
produced an approximately 20% increase.
Two trends are evident due to the increasing flux level. First, the optimum
exposure time for the flux level at RMC (2 to 3 x 104 n/cm2 s) would be 6 minutes, but
decreases to 3 minutes for the 10 kW power setting (3.0 x 105 n/cm2 s) and decreases
M e r to 2 minutes for the 100 kW power seîthg (3.0 x 1 o6 n/cm2 s); therefore, as flux
inmeases, optimum time decreases. Second, the intemity difference value for a given
exposure time increased with increasing flux values.
4.4.2.2 Simulateci Water - Three Cells
The r d t s from calculating the intensity difference for a representative height of
PE (10 mm) are shown in Figure 4.2 1. The method for calculating the relative mor and
its placement is identical to the graph from the previous section (4.4.2.1).
hmnsity Difbr9n~9 for Varying Flux Values (hgQpdlOmm(pr3CaIsdPE)
-- -
Figure 4.21 - Results of Varying Flux Levels for Three Cells of PE
As previously observed, al1 of the curves inaease as the exposure tirne increases.
As well, the curves are becoming asymptotic with the longest exposure tirne. As seen
previously (in Figure 4.20), the intensity difference increases with exposure the. The
poor fit of the curve to the Penn State (100 kW) data is due to the large variations of the
intensity difference values caused by the large mor (39.3%) associated with these data.
For a given curve, the intensity difference values are less than the same curve from the
single ceil graph in Figure 4.20.
The curve fiom the RMC data increases constantly, but does not appear to
becorne asymptotic at the 10-minute exposure tirne. The beginning of the RMC curve
increased about 26% for the first 5 minutes and then rose only 12% for the remaining 5
minutes, making a total intensity difference increase of approximately 3 8%. Accounting
for image noise and the lack of ability to differentiate between 10% contrast, an optimum
expotaire time would be 7 minutes.
The curve from the Penn State data (10 kW) increased rapidly and became
asymptotic at a 5-minute exposure time with the cuve increasing in slope again at the 10-
minute exposure tirne. The second nse in the curve is from fitting a third order
polynomial c w e to the data. This conclusion is derived fiom the fact that only the Penn
State (10 kW) cuves (for the single ce11 and three cells) have this characteristic and,
therefore, this anomaly should be ignored. The beginning of the Penn state (10 kW)
cuve increased about 45% for the f3st 5 minutes and then rose only 1% or 2% for the
remaining 4 minutes (the 1s t data point is ignored) making a total intensity difference
increase of approximately 47%. These values are the same as the single ce11 PE for this
power setting. Accounting for image noise and the 10% contrast factor, an optimum
exposure t h e would be 3 minutes.
The curve fiom the Penn State data (100 kW) increased rapidly and becomes
asymptotic at the Cminute exposure time. Applying the same logic as before (noise and
10% factor), an optimum exposure time would be 2 minutes.
The lower flux value fiom Penn State (IO kW) produced the largest increase in
intensity difference over the 5 minute range. Compared to the data fkom RMC which
showed approximately 26% increase in intensity difference, the data fiom Penn State (10
kW) showed approximately 45% ciifference and that data fiom Penn State (100 kW) had
approximately 22% increase.
The optimum exposure time for the flux level at RMC (2 to 3 x lo4 n/cm2 s)
would be 7 minutes, but decreases to 3 minutes for the 10 kW power setting (3.0 x 10'
n/cm2 s) at Penn State and decreases M e r to 2 minutes for the 100 kW power setting
(3 .O x 1 o6 n/cm2 s). Therefore, as flux increases, the optimum time decreases.
There was a good correlation between optimum exposure times for the single ce11
and the three cells for a given flux level. The small increase of one minute to the
optimum exposure at the lowest flux level for three cells of PE compared to a single ce11
is not significant and is within the experimental error. Since the target is larger than the
single cell, reducing the exposure time to one minute would not degrade the image
contrast significantly and therefore an optimum exposure t h e for the three cells would
be 6 minutes.
Overall, as the flux increases, the intensity difference, which is the contrast
between the PE and the test piece, increased significantly, for both the single and three
cells. The optimum exposure tirne decreased as flux values increased £kom seven minutes
to 2 minutes. The calculation of the optimum exposure time was used to understand the
effects of the camera system for varying flux levels. These results will not be used to
determine the optimum exposure time at RMC as PE only simulated the effects of water
ingress. In practical terrns, as the flux values increased, the easier it was to see the pieces
of PE which simulated water ingress. Logically then, water shoutd also be easy to see at
these higher flux levels.
4.5 CF188 Rudder
To understand the results of the neutron radioscopy performed on the CF188
nidder, a basic knowledge of the interna1 structure is required. A print of a neutron
radioscopy image of a serviceable section of a rudder with its components labelled is
shown in Figure 4.22 (Note that more detail is observed in the actual computer monitor
image). In Table 4.5 is listed the components corresponding to the capital letters in
Figure 4.22. Al1 images from this section are shown in a Iarger format in Appendix F.
Figure 4.22 - Exposure 2 of Rudder frorn Penn State
Since the neutron beam at Penn State did not completely cover the entire area of
Table 4.5 - Interna1 Components of CF188 Rudder
the scintillation screen, d l RMC images were cropped so that the images can be
Letter A B C D E
compared. Only the data inside the circle were used in the histograms as the data
Component Al-um Honeycomb pattern fiom FM300 adhesive Porous Adhesive - Connects honeycomb to main spar Main Spar - Aluminium material 1 0/32 Steel Anchor Nuts (four in total) 0.050 inch Aluminium Sheet - Fairing support
associated with the dark area outside the circle were excluded fiom the histogrms. Al1
images are presented in a format similar to Figure 4.22 and contain similar intemal
components as indicated in Table 4.5.
Figure 4.23 shows a series of images taken at RMC of Exposure 5 taken at a
4-minute exposure. The first image (a) is without any data manipulation and the second
image (b) had the conîrast and brightness manipulated to make the original image visible.
The 1st image (c) has had a 2 x 2 erosion filter applied to the original image and then the
brightness and contrast were manipulated.
(a) no enhancements (b) basic enhancemats (c) advanced enhancements
Figure 4.23- Sequence of Exposure 5 Images taken nt RMC (10 kW)
Although not visible, in Figure 4.23% there are some white spots indicating
gamma contamination or system noise. Appendix F, Figure F.2, clearly shows the
evidence of system noise. Figure 4.23b has very little definition, as some of the
individual components shown in Figure 4.22 are not clearly distinguishable, but water is
indicated, as is the hinge assembly. Water ingress and its location were previously known
liom neutron radiography. In Figure 4.23 b and c, water ingress is visible and the pattern
of water is identical to known images of this area. Figure 4 . 2 3 ~ also indicates clearer
details such as the main spar, porous adhesive and the anchor nuts.
The histogram associated with the original image (Figure 4.23a) of exposure 5
taken at RMC is shown in Figure 4.24. Thresholding was applied to the image to identiS
the features related to the curve and are indicated in Figure 4.24.
O
1024 1536 2048 2560 3072 3584 4096 inœnsltll Value
Figure 4.24 - Histogram of Exposure 5 from RMC (10 kW)
The intensity values of the graph in Figure 4.24 are in the low thousands and
therefore the original image is totally black as previously seen in Figure 4.23a. The
intensity values have a small range, from 1536 to approximately 3500, which is only 3%
of the total number of intensity values available. This small intensity range produced an
image with a small dynarnic range, which reduced image contrast between components as
shown in Figure 4.23b.
Figure 4.25 shows a series of images of Exposure 5 taken at Penn State with the
reactor at 1 0 kW and an exposure time of 2 minutes. The original image (a) is without
any data manipulation and the second image (b) had the contrast and brighmess
manipulated to make the original image visible. Due to the increased flux levels, no
enhancement filters were requred.
(a) no enhancements (b) basic enhancements
Figure 4.25 - Sequence of Exposure 5 images taken nt P ~ M State (10 kW)
Again in the original image, Figure 4.25% no details are visible without some
contrast and brightness alterations appIied to the image. Al1 the major components are
visible in the enhanced image as seen in Figure 4.2% There are visible intensity
differences fiom the main spar, a porous adhesive area and water. The honeycomb
pattem is dso visible in both areas, with and without water.
The applicable histogram of Figure 4.25% with labels indicating the major
components of the rudder, is presented in Figure 4.26.
Penn StibeData8t10 kW-Exp5of Rudder @ -30 C, 512 x 512 d U h , Wi inrsiel-16
Figure 4.26 - EIistogram of Exposure 5 from Penn State (10 kW)
The intensity values of the graph in Figure 4.26 are in the low thousands and
therefore the original image is totally dark as previously seen. The histogram covers 4%
of the available range of intensity values, fiom 1 536 to approxùnately 4 1 20, producing an
image with a larger dynamic range than the image fkom W. This larger range
indicates that the intemal components should produce visible density differences on an
image, which c m be seen in Figure 4.2%
An image of Exposure 5 taken at Penn State with the reactor at 100 kW and an
exposure time of 1 minute is shown in Figure 4.27a and b. At this power setting? the flux
produced was two orders of magnitude greater than available at RMC. Due to the higher
flux values, the only enhancement required was a small change to the contrast and
brightness to produce the image in Figure 4.2%.
