Professional Documents
Culture Documents
An Automated Radiographic NDT System For Weld Inspection - Part I - Weld Extraction PDF
An Automated Radiographic NDT System For Weld Inspection - Part I - Weld Extraction PDF
An Automated Radiographic NDT System For Weld Inspection - Part I - Weld Extraction PDF
157-162, 1996
Copyright © 1996 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
ELSEVIER PII: S0963-8695(96)00009-6 0963-8695/96 $15.00+ 0.00
An a u t o m a t e d r a d i o g r a p h i c N D T
system for weld inspection:
Part I - W e l d e x t r a c t i o n
T. Warren Liao *t and Jiawei Ni*
*Industrial and Manufacturing Systems Engineering Department, *Electrical and Computer
Engineering Department, Louisiana State University, Baton Rouge, LA 70803, USA
Received 17 July 1995; revised 25 October 1995
Welded structures often have to be tested nondestruc- in situations, such as on site, where portability of the
tively, particularly for critical applications where weld equipment is needed. Gamma-ray sources are much used
failure can be catastrophic, such as in pressure vessels, for the radiography of pipe welds.
load-bearing structural members, and power plants.
The quality of radiographic images is a key factor for
Nondestructive testing techniques for welded joints
detecting weld flaws. The image quality of film radio-
usually include visual, radiographic, magnetic particle,
graphy is usually measured by the flaw sensitivity using
liquid penetrant, and ultrasonic testing methods. Of various image-quality indicators (IQIs). To increase the
interest in this paper i~ the radiographic testing (RT)
ability to detect small and narrow cracks, the RT
technique.
technique must be optimized by a proper selection of
Conventionally, a radiographic image is produced by all influential factors. The important factors in film
permitting the X-ray or gamma-ray source to penetrate radiography include film type, film density, energy level
an object and expose a photographic film. Film radio- of the source, focus-to-film distance, exposure time, and
graphy produces 'negative' images which are normally developer temperature. Discussions related to these
viewed in 'static' mode using a view box by a certified topics can be seen in References 1-6. Article 2 of
inspector. The intensity of exposure of a particular point Subsection (a) of ASTM Section V specifies RT
on the film depends primarily on the object density, the requirements for welds. Weld flaws are typically
source-object-film geometry, the type of film, and the characterized by a local variation in material density,
developing process. The information content of a which manifests itself as a local variation in image
radiograph depends upon the absorption characteristics intensity in the resulting radiographic image.
of the object, which are determined by the thickness, Radiographic inspection is performed to reach a decision
density, and material properties of the object exposed at about the acceptability of the component or product
a selected energy level. X-ray is primarily used in the being tested. Before any evaluation can be made, the
examination of welds in thinner gauge materials where interpreter must be certain that the images are
flaws are likely to be small, fine and difficult to detect. satisfactory. In addition, the interpreter must have a
Gamma-radiography is used for thicker welds, especially solid understanding of the following in order to be
successful with the interpretation of the radiographic
film for welding quality.
tCorresponding author: 3128 CEBA, IMSE Dept., LSU, Baton
Rouge, LA 70803, USA. E-mail: liao@rsip.lsu.edu • principles of radiographic examinations
157
T. W. Liao and J. Ni
• welding processes, their associated flaws, and their (250-pixel by 300-line) was also produced. Each pixel has
images as they appear on the radiograph 12 bits. The digitized images are all stored in VICOM file
• the acceptance criteria as specified in the referenced format. The images contain reference objects to identify
codes and standards. positions in the welds or to calibrate the intensity of the
X-rays. These objects include rulers, penetrometers,
Reference 7 provides a guide to types of discontinuities densitometers, identification letters, space between weld
identified in American Petroleum Institute (API) strips, etc. Only the items within a weld are of interest.
Standard 1104 and their images as they appear in the
radiographic film. Twenty-five probability-of-detection (POD) tapes, pre-
pared originally for the project conducted by Martin
It takes time to train a film interpreter. In addition,
Marietta Electronics, Information & Missiles Group for
human interpretation of weld quality based on film
Martin Marietta Manned Space Systems (M3SS), were
radiography is very subjective, inconsistent, labour
used in this study. Each POD tape was prepared as
intensive, and sometimes biased. It is thus desirable to described above and was designated with a number. The
develop some forms of computer-aided systems to assist existence of weld anomalies was established by experts
the human interpreter in evaluating the quality of welded examining the X-rays using existing tools at M3SS. The
joints. This involves the digitization of film radiography
expert identified the type of anomaly, its approximate
and the development of algorithms to extract welds and
location (measured in inches with a ruler) relative to the
to identify flaws in the welds. Nowadays, radiographic
upper left corner of the image, and the 'size' of the object.