(a) no enhancements (b) with basic enhancements
Figure 4.27 - Sequence of Exposure 5 images taken at Penn State (100 kW)
Unlike the previous undtered images, the rudder components are visible and,
adjusting only the brightness and contrast, a good image of Exposure 5 can be seen in
Figure 4.27b. Al1 intemal components show good dennition and dynarnic intensity
range. The honeycomb pattern c m be clearly seen in a larger image in Appendix F,
Figure F.8.
The applicable histogram for Figure 4.27a, with labels indicating the major
components of the rudder, is presented in Figure 4.28.
Figure 4-28 -Eistogram of Exposure 5 from Penn State (100 kW)
This histogram covers 14% of the total available intensity scale, fiom
approximately 3000 to 12000. Since the histogram spans past the intensity value of
10,000, the image should be visible as can be seen in Figure 4.2%. The large dynarnic
range of this histogram should translate to significant differences in intensity values for
given interna1 components of the rudder, which can also be seen in Figure 4.27b.
The image produced at a 100 kW power setting (Figure 4.27) has the best
definition and the largest range of intensities compared to the other two images taken at
RMC and Penn State at 10 kW. Al1 three images fiom RMC (10 kW) and Penn State at
10 and 100 kW show the intemal components of the rudder water ingress, Figure 4.29.
Due to the clarity of the images produced fkom the Penn State reactor at 100 kW, these
images can be used as a cornparison standard.
Figure 4.29 - Best Results of Exposure 5 from (a) RMC at 10 kW, (b) Penn State at 10 kW, and (c) Penn State 100 kW, with image enhancements
Plotting nomalised histograms of Exposure 5 at RMC (1 0 kW) with an exposure
time of 4 minutes and Exposure 5 at Penn State (100 kW) with an exposure time of 1
minute produced results shown in Figure 4.30. Both sets of data were normalised to the
centroid of the honeycomb peak. In Appendix G, the images for each rudder exposure
were correlateci to the applicable histograms, similar to Figure 4.30.
--+ +- Horieycomb
1 .
- RMC -min - Pem State 100 kW EYP 5 1 min
0.6 -
a6 - 1 1 Hqe 1-r ~ d h ~ 3 ~ ~
r\ A
Anchor NuLs a4 8
0.2 *
0 --- O Q 1 0.2 a3 a4 a5 a6 o. 7 as 0.9 1
Nomrlltad hbmrfly Vilue
Figure 4.30 - Normaiised Graph at Two Different Power Settings
Normalising the data allowed a direct comparison between the two graphs and
compensateci for the large magnitudes of pixel count and intensities. The goal in
normalising the data was to confirm that both images produced sirnilar shapes and
features that were related to the intemal components of the nidder. Although the
histogram graphs are not identical, the shapes are similar and features on the graph do
relate to specific items in the images. The diffaence between the two graphs shows that
the histogram fkom the image taken at Penn State uses a larger percentage of the intensity
scale. The curve of the histogram fiom the image taken at RMC is compressed on the lefi
side of the peak and elongated on the nght side of the peak relative to the other
histogram. Overall, the image taken at RMC contains al1 the required data to produce an
image that is capable of indicatuig the honeycomb pattern, main spar, porous adhesive,
anchor nuts, duminium structure and water ingress.
4.6.1 Resolution
The resolution of CX and SR radiography film was detemüned utilising the MTF
technique and then these results were compared to the resolution calculated for the
radioscopy system. The CX film was exposed for 18 minutes and the SR film for 135
minutes. The 10% MTF values were calculated as descnbed in Section 2.8.3. The Ut
values and calculated cyclelmm values, with absolute mors for both types of film, are
shown in Table 4.6.
r I I
SR- 135 1 39.3 1 1 0.1 1 + O.OU-.O1 (1 8Yo/9%) 1 170 (+ 201 - 20)
Table 4.6 - Resolution Variables for Film at RMC
These resolution calcdations indicate that both films have very similar resolution
Exposure t h e 1 PixeYmm CX - 18 minutes 1 39.3 1
values when the error is taken into account, with the SR film having a smaller resolution
variation than the CX film. The resolution of SR film cdcdated fkom half a cycle is
ut (mm) 0.12 + 0.02 (16%)
Cycles/mm at MTF Values of 10 % 160 (+ 28/ - 20)
0.0029 +0.0001/- 0.0003 mm and the resolution for CX film is 0.003 1 +0.0003/- 0.0005
mm.
Comparing the resolutions of radioscopy and radiography against expowe time
produces the results shown in Figure 4.3 1.
Worklng Rem Won 8t RMC of NRS
Figure 431 - Resolution Comparison
The largest resolution area associated with the Weutron Radiography at RMC"
indicates the estimated range available for film. The maximum value of the range was
calculated by using the full cycle resolution value for CX film, and the minhum value of
the range was calculated by using the smallest half cycle resolution value £kom the SR
film. The exposure time was calculated for "Neutron Radiography at M C " by
arnalgamating the exposure times for both the CX and SR film, which created a range of
possible exposure thes .
The resolution calculated for both films used the maximum and minimum half
cycle values and the exposure times are actual times used at RMC for a given type of
film. The exposure tirne for the 'heutron radioscopy at RMC" indicates a range of
exposure times (fiom 30 seconds to 10 minutes) rather than a single optimum exposure
time. Graphing the optimum exposure time would make the area associated with film
look even larger and therefore, a range of possible exposure times was shown. Figure
4.3 1 shows the general inverse relationship between resolution and exposure tirne. It also
shows that radioscopy has fast exposures at the cost of poorer resolution, while film has
very fine resolution at the cost of long exposure times.
The results fkom the modifiai ASTM E545-75 Type A were not significant as the
SR film showed al1 12 holes while the CX film showed 11 holes. The holes were too
large in the standard to be useful for the SR film. For film, the SI for the ASTM E545-91
should be used.
4.6.2 Time and Cost Study
For a cost cornparison between neutron radiography and radioscopy at RMC,
capital cos& will not include items that are now considered part of the Facility such as the
reactor, the beam stops, positionhg system and permanently installed radiation meters.
These items are not included, as both techniques require thern for inspection of flight
control surfaces using a neutron beam. The capital costs for both techniques are shown in
Table 4.7.
Table 4.7 - Cornparison of Capital Cost of Radioscopy and Radiography
- -- - -
The estimate of total cost for both techniques show that radioscopy is
approximately 23% greater for the initial capital purchase. Capital costs are usually a
single cost, while consumable costs are small but ongoing.
For consumables and labour, the mst per image is indicated in Table 4.8. The
operating costs of the reactor (Le., fuel usage, operator salary, etc) was not included in
this analysis as the reactor is used simultaneously for the application of NAA.
Radiography 1 Cost
Cassettes & Screens(2) Film Processor Vacuum Pumps (2) BP1,SI Densitometer Dark Room Oight, timers, door etc) Total
lbdioscop y 1 Cost (Cdn dollars)
2 x 5930 16 000 2 x 300 4000 2000 2000
$36 460
CCD Camera System Image software Scintillation Screen Bellows Cornputer Volt meter
Total
(Cd. dollars) 32 920 4000 5000 710 2000 300
$44 930
Table 4.8 - Cost per Image Radiography Cost per image
CX Film (50 shcas/box )
AA Film (50 shects/bo~)
Developer Fixer Density Strip Misc. (~nvclopes, ~IOVCS,
Costhnage (CX)/(AA)
Cost per image (Cdn dollars)
As indicated in Table 4.8, radioscopy is about 6 to 10 t h e s less in cost per image
Radioscopy
($240/50) = 4.80 ($583/50) = 1 1 -66 ($66/100') = 0.66 ($36/1001) = 0.36 ($1 56/500L) = 0.3 1 Estimate = 0.50
$6.63/ 13.49
Labour costhnage (CWAA) Total codimage (CX/AA)
due its simpler image storage method and shorter exposure time. The final cost analysis
Image storage (np disk)
Printing image ciara)
Costhage
is related to a time study that was based on an actual practice and as close as possible
Note: 1 - Estirnated number of images 2 - Labour wst at %38/hour for a 18 minute exposure (CX) or 25 minute acposure (AA) 3 - Labour cost at S38hour for a 4 minute expomre
$1 1.40/ 15.83 $18.03129.32
replicates the real activities associate- with that component. Figure 4.32 incorporates
two flow charts, showing the steps required in taking a single image of a rudder using
Labour cost/imap;e Total costhmage
radiography and radioscopy, and the time required for each step.
$2.53 $3.13
Figure 432 - T h e Study Flow Chart for Rudder
It takes eight images and two position changes to radiograph a CF188 rudder.
Using the flow chart for radiography, a total of approximately 204 minutes (using an 18-
minute exposure time) is required to radiograph a rudder. It takes approximately 76
minutes (using a 4-minute exposure tirne) to image a rudder using radioscopy. Using
radioscopy, there was a 2/3 reduction in person hours compared to radiography and a
78% reduction in reactor time ([(18 minutes * 8) - (4 minutes*8)]/ (18 minutes * 8) * LOO
= 78%). The method used to determine the total time and cost to radiograph a rudder was
extrapolated for the remainuig CF188 flight control surfaces. The tirne and cost
appreciation for each component is shown in Table 4.9, based on usage of CX film.
Table 4.9 - Operational Cost and Inspection T h e of Radiography and Radioscopy 1
1 Flight Control Surface 1 Operational Inspection Time 1 Operational cost ( labour + 1
Although neutron radioscopy is iaitially more expensive than neutron radiography
(23%), the cost per image is less expensive (83%). A tirne study has indicated that when
radioscopy is used to inspect a rudder, the total inspection time is reduced by
approximately 60%, translating into a total cost savhgs of 70% relative to neutron
radiography. These savings can be extrapolated to a complete inspection of al1 twelve of
the flight control surfaces, as indicated at the bottom of Table 4.9. Considering the
higher capital cost of radioscopy (Table 4.7) with its lower operating cost (Table 4.9) the
break even point is at about three sets of aircrafk components. Moreover, neutron
radioscopy requires about 78% less reactor time to image a nidder than neutron
radiography, fkîher reducing the cost of this technique. It is the reduction in labour and
reactor time that is significant, making radioscopy economically favourable.
-
Inboard Leading Edge Flap* Outboard L&g Edge Flap* Trailinp; Edge Flap* Ailerons* Horizontal Stab* Rudder
Total for half set Total set (complet e aircraf t)
Note: * - extrapolated inspection and cost estimates, # - based on CX film usage
(mulutes) ~adiographf
33 1 129 454 179 708 204 2005 4010
consumables) Radioscopy
123 49 166 67 206 76 74 1 1482
~adiogra~hf $295.85 $1 14.86 $406.9 1 $159.79 $634.1 O $182.26 $1793.76 $3587.52
Radioscopy $85.70 $34.03 $1 15.79 $46.63 $181.47 $52.93 $5 16.70 $1033.40
CHAPTER 5
CONCLUSION
5.1 Introduction
The Neutron Radiography System (NRS) installed on the SLO WPOKE-2 Facility
at RMC has the ability to inspect CF1 88 aircrafi flight controI surfàces for water ingress.
The low flux value of 2 to 3 x 104 n/cm2 s at the image plane necessitates using a fast film
in order to have a reasonable exposure t h e to produce a image with a sufficient
background density. A fast film (CX) has been adapted to reduce exposure times to 18 to
25 minutes fiom the standard film (SR) which takes more than 2 hours. A CCD camera
and scintillation screen combination could replace the film and produce images with lower
exposure times without degrading the ability to detect water ingress into the aircraft flight
control surface. The goal of this thesis was to install. characterise and develop a near real-
tirne radioscopy system on the NRS at RMC in order to expedite the inspection of
composite aircraft flight control surfaces.
5.2 Installation of Hardware
The first phase of the study was the installation of the hardware, which included
the CCD camera, mirror, lem, scintillation screen, iight-tight enclosure and image
processing software. This phase was a relatively straight fonvard endeavour, as the
technology tends to be an assembly of off-the-shelf cornponents and requires the user to
become familiar with the operation of the software and hardware.
Characterisat ion of the Svstem
The second phase was the characterisation of the CCD camera system to detemine
t s capabilities and performance. Images taken to establish the system's capabilities did
not have a component or object between the neutron bearn and the scintillation screen so
that only the camera and scintillation screen were analysed. Parameters that affected the
performance of the CCD camera were the relative darkness of the difiùser, the camera
temperature, the quantity of neutron flux, the exposure time of an image and the
resolution.
When "dark images" were taken, the darkest images were produced by a capped
camera lem (i.e., no diffuser) for a given exposure tirne, followed by the portable difiser
used at P ~ M State and the h e d diffuser used at RMC. Producing an image with an
intensity value of zero is not obtainable due to system noise and dark current. The lowest
the carnera system could achieve was a peak intensity value of 1140, which was taken
without a diffuser and at a 30 second exposure tirne.
The camera temperature had a large effect on the system's performance. For a dark
image, as the temperature of the camera decreased, the image became darker and became
more centralised around the peak. At -1 5 OC, the peak intensity value was recorded as
1397 and, when the temperature was lowered to -3 0 OC, the peak intensity value dropped
to 1205, for a 2-minute exposure t h e . This drop in peak intensity values is directly
proportional to a reduction in dark current.
The neutron flux was also varied by using two different reactors and power levels
and, as the neutron flux increased, the overail intensity of the unified image also increased,
producing a larger dynamic intensity range. The unified images taken at RMC at 10 kW
produced peak intensity values korn approximately 1400 to 4000 depending on the
exposure time. The dynamic range produced fiom the intensity values fiom the dark
image and unified image varied fiom approxirnately 200 to 2000. A saturated image was
acquired at Penn State at 100 kW (3.0 x 106 n/cm2 s at the image plane) nom a 5-minute
exposure.
In general as the exposure time for an image increased for either situation (dark
image or unified image), the intensity value increased and became less centralised around
its peak. Long exposure times produced images with a large range of intensity values with
additional system noise; conversely, very short exposure t h e s produced Mages that
lacked detail due to low contrast conditions.
To measure the resolution, the MTF method was used at M C , and a line pair
gauge was used at Penn State. To measure the system's sensitivity, a modified ASTM
E545-75 Type A indicator was used. For exposure times ranging from 30 seconds to 10
minutes at RMC, a measured resolution of 0.47 (+0.34/-0.14) cycles/mm at 10% MTF was
calculated. The line pair gauge recorded a resolution of 0.656 Ip/mm for both power
settings and al1 exposure times at Penn State. At RMC, the measurable spatial resolution
for the CCD system was 1 .O6 mm, which is half of the d c y c l e measurement.
In practical terms, this series of analyses showed that the dark image should be as
black as possible; therefore, every effort should be made to eliminate light leaks, reduce
the camera temperature, and reduce the absorption and reflective surfaces in the diffuser.
The analysis has also indicated that the unified image should be as bright as possible;
therefore, the greater the neutron flux, the brighter the screen, recognising that CCD pixel
well saturation is a h i t i ng factor.
Ideally, the further the intensity values of the dark image are nom the unified
image, the greater the dynamic range of the systea A large dynamic range translates into
a larger intensity separation between dfierent materials with different neutron absorption
and scattering properties.
A method of revealing the intensity values of observed indications in the
radioscopy image was devised. A histogram was made of the image such that the
dynamic range is indicated between the dark and unified image. Within this range,
indications such as moisture and components of the object cm be displayed.
5.4 Develo~rnent of Neutron Radiosco~v Svstem
Once the NRS was characterised, it was then tested by irnaging test pieces and then
a flight control surface. A honeycomb test piece with water inserted into it was used to
detennine the optimum exposure t h e , camera gain setting and digital enhancement
technique. Using the best possible camera settings determined fkom the test piece, a
CF188 rudder was then imaged at three different flux values. The images taken at the
higher flux values at Penn State were compared to the images taken at RMC to detennine
the quality of these images.
Images taken of the nidder were transformed into histograms and normalised. This
normalisation around the peak centroid allowed the same image of the rudder to be
compared at different flux levels. Thresholding was applied to the images, which allowed
the different characteristics on the histogram to be identified. It was discovered that water
has a characteristic location on the histogram for all flux levels.
The optimum exposure time for radioscopy at RMC was detennined by
comparing the intensity diffaence (cell intensity value subtracted fiom the background
intensity value) for varying exposure tirnes. This cornparison was completed for a set of
single cells and a set of three cells, with each cell or cells having varying volumes of
water. For the water indication, a 4 to 6 minute exposure time for both the single ce11 and
the three celk produced the best image, having good contrast between the cell and
background.
Using intensity dflerence to determine the effects of camera gain, a 3-minute
exposure t h e showed that no measurable benefit was produced, therefore, it was
concluded that the gain should left at the zero setting.
A large number of enhancement techniques were applied to images of the
honeycomb test piece taken at a four-minute exposure time with no camera gain. Of al1
the enhancement techniques applied, about t en techniques produced reasonable images.
Of the ten, three techniques produced very good images, these being equalisation of the
histogram, a J flat filter and an erode fiher. Of the three, the erode filter produced an
image that reduced the noise considerably and indicated the various volumes of water for
single cells and three cells.
The images of the CF188 rudder taken at RMC were compared to images taken at
the Penn State facility that had flux levels at the image plane two orders of magnitude
larger. Images produced at the Penn State facility with the same camera system showed
larger intensity dflerences between intemal components and better quality images overall,
at a quarter of the exposure time compared to images taken at RMC. Enhanced images of
the CF188 rudder taken at RMC have very little system noise, show all internal
coqonents, indicate some contrast differences between internal components, and clearly
indicate water ingress.
Analysing the CCD camera based neutron radioscopy system installed at the
SLO WPOKE-2 Facility at the RMC has shown that it is:
1. a viable technology that is capable of producing quality neutron images at a
very low neutron flux level;
2. a capable technology that can produce images at a shorter exposure time
relative to film;
3. a sensitive technology, capable of detecting small amounts of water;
4. a cost effective technology relative to film based systems; and
5. a viable technology for inspecting CF1 88 flight control surfaces.
Therefore, the SLOWPOKE-2 Facility at RMC has a neutron radioscopy system
that has met and exceeded al1 performance objectives and is now ready to supplement the
film based inspections carried out on CF188 flight control surfaces.
CHAPTER 6
RECOMMENDATIONS
Recommendat ions
The following are the recomrnendations for the CCD camera-based neutron radioscop y
systern installed at the SLOWPOKE-2 Facility at the M C , based on the work completed
for this thesis.
1. The radioscopy system should be used as an initial screening inspection of the
CF188 flight control surfaces so that water ingress can be located. Then (the more
costly) neutron radiography can be used in these locations that have indicated
problems, so that higher resolution and permanent images of the area can be taken.
Using neutron radioscopy in this fashion eliminates the expensive process of
taking radiographie images of areas that have no indication of water ingress.
2. An investigation into applying pattern recognition or neural network techniques on
the histograms, taken from the neutron images of flight control surfaces, to analyse
automatically the area on the histogram associated with water ingress should be
undert aken.
3. An increase in flux at the image plane would provide a slightly lower exposure
tirne, less requirernent for enhancement and, most importantly, a larger dynamic
range which would reveal the presence of water in the histogram. With the
SLOWPOKW-2 operating at full power (20 kW), the tlux should be about halfof
Penn State at 10 kW. It is recommended that the flux at the image plane be
measured and the shielding required be assessed regarding the feasibility to
perfonn neutron radioscopy at full power (20 kW) at RMC.
REFERENCES
Standard Practices for Thermal Radiomwhv of Materials, American Society for Testing and Materials, ASTM €748-95, pg. 2,1996.
Kobayasi, H., Neurron Radiography, 3, Kluwer, pg. 42 1, 1990.
Lange, KI., Leeflang, H.P., Mark af, J.F.W., Neutron Radiosco~v usinn Electronic Cameras with Zoom F Ca~abiliw and Hi& Detectivity, 6 European Conference on Non Destructive Testing, pg. 8 17-82 1, 1994.
Casali, P., Chiro, R , Partemi, Neutron Imagine Usinp; a Low-Cost Cooled CCD Camera, Nuclear Energy in Central Europe: Present and Perspectives, Portoroz, Slovenia, pg. 566575, June 1993.
Brenizer, J.S.,Berger, H., Stebbings, C.T., Gillies, G.T., Performance Charactenstics of Scintillators for use in an Electronic Neutron Imaginn Svstem for Neutron Radiom~hy, Science Instruments, Vol. 68, No. 9, pg. 337 1-3379, September 1997.
Berger, EL, Polanslq, D., Brenizer, J.. Bueno, C., Electronic Neutron Imaaq, Indusnial Quality Inc Final Report, Report Number N68936-93-C-0274, January 1994.
Czirr, J.B., MacGillivray, G.M, MacGillivray, RR, Seddon, P.J., Performance and Characteristics of a New Scintillator, Nuclear Instnunents and Methods in Physics Research, A 424, pg. 15-19, 1999
Brenizer, J.S., Berger, iX, Gillies, G.T., Mengers, P., Stebbings, C.T., Polansky, D., Rogerson, D.J., Develo~ment of a New Electronic Neutron Imaging System, Nuclear Instruments and Methods in Physics Research, pg. 9- 13, March, 1999.
Lanza, RC., The Use of Neutron Tomoerraphv Techniques for the Detection of Corrosion D a m e in A i r d Structures, Amencan Society of Mechanical Engineers, Aerospace Division. AD-04. Vol. 47, pg. 253-259, 198 1.
10. Balasko M, Svab E, Neutron Radioeraphy in Research and Developinent, Nukieonika 39, No 1-2, Pg. 3-22, 1994.
1 1. Leeflang, KP., Markgaf, J.F.W., Detection of Corrosion on Airmafi Component by Neutron Rad iom~hy , Non-destructive Testing Evaluation, Vol. 1 1, pg. 1 15- 128, 1994.
12. Dupuis, J.P., Huriet, J.R, Industrial Inspection of Aeronautic Samples - First step with the DIANA, 5 World Conference on Neutron Radiography, pg. 657-664, June 1996.
13. Bickerton, M., Deveio~ment of Im~roved Techniaues for the Neutron Radioai-aphv of CF188 Flinht Control Surfaces, Department of Chemistry and Chernical Engineering, Royal Military ColIege of Canada, Kingston Ontario, 1 998.
14. Lanza, RC., Shi, S., McFarland, E.W, A Cooled CCD Based Neutron lmaging Svçtern for Low Fluence Neutron Sources, E E E Transactions on Nuclear Science. Vol. 43, No. 3, June 1996.
1 5. Non-destructive Ins~ection Reuort Canadian Forces CF- 1 8A, United States Air Force Sacramento Air Logistics Cen ter, McClellan Aïr Force Base, California. 1995.
16. DCIEM Report No 98-TM-44, Progress in the Non-destructive Evaiuation of CF-18 Comwsite Flight Controls for Water ïngress and Related Damage, DCIEM Air Vehicle Research Detachment, Ottawa, Ontario. June 1998.
17. Bickerton, M., Develo~ment of Improved Techniaues for the Neutron Radiomphv of CF188 Flinht Control Surfaces, Department of Chemistry and Chernical Engineering, Royal Military CoUege of Canada, Kingston Ontario, 1998.
18. Standard Practices for Thermal Radioerarihv of Materials, American Society for Testing and Materials, ASTM E748-95, pg. 2,1996.
19. Whittemore, W., Neutron Sources and Facilities - Present Status and Future, 6th World Conference on Neutron Radiogaphy, Japan, May 1999.
20. MacGiIlivary, G.M., Neutron Radiorzraphv Princi~les and Practices, Course notes fiom Neutron Radiography course given at M C 15-19 Feb, section 2.5, pg.11. 1999.
2 1. Standard Practices for Thermal Radionra~hv of Materials, Amencan Society for Testing and Materials, ASTM E748-95, pg. 8,1996.
22. Balasko, M., Svab, E., Neutron Radiographv in Reasearch and Develo~ment, Nukeonika 39, No. 1-3, pg. 7, 1994.
23. Standard Practices for Thermal Radiographv of Materials, American Society for Testing and Materials, ASTM E748-95, pg. 9,1996.
24. Standard Practices for Thermal Radio41.aohv of Materials, Amencan Society for Testing and Materials, ASTM E748-95, pg. 7,1996.
25. Standard Guide for Radiosco~y, American Society for Testing and Materials, ASTM E1000-92. pg. 396- 397,1996.
26. Brenizer, J.S., Berger, H., Stebbings, C.T., Gillies, G.T., Performance Characteristics of Scintillators for use in an Electronic Neutron ha-gin& Svstem for Neutron Radiomphy, American Institute of Physics, Rev. Science Instruments, Vol. 68, No. 9, September, pg. 3371-3379, 1997.
27. Berger, K, Polansky, D., Brenizer, J., Bueno, C., Electronic Neutron Imaninq, Industrial Quality Inc Final Report, Report Number N68936-934-0274, pg.4, January 1994.
28. Pleinert, H., Lehmann, E., Komer, S., Design of a New CCD- Camera Neutron Radiofsaphv Detector, Nuclear Instruments and Methods in Physics Research, Elsevier Science Publishers B.V. A 399. pg. 386, 1997.
29. Lanza, R, Shi, S., McFarland, E., A Cooled CCD Based Neutron Imaging System for Low Fluence Neutron Sources, IEEE Transactions on Nuclear Science, Vol. 43, No. 3, pg. 1347, June 1996.
30. Hi& Performance Digital haaine, Optikon Briefing Notes f?om Neutron Radiography Course given at RMC 1 5- 19 Feb 1999.
31. Healey, G., Kondepudy, R, Radiometric CCD Camera Calibration and Noise Estimation, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 16, No. 3, pg. 267-268, March 1994.
32. Buil, C., CCD Astronomv Construction and Use of an Astronomical CCD Camera, William Bell Inc, Virginia, pg.4-8, 199 1.
33. Buil, C., CCD Astronomv Construction and Use of an Astronomicd CCD Camera, Wiliam Bell Inc, Virginia, pg.34, 199 1.
34. Buil, C., CCD Astronomv Construction and Use of an Astronomical CCD Camera, William Beil Inc, Vuginia, pg. 26, 199 1.
35. Buil, C., CCD Astronomv Construction and Use of an Astronomical CCD Camera, W i a m Bell h c , Virginia pg. 30, 199 1.
3 6. Jain, A, Fundamentals of Digital image Processing, Prentice Hall, Englewood C l m , NJ, pg. 5 1, 1989.
37. Buil, C., CCD Astronomv Construction and Use of an Astronomicd CCD Camera, Wfiam BeU Inc, Vxghia, pg. l69,;99 1.
3 8. Jahne, B., Digital Image Processinn. Conceuts, Alnorithms, and Scientifk Au~lications, Springer, New York, Pg. 21 1, 1997.
39. Spiegel, M., Probabilitv and Statistics, McGraw-Hill Book Company, pg. 78, 1975.
40. Avallone, A., Baumeister, T., Mark's Standard Handbook for Mechanical Engineers, McGrall-Hill. New York, pg. 17-21, 1978.
41. Buil, C., CCD Astronomy - Construction and Uses of an Astronomical CCD Camera, Willmann-BeU. Virginia, pg. 37,199 1.
42. MCD IHandbook, Andor Technology Training Notes, Pg. 53, Jan 98
43. Euil, C., CCD Astronomv - Construction and Uses of an Astronomical CCD Camera. Willmann-Bell. Virginia, pg. 43, 199 1
44. MCD Handbook, Andor Technology Training Notes, Pg. 82. January 1998.
45. Borne, J., Processine Digital Images, Department of Mathematics and Cornputer Science, Royal Military College of Canada, Kingston, Ontario, Canada, pg. 1 1, 1996.
46. Jain, A., Fundamentals of Digital Image Processinq, Prentice Hall, Englewood C l f i , NI, 1989.
47. Russ, J., The Image Processine Handbook, CRC Press, A m Arbor, pg. 34, 1992.
48. Jain, A., Fundamentals of Digital I m a ~ e Processinq, Prentice Hall, Englewood CliEs, NJ, pg. 51, 1989.
49. image-Pro Plus Ver 4 for Windows Reference Guide, Media Cybemetics, pg. 1-19, 1998.
50. Image-Pro Plus Ver 4 for Windows Reference Guide, Media Cybemetics, pg. 1-169, 1998.
51. Motomura, Y., Yoshii, K., Sonoda, IL, Application of Momholonicd Filters to Noise Suupression in Neutron Radiogra~hv Image, 5& World Conference on Neutron Radiography, Berlin, June 17-20, 1996.
52. Buil, C., CCD Astronomy, Wilùnann-BeU Inc, Richmond, Virginia pg. 253, 1991
53. Miday, D., Resnick, R, Physics, Parts 1 & 2, John Wiley & Sons Inc, Toronto, pg. 999, 1978.
54. Standard Guide for Comuuter T o m o m n h ~ (CT) Imagine;, American Society for Testing and Materials, ASTM E1441-95, pg. 712, 1997.
55. Standard Tenninologv for Non-destructive Examination, American Society for Testing and Materiais, ASTM El3 1697b, pg. 636, 1997.
56. Panchal, C., Sinha, A., Bhawa, B., Shyan, A., S~ ivasan , M., Eduat ion of ksolvinn Power of Neutron Radioscouv System fiom edge Remonse Function Measurements, Insight, Vol. 39, No. 2, February, pg. 98, 1997.
57. Tobin, K., Brenizer, J., Mait, J., Modulation Transfer Function Technique for Real Time Radioscoriic Svstern Characterization. Ap~lied &tics, Vol. 28, No. 23, Decernber. 1989.
58. Gaskill, D., Linear Systems. Fourier Transforms and Outics, Wiley, New York. 1978.
59. Panchal, C., Sinha, A., Bhawa, B., Shyan, A., Srhivasan, M., Evaluation of Resolwlg Power of Neutron Radioscopv Svstem fiorn edne Remonse Function Measurements, Insight, Vol. 39. No. 2. February, pg. 99, 1997.
60. Russ, C., The Image Processinn Handbook, CRC Press, Ann Arbor, pg. 14, 1992.
61. Jahne, B., Digital Image Processinn, Conceuts, Alnorithms. and Scientific Applications, Springer, New York, Pg. 85, 1997.
62. Jain, A, Fundamentais of Digital Image Processinq, Prentice Hall, Englewood Clifi, NJ, pg. 59, 1989.
63. Jain, A., Fundamentals of Digital h a n e Processing, Prentice Hall, Englewood Ciiffs, NJ, pg. 51, 1989.
64. Standard Guide for Radiosco~y, ASTM E 1000-92 (Reproved 1 W6), pg. 405, 1996.
65. Bickerton, M., Develo~ment of Imuroved Techniques for the Neutron Radiograuhy of CF188 Flight Control Surfaces, Department of Chemistry and Chernical Engineering, Royal Military College of Canada, Kingston Ontario, pg.77, 1998.
66. Lewis, W.J., B e ~ e t t , L.G.I., Kirby, C.T., Enhancement to Neutron Radiolow Facilitv at the SLOWPOKE- 2 Facilitv at RMC, 5 World Conference on Neutron radiography, Berlin, pg. 425, 1996.
67. Brenizer, J.S., Hughes, D.E., Bryan, M.E., Gould, R., Flinchbaugh, T.L, Sears, C.F., Recent Improvements to the Pemsylvania State University's Neutron Radiom~hv Facility, 6th World Conference on Neutron Radiography, Japan, May 1999.
68. Lepine, B., McRae, K., Progress in the Non-Destmctive Evaluation of CF-18 Composite Flight Controls For Water ingress and Related Damage, DCIEM Report No. 9 8 -TM-44, pg. 3, 199 8
69. Halliday, D., Resick, R, Phvsics part 1 and 2, John Wiley and Son, Pg. 867, 1978.
70. Availone, A., Baumeister, T., Mark's Standard Handbook for Mechanical Ennineers, McGrall-Hill, NewYork. Pg. 17-2 1, 1978.
7 1. Standard Method for Determinin2 Image Oualitv in Thermal Neutron Radiorrra~hv Testing, ASTM E 545- 75, American Society for Testing and Matenals, PA, USA, pg. 600, 1975.
Avallone A, Baurneist er T, Mark's Standard Handbook for Mechanical Enaineers, McGraU-Hill, NewYork, 1978.
Balasko M, Svab E, Neutron Radioera~hy in Research and Develooment, Nukeonika 39, NO. 1-2, 1994.
Berger H, Polansky D, Brenizer J, Bueno C, Electronic Neutron Irnaeing, Industrial Quality Inc Final Report, Report Number N68936-93-C-0274, January, 1994.
Bickerton M, Development of Iqroved Techniques for the Neutron Radiogra* of CF 1 88 Flieht Control Surfaces, Deparûnent of Chemistry and Chernical Engineering, Royal Military College of Canada, Kingston Ontario, 1998.
B o u e J, Processine. Digital Images, Department of Mathematics and Cornputer Science, Royal Military College of Canada, b g s t o n , Ontario, Canada, 1996.
Brenizer J.S, Berger H, Gillies G.T, Mengers P, Stebbings C.T, Polansky D, Rogerson D.J, Develo~ment of a New Electronic Neutron Imag;in~ System, University of Virginia, Nuclear Instruments and Methods in Physics Research, March, 1999.
Brenizer J. S, Berger H, Stebbings C.T. Gillies G.T, Performance Characteristics of Scintillators for use in an Electronic Neutron Ima d e . SY stem for Neutron Radiographv, -
Arnerican Institute of Physics, Rev. Science Instruments, Vol. 68, No. 9, September, 1997.
Brenizer J.S, Hughes D.E, Bryan M.E, Gould R, Flinchbaugh T.L,Sears C.F, Recent Im~rovements to the Pennsvlvania State University's Neutron Radioma~hv Facilitv, 6th World Conference on Neutron Radiography, Japan, May 1999.
Buil C, CCD Astronomv - Construction and Uses of an Astronomical CCD Carnera, Wilhnn-Bell. Virginia, 199 1.
Casalî, P. Chiro, R Partemi, Neutron Imagine Usine. a Lo w-Con Cooled CCD Carnera, Nuclear Energy in Central Europe: Present and Perspectives, Portoroz, Slovenia, June 1993.
10. Czirr J.B, MacGillivray G.M, MacGillivray R.R Seddon P.J, Performance and Characteristics of a New Scintillator, Nuclear Instnunents and Methods in Physics Research, A 424, 1999
1 1. DCIEM Report No 98-TM-44, Promess in the Non-destructive Evaluation of CF- 18 Cornosite Flight Controls for Water Ingress and Related Damage, DCIEM Air Vehicle Research Detachment, Ottawa, Ontario, June 1998.
12. Dupuis J.P, Huriet J.R, Industrial Iwection of Aeronautic Sarnples - First ster, with the DIANA, 5 World Confêrence on Neutron Radiography, June 1996.
13. Gaskill D, Linear Svsterns. Fourier Transforrns and O~tics, Wiley, New York, 1978.
14. Haiiday D, Resnick R, Phvsics. Parts 1 & 2, John Wiley & Sons Inc, Toronto, 1978.
15. Healey G, Kondepudy R, Radiometric CCD Carnera Calibration and Noise Estimation, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 16, No. 3, March 1994.
16. H i ~ h Perfomiance Digital Imaging, Optikon Briefing Notes from Neutron Radiography Course given at M C 15- 19 Feb 1 999.
17. Image-Pro Plus Ver 4 for Windows Reference Guide, Media Cybemetics, 1998.
18. Jahne B, Digital Image Processing, Concepts. Al~orithms. and Scientïfic A~~lications, Springer, New York, 1997.
19. Jain A, Fundamentals of Digital Image Processing, Prentice Hall Englewood Cliffs, NJ, 1989.
20. Kobayasi H, Neutron Radioaa~hv, 3, Kluwer, 1990.
2 1. Lange H.1, Leeflang H. P, Markgraf J.F, Neutron Radiosco~v using Electronic Cameras with Zoom Ca~ability and High Detectivity, 6th European Conference on Non Destructive Testing, 1994.
22. Lanza R, Shi S, McFarland E, A Cooled CCD Based Neutron Imaging System for Low Fluence Neutron Sources, IEEE Transactions on Nuclear Science, Vol. 43, No. 3, June 1996.
23. Leeflang H.P, Markgaf J.F. W, Detection of Corrosion on Aircrafl Component by Neutron Radioera~h~, Non-destructive Test ing Evaluation, Vol. 1 1, 1 994.
24. Lepine B, McRae K, Proaess in the Non-Destructive Evaluation of CF- 1 8 Com~osite Flight COntrols For Water ingress and Related Damaee, DCIEM Report No. 98 -TM-44, 1998.
25. Lewis W.J, Bennett L.G.1, Kirby C.T, Enhancement to Neutron Radiolow Facility at the SLO WPOKE-2 Facilitv at RMC, 5 World Conference on Neutron radiography , Berlin, 1996.
26. MacGillivary G.M, Neutron Radioeraphv Princip les and Practices, Course notes kom Neutron Radiography course given at RMC 15- 19 Feb, section 2.5, 1 999.
27. MCD Handbook, Andor Technology Training Notes, January 1998.
28. Motomura Y, Yoshii K, Sonoda K, Application of Momholotrical Filters to Noise Sumression in Neutron Radiowaphv Image, 5th World Conference on Neutron Radiography, Berlin, June 1 7-20, 1 996.
29. Non-destructive inspection - Report Canadian Forces CF- 1 8A, United States Air Force Sacramento Air Logistics Center, McClellan Air Force Base, California, 1995.
30. Panchal C, Sinha A, Bhawa B, Shyan A, Srhivasan M. Evaluation of Resolvine Power of Neutron Radioscoov Svstem fkom E d ~ e Remonse Function Measurements, Insight, Vol. 39, No. 2, Febmary, 1997.
3 1. Pleinert H, Lehmann E, Korner S, Design of a New CCD- Camera Neutron Radioaa~hv Detector, Nuclear Instniments and Methods in Physics Research, Elsevier Science Publishers B.V, A 399, 1997.
32. Russ J, The Image Processine; Handbook, CRC Press, Ann Arbor, 1992.
33. Spiegel M, Probabilitv and Statistics, McGraw-Hill Book Company, 1975.
34. Standard Guide for Radioscooy, Amencan Society for Testing and Materials, ASTM E 1000-92, 1996.
35. Standard Guide for Comouted Tomowa~hv (CT) Imaeing, American Society for Testing and Materials, ASTM 1441-95, 1997.
36. Standard Method for Detennininrr Image Oualitv in Thermal Neutron Radiomaphy Testing, ASTM E 745-75, American Society for Testing and Materials, PA, USA, 1975.
3 7. Standard Terrninolom - - for Non-destructive Examinations, American Society for Test ing and Materials, ASTM E1316-97b, 1997.
38. Tobin K, Brenizer J, Mait J, ModuIation Transfer Function Techniaue for Real Time Radiosco~ic Svstem Characterization, Applied Optics, Vol. 28, No. 23, December, 1989.
39. Whittemore W, Neutron Sources and Facilities - Present Status and Future, 6th World Conference on Neutron Radiography, Japan, May 1999.
APPENDICES
APPENDIX A
CHARACTERISTICS OF THE SLOWPOKE AND BEAZEALE REACTOR
A. 1 SLOWPOKE-2 Facility
The SLOWPOKE-2 Facility (Safe Low Power c(K)ritical Experirnent) is a small
light - water, pool reactor designed by the Atomic Energy of Canada Limited (AECL) and
is capable of producing a maximum thermal power of 20 kW. Onginally it was designed
for Neutron Activation Analysis (NAA), but has been modified extensively to facilit ate
neutron radiography and neutron radioscopy. The major sections of the SLO WPOKE-2
Facility include the reactor and the Neutron Radiology System (NRS) (Figure A. 1). The
NRS coosists of a Neutron Bearn Tube (NBT) and a Neutron Carnera Assembly (NCA).
The terrn NRS will be used, but only in reference to the image plane (scintillation screen),
bellows, diffuser and carnera assembly.
The reactor core is located at the bottom of a seaied cylindrical aluminium
container suspended in a stainless steel-lined reactor pool that is 5.87 meters deep and
2.46 meters in diameter. The reactor core, located inside the critical assembly, is
manufactured fiom a zirconium alloy cage containing 198 fuel elements and is 4.4 m
below the surface of the water.' Both the pool and reactor container are filled with
' Bickerton, M., Develo~ment of lmproved Technimes for the Neutron Radioaa~hv of CF188
Flinht Control Surfaces, Deparunent of Chemistry and Chernical Engineering, Royal Military College of
Canada, Kingston Ontario, pg. 12, 1998.
deionized light water that acts as a codant, radiation shield and, in the case of the critical
assernbly, as a moderator. Each element, formed &om Zircaloy-4 tubing, is packed with
pellets of sintered uranium dioxide enriched to 19.89% in Uranium 235. The core is
located within a beryllium annulus and berylliurn plate reflectors located above and below
the annulus. The berylliurn acts as a moderator and neutron reflector as well as the
housing of five inner radiation sites. A small thermal column of heavy water was installed
in place of an outer irradiation site, located between the beryIlium annulus and the reactor
container wall adjacent to the bottom end of the NBT (Figure A. 1 ). At half reactor power,
the flux measured at the location adjacent to the column of heavy water is 1.4 x 10" n/cm2
s vs 5.1 x 10'' n/cm2 s measured at the reactor container wall at a different radial location.'
Lewis, W.J., Andrews, W.S., Bennett, L.G.I., Beeley P.A., Measurernent in Sumort of a Neutron
Radioma~hv Facilitv for the SLOWPOKE- 2 at RMC, Nuclear lnstnunents and Methods in Physics
Research, Elsevier Science Publishers B.V. A299, pg. 430-433, 1990.
Figure A.1 - SLOWPOKE-2 Facilit. at RMC
The NBT is neutrally buoyant in the reactor pool and has a divergent beam
with a measured L/D of approximately 100 starting fiom an aperture of 5 cm in diameter
and diverging to a beam of 60 cm in diameter at the lower beam stop. ' It is positioned
tangentially to the reactor container and, when the bottom end is swung against the reactor
container with a hand pumped hydraulic actuator, it is tilted 8.5' from the vertical. This
action places the iLluminator (a graphite block) adjacent to the centre line of the hiel cage
and almost in contact with the exterior of the reactor container. Neutrons passing through
the annula berylliurn refiector and the heavy water thermal co1um.n (which are both
located inside the reactor container) then pass into the illuminator which redirects the
thermal neutrons through the aperture and up the NBT. The action of swinging the NBT
away fiom the reactor container to the vertical position provides a shuttering fùnction.
The NCA consists of semi-conformal shielding, a positioning system, an image
plane, a belbws and three beam stops. The semi-conforma1 shielding is constnicted fiom
two layers of borated polyethylene and is attached to an actuated frame, which in tuni is
attached to the middle beam stop. The shielding is lowered and confoms to a part under
inspection even if it protrudes fiom the perimeter of the middle beam stop. The
positioning system is an aluminium structure Iocated between the bottom and middle
beam stop. A cornputer program is used to control an X-fiame and Y-fiame assembly, so
that the part being radiographed can be incrementally moved in either the x or y direction.
At the image plane, an aluminium fiame holds either a vacuum cassette with a gadolinium
3 Lewis, W.J., Bennett, L.G.I., Kirby, C.T., Enhancernent to Neutron Radiolo~v Facilitv at the
SLOWPOKE-2 Facility at RMC, 5 World Conference on Neutron radiography, Berlin, pg. 425, 1996.
converter screen and film for neutron radiography, or a scintillation screen, for neutron
radioscopy. The image plane resides above the positionhg system and has the capability
of moving vertically (z-direction) to accommodate part movement and varying part
thickness. The neutron flux at the image plane has been measured using the thin gold foi1
method and was determined to be 2 to 3 x 104 dcm' S.^ The bellows is only installed for
neutron radioscopy and produces a light tight area between the scintillation screen and
middle beam stop. The lower beam stop is positioned above the NBT and has an angled
square hole 44 cm x 44 cm to match the beam output of the NBT. The middle beam stop
has a similar 44 cm x 44 cm square hole opposite to the lower beam stop that tapers to a
12.7 cm x 12.7 cm square hole at the top of the rniddle beam stop. The tapered hole in the
middle beam is a diffuser used to accommodate the CCD camera. The upper beam stop
fits on top of the middle beam stop, covering the small square hole and housing the CCD
A.2 Breazeale Reactor at Pennsylvania State University
The Breazeale Reactor at the Penn State University Radiation Science and
Engineering Centre (RESC) is a medium sized, light water, pool reactor and is capable of
Bickerton. M, Development of Improved Techniques for the Neutron Radionrauhv of CF188
Flierht Control Surfaces, Departmen t of C hemistry and Chernical Engineering, Royal Military CoUege of
Canada, Kingston Ontario, pg.77, 1998.
Bennett, L.G.L. Lewis, W.J., MacGiUivray, G.M., Francescone, O., Enhancement to Neutron
Radiolom Svstem on the SLOWPOKE-2 Facilitv at RMC, 6th World Conference on Neutron radiography,
Japan, 1999
producing a maximum power of 1 MW. The major sections of the Penn State Breazeale
Nuclear Reactor (BNR) include the reactor, beam hibe and shielding as indicated in Figure
A.2.
Figure A2 - Top View of BNR
Uriginally, the reactor had a plate-type fuel core, but in the mid-sixties the reactor
core was converted to a TRIGA design. The reactor can produce a steady state operation
fiom 100 W to 1000 kW, with a capability to pulse the power up to 2000 MW for
approximately 10 miUi~econds.~ The reactor core connects to a bridge that runs above the
reactor pool that allows the reactor core to move in lateral, longitudinal and rotational
directions. The three-axis movement permits the reactor core to be placed tangentially
against a heavy water tank that connects to the NBT, reducing the gamma content of the
neutron beam. The reactor fiel is composed of zirconium hydride mixed with uranium-
235. The heavy water tank is a right cylinder radially aligned with the #4 beam port and is
6 1 cm in diameter and 30.5 cm thick. The fiel rods are placed in a fiame assembly and
are in direct contact with the reactor pool water, which acts as a coolant, radiation shield
and moderator.
There are seven beam ports. but only #7 and #4 beam ports are in operation. The
#7 beam port is used for neutron gauging, while the #4 beam port has a beam tube
installed in it and is used for neutron radiography. The beam tube is a horizontal tapered
collimator that is 297 cm in length and has a diameter of 12.7 cm at the heavy water tank
and expands to 19.05 cm at exit of the beam tube. A gadolinium aperture and bismuth
filter is set at the entrance to the bearn tube adjacent to the heavy water tank. The beam
tube has an LID ratio of 34 and, at the end of the beam tube a flux of 3.0 x 106 dcm2 s,
was measured with the gold foi1 method. '
ïhe Penn State Breazcale Reactor: A facilitv of the Past. Present. and Future of the Pennsvivania
State University, Penn State Department of Publications, pg. 12.
7 Brenizer. J.S., Hughes, D.E., Bryan, M.E., Gould, R. Flinchbaugh, T.L., Sears, C.F., Recent
Improvements to the Pemsylvania State Universitv's Neutron Radiomhy Facility, 6th World Conference
on Neutron radiography, Japan, 1999
The shielding is a fully enclosed room rneasuring 3.8 m by 2.4 m and constxucted
fiom concrete-filled steel boxes sumounding the exit of the beam tube. This back wall of
the structure acts a beam stop and houses the camera, i n t e n s w g screen, support
equipment and the article being radiographed. For this research, a light tight diffiser,
which was constructed so as to simulate dimensionally the middle beam stop fkom the
NRS at the SLOWPOKE-2 Facility, held a CCD carnera and scintillation screen at the
proper positions inside this shielded r w a When the diffuser was properly aligned, the
neutron beam illuminated a 35 cm diameter circle on the scintillation screen.
APPENDIX B
IMAGE SPECIFICATIONS
B. 1 Installation and Optimisation of System at RMC Reduction of Light Leaks
-Hoc, 2 sets of 30 second to 4 minutes exposure, 5 1 1 x 5 1 1 resolution. Statistical Results
-30°C, 30 images at 4 minute exposure, 5 1 1 x 5 1 1 resolution.
B .2 Camera Paramet ers Dark Image - Temperature Change
RMC, -lS°C, 30 second to 4 minute exposures, 5 12 x 5 12 resolution. RMC, -30°C, 30 second to 10 minute exposures, 5 12 x 5 12 resolution.
Dark Image - Diffuser Changes No Diffuser, -30°C, 30 second to 10 minute exposures, 512 x 5 12
resolution. D f i s e r at RMC, -30°C, 30 second to 10 minute exposures, 51 1 x 51 1
resolution. Diffuser used at Penn State -30°C, 30 second to 10 minute exposures, 5 12 x
5 12 resolution.
Unified Images - Neutron Flux Effects RMC (10 kW), -30°C, 30 second to 10 minute exposures, 5 12 x 512
resolution. Penn State (1 kW) -30°C, 30 second to 5 minute exposures, 5 12 x 5 12
resolution. Penn State (10 kW) -30°C, 30 second to 10 minute exposures, 512 x 512
resolution. Penn State (100 kW), -30°C, 30 second to 5 minute exposures, 512 x 512
resolut ion.
Dynamic Range Combination of data fiom dark images and unified images taken at RMC
(10 kW), Penn State (1 kW, 10 kW, 100 kW).
B.3 Resolut ion System Resolution Calculations
RMC (10 kW), -30°C, 30 second to 10 minute exposures, 5 12 x 5 12 resolut ion.
Penn State (10 kW),-30°C, 30 second to 10 minute exposures, 5 12 x 5 12 resolut ion.
Penn State (100 kW), -30°C, 30 second to 5 minute exposures, 512 x 512 resolution.
S ystem sensitivity RMC (10 kW), -30°C, 30 second to 10 minute exposures, 512 x 512
resolution. Pem State (10 kW),-30°C, 30 second to 10 minute exposures, 512 x 512
resolut ion. Penn State (100 kW), -30°C, 30 second to 5 minute exposures, 512 x 512
resolut ion.
B.4 Water Ingress Simulation
Wat er Ingress Water Ingress - Single Cells
RMC (10 kW), -30°C, 30 second to 10 minute exposures, 512 x 512 resolut ion.
Water Ingress - Three CeUs RMC (1 0 kW), -30°C, 30 second to 10 minute exposures, 5 12 x 5 12
resolution.
Water Ingress - Camera Gain Camera Gain - Single Cells
RMC (10 kW), -30°C, 3 minute exposures, 5 12 x 5 12 resolution, at gain settings 0,4, 8, 12, 16.
Camera Gain - Three Cells RMC (10 kW), -30°C, 3 minute exposures, 512 x 512 resolution, at
gain settings 0,4, 8, 12, 16.
Water Ingress - Digital Enhancement RMC (1 0 kW), -30°C, 4 rninute exposures, 5 12 x 5 12 resolution
with digital enhancements applied.
S imulated Water Simulated Water - Single Ce11
RMC (10 kW), -30°C, 30 second to 10 minute exposures, 5 12 x 5 12 resolution.
Penn State (10 kW), -30°C, 30 second to 10 minute exposures, 5 12 x 5 12 resolution.
Penn State (100 kW), -30°C, 30 second to 5 minute exposures, 5 12 x 5 12 resolution.
Simulated Water - Thfee Cells RMC (10 kW), -30°C, 30 second to 10 minute exposures, 5 12 x 5 12
resolution. Penn State (10 kW), -3OoC, 30 second to 10 minute exposures, 512
x 5 12 resolution. Penn State (100 kW), -30°C, 30 second to 5 minute exposures, 512
x 5 12 resolution.
B.5 CF188 Rudder RMC rudder exposure number 1 to 7 less S), -30°C, 2,4,6 minute exposures, 512
x 5 12 resolution, rudder exposure number 5, -30°C, 30 second to 10 minute exposures, 5 12 x 5 12 resolution.
Penn State (10 kW for rudder exposure number 2 to 6, less 5), -30°C, 2, 4, 6 minute exposures, 512 x 512 resolution, expoçure number 5, -30°C, 30 second to 1 0 minute exposures, 5 1 2 x 5 12 resolution.
Penn State (100 kW for rudder exposure number 2 to 6), -30% 1, 2, 3 miaute exposures, 5 1 2 x 5 12 resolution
B .6 Comarison of Radiosco~y to Radio~aphy Resolution
RMC (10 kW) Two film shots (CX at 18 minutes and SR at 2 hours and 15 minutes)
APPENDE C
SYSTEM OPTIMISATION AND CAMERA PARAMETERS
C. 1 Installation and Optimisation of Systern
Figure C. 1 is a histogram before the light leaks were reduced at RMC while Figure C.2 is a histogram after the light leaks were reduced.
Figure C.1- Before Light Leaks Reduced
O 256 512 ïôû lû24 1280 1536 1792 aY8 PO, 258) aBi6 3072 3328 35ô4 3840 4096
hdsrulty Value
Figure C.2 - After Light Leaks Reduced
C.2 Camera Parameters
C.2.1 Dark Image - Temperature Change
The following graphs are of dark images at two different camera temperatures (-15°C -30°C). Resolution, intensity interval and diffher type were held constant, while the exposure time was varied.
Figure C 3 - Dark Image at -15OC
- - - -
Figure C.4 - Dark Images at -30C0
C.2.2 Dark Image - Diffuser Changes
The following are graphs of dark images with camera temperature, resolution, and intensity interval held constant, while varying the exposure tirne and the diffuser type.
Figure CS - Dark Images taken with No Dinuser
l I '.
-10 min I
Figure C.6 - Dark Images taken with diffuser used at Penn State
Figure C.7 - Dark Images taken with Diffuser used at RMC
C.2.3 Unifies Image - Neutron Flux Effects
Several images were taken at four different power settings; at RMC, the images were taken at hdf power (10 kW) and at Penn State the images were taken at 1, !O and 100 kW.
- - - - --
Figure C.8 - Unified Images taken at RMC at 10 k W
Figure C.9 - Unified Images taken nt Penn State at 1 kW
-2 min
-3 mfn
- a O
0 loooo! - -8 min
Figure C.10 -Unified Images taken at Penn State at 10 kW
- -
Figure C.11 - Unified Images Taken at Penn State at 100 k W
C.2.4 Dynamic Range
The following graphs show the dynamic envelope of the carnera system for RMC (10kW) and Penn State (1 kW, 10 kW and 100 kW) by plotting the FWHM value for the dark and unified images for the indicated exposure tirnes. The dynamic range in the text was calculated using peak centroid values, but the following graphs indicate dynarnic range using FWHM values. The dynamic range is similar in magnitude for both methods.
Figure C.12 - Dynamic Range of System from Images taken at RMC
Figure C.13 - Dynamic Range of System from Images taken nt Penn State (1kW)
Figure C.14 - Dynamic Range of System from Images taken at Penn State (10 kW)
Figure C.15 - Dynamic Range of System from Images taken at Penn State (100kW)
APPENDIX D
ERROR CALCULATION OF LNTENSITY DIFFERENCE
D. 1 Error Theorv
Standard error calculàt ions were used throughout the report. The basic de finitions of the two types of error, absolute and relative uncertainty will be given. Then the basic method for adding and multiplying mors will be shown and an error calculation will be developed for intensity difference formula.
The absolute uncertainty of the measurement is an estimate of the range in which the measurement could fa11 due to the limited accuracy of the measurement technique, e.g., 0.52 + 0.02. Where 0.52 is the best estimate and 0.02 is the absolute uncertainty. The relative uncertainty is a measure of the quality of the measurement and is fiequently stated as a percentage uncertaint y emor. Relative uncertaint y or ftact ional uncertaint y is represented as M x or in the exarnple above relative error = 0.02/0.52 x 100 = 3.8% where âx = 0.02 and x = 0.52.
The uncertainty in a sum or difference is the sum of the absolute uncertainties, i.e., if C = A + B then AC = A A + AB. The percentage relative uncertainty in a quotient or a product is the sum of the relative uncertainties, Le., Z = AB then AZ/Z = M A + D I B .
The formula for intensity difference is as follows:
ID (%) =(C-B)/B x 100 Where:
ID = Intensity Difference C = Cell Intensity Value B = Background Intensity Value
D.2 Error Calculation
The following is the formula for the error calculation for intensity difference.
The uncertainty error for Z.
Where Z = (C-B) is
The relative error for ID.
Where ID = (C-B)/B = 233 is
AZ is substituted to obtain the relative error for intensity difference.
D.3 Samde Calculation of Intensitv Difference
A sample calculation of the intensity difference for a single ce11 of water, with a volume of 0.04 mi taken at a m u t e exposure will be carried out. The data for this calculation is presented in Table D. 1.
sure
The absolute value (AC) for the ce11 can be calculated by the following:
MaXLmum value (3 129) - mean value (2336.2) = positive error value = 792.8 Minimum value (22 14) - mean value (23 3 6.2) = negative error value = - 1 22.2
A large intensity value (relative to the background intensity value) is associated with the noise of the system; hence, the reason for the large value in the maximum colurnn in Table D.1. Using the maximum intensity value in an error calculation causes a very large calculated error for the system. One solution to make the error more realistic is not to use the maximum value but to use the minimum error for both the negative and positive value (i.e., 2336 * 122.2).
Another method for calculating the AC value is to use the standard deviation value. Using the standard deviation produces an error value of:
2 x Standard deviation = 2 x 215.6 = AC = 431.2
Using the standard deviation produces a value that is less consewative then using the minimum value, but more conservative than using the maximum value and therefore will be used for calculating the AC value in this Appendix.
The minimum intensity value for the background will be used for the absolute error calculation, as the threshold technique was not used on the background and therefore the maximum value tends to be very large. The calculation for hB is as follows:
Minimum value (2363) - mean value (2880.5 1) = negative error value = -5 17.5 1
Therefore: AE3 = * 517.51
The calculated the relative error for a 5-minute exposure of a 0.04 mL ceii is as follows:
D.4 Error Calculation for Intensity Difference of a Single Cell
The relative error was calculated for a single volume of water (0.14 mL) for exposure times f?om 0.5 to 10 niinutes as described in the above paragraphs. The relative errors for the various exposure times are shown in Table D.2.
Table ater
Averaging the relative errors produced an average relative error of 29%, which was used in Figure 4.1 1.
APPENDIX E
ENHANCED MAGES OF TEST PECE
2 Minute Exposure 2 Minute Exposure wlcontrast Changes
4 Minute Exposure 4 Minute Exposure wlcontrast Changes
Equalization of H istogram
Erode Filter
Equalization of Histogram w l Contrast Changes
Erode Filter wfcontrast
J Flat Filter J Flat Filter wlcontrast
M edian Filter M edian Filter wlcontrast
M ultiplying Image M ultiplying image
wIContrast Changes
Ratio Flat Ratio Flat
wlcontrast Changes
Ratio Optical Ratio O ptical wlcontrast Changes
APPENDIX F
NEUTRON RADIOSCOPY IMAGES OF CF 188 RUDDER
The following images were taken at RMC (10 kW) and Penn State (10, 100 kW) of a CF188 Rudder for an exposure time that is indicated in figure caption for the image.
Figure F.1- Exposure 2,3-minute exposure time taken at Penn State at 100 k W
Figure F.2 - Exposure 5, Cminute exposure time taken at RMC
Figure F.3 - Exposure 5,4-minute exposure time taken at RMC with contrast
Figure F.4 - Exposure 5,4-minute exposure time taken at RMC w/ Erode & contrast
Figure F.5 - Exposure 5, Zminute exposure time taken at Penn State, 10 kW
Figure F.6 - Exposure 5, Zminute exposure time taken at Penn State, 10 kW with contrast
Figure F.7 - Exposure 5,l-minute exposure time taken at Penn State, 100 kW
Figure F.8 - Exposure 5,l-minute exposure time taken at Penn State, 100 kW with contrast
APPENDIX G
COMPARISON O F NORMALISED DATA
Figure G 1 - Exposure 2 of Rudder, taken at Penn State (100 kW), 4-minute Exposure
Nomlised Data of Exp 2 of Rudder O -3OC. 512 x 512 m8drilm-1. W m ~ ~ l y rilarol-13
Normilized htansïty Value
- -
Figure G2 - Normalised Histogram of Exposure 2 at Two Different Power Settings
Figure G3 - Exposure 3 of Rudder, taken at Penn State (100 kW), 4-minute Exposure
- N o m l i D m ot Exp 3 of Rudder
O Ja C. 512 x 512 mscirrrm. rirms~ty nlwiai -16
-- --
Figure 6 4 - Normaliseci Histogram of Exposure 3 at Two Different Power Settings
Figure G5 - Exposure 4 of Rudder, taken at Penn State (100 kW), Cminute Exposure
--
N o m i a l i d Daia of Exp 4 of Rudder O -J) C. 512 x 512 mcUm. Wenscy 1ds91-16
-- -WC EV a n - Pem Çcae 100 kW Exp clmn,
\
O O 1 0 2 O 3 O 4 O 5 O 6 O 7 O a 0 9 1
Nomalüed htensity Value
-- .- ---
Figure 6 6 - Normalised Histogram of Exposure 4 at Two Different Power Settings
Figure G7 - Exposure 5 of Rudder, taken at Penn State (100 kW), 4-minute Exposure
-- Nonnalkgd Data of Ex(, 5 of Rudder
O -30 C. 512 I 5.12 resdim. n m a y inaisi -16
Normal id htensity Value
Figure GS - Norrnalised Histogram of Exposure 5 at Two Different Power Settings
Figure G9 - Exposure 6 of Rudder, taken at Penn State (100 kW), Cminute Exposure
Figure GlO - Normalised Histogram of Exposure 6 at Two Different Power Settings