film images can be digitized by a digital radiography
The POD test was designed to produce the data from
system (digitizer) without losing the useful information
which a probability of detection curve as a function of
captured in the original radiograph [8]. Digitized radio-
the object's size could be produced. It is intended to give
graphic images not only enable more-efficient manage-
a first measure of the confidence that objects of a
ment of radiographic inspection data, but also make
particular size have been found. The scope of this paper
automatic inspection of welds [9] and closed-loop process
is restricted to weld extraction. The results of flaw
control (1°] possible.
detection will be reported in another paper.
The Advanced Quality Technology Group of Martin
Marietta Manned Space Systems has been supporting
research projects related to computer-aided or auto- Weld extraction methodology
mated inspection of the Space Shuttle External Tank
Welds [1]-13]. An ongoing project, partially supported Since only the items within a weld are of interest, it is
by Martin Marietta Manned Space Systems, is being preferred to extract the welds (four in total) from each
conducted at Louisiana State University for the devel- image before applying flaw detection algorithms. The
opment of an automated weld inspection system. The 250-pixel by 300-line decimated images were used for this
objective is to provide machine-aided support to the X- purpose to reduce the amount of data needed to be
ray film examination staff, while improving quality and processed, thus shortening the processing time. Figure 1
reducing cost. This paper presents a methodology shows a typical decimated image (POD 15). The extrac-
developed for welds extraction as a part of the tion methodology was developed using the Khoros
automated inspection system. software package[14] installed on a Sun SPARC5 work-
station. Data conversion from VICOM to VIF format
The remainder of the paper is organized as follows. The
was performed on all images to enable their processing in
acquisition of radiographic images is described in the
the Khoros environment.
next section. The weld extraction methodology is then
presented. The test results are then provided and The methodology was developed based on the observa-
discussed, followed by the conclusions. tion that the intensity plot of a weld looks more like
Gaussian than the other objects in the image. This was
manifested by an intensity plot of a line image, as shown
in Figure 2, extracted from Figure 1. The weld extraction
Acquisition of radiographic images methodology consisted of the following steps:
X-ray film strips of about 3.5 inches wide by 17 inches (1) Divide the image into four approximately equal
long were digitized four at a time using the NDT SCAN parts (each part is called a subqmage) vertically.
II digitizer. Refer to Reference 8 for a detailed These four sub-images are called imagek, k = 0, 1,2,
description of the digitizer. Four strips were laid on the and 3. The first three sub-images (from right to left)
digitizer at once with the scan direction orthogonal to the are 62 pixels wide, computed by taking the smallest
long axis of the strips. The strips were digitized at 70 #m integer of 250/4. The width of the fourth sub-image
resolutions (about seven lp/mm) to produce 5000-pixel is 64 ( - - 2 5 0 - 3 × 62). Each sub-image contains
by 6000-line images. This results in large X-ray images exactly one weld.
(30 million pixels) that must be processed to find (2) S e t k = 0 .
anomalies in welds. Along with each 5000-pixel by (3) Process imagek.
6000-line image, a decimated image 20 times smaller (4) Set the no. of iterations I = 0.
158
An automated radiographic N DT system
159
T. W. Liao and J, Ni
160
An automated radiographic NDT system
Peaks D e f e c t e d
Intensity Plot (y=3o) 4-
p
n --
o
0 --
el
>
J
O
m
Oo
>~-
e
J
Oo
, J
0 --
~o ,t.o el
Inten~
F----------
l-ty PLot (y=130)
o
0
I
20 &O
I
A I
61
;° T rough s Defected
O
J
io
I I 7 OO _
20 ~o 81
0
Intensify Plot (y=230) J I
o
--o
>
e
e o
k,
JU 0
Figure 6
,
0
A
20
I
161
T. W. Liao and J. Ni
Acknowledgements
References
1 Yan, Li. 'Optimizing radiographic NDT techniques for welds',
N D T & E Intern 27, 1 (1994) pp 15-20
Figure 7 Welds extracted (marked by black lines) from Figure 1 2 Halmshaw, R. 'Flaw sensitivity in relation to standards for film
radiography', Mater Eval (June 1992) pp 678-683
3 Bowles, W. and Mayberry, G. 'Selecting radiographic film: consid-
extracted. For presentation purposes, each extracted erations for the shop', Mater Eval (March 1988) pp 426-427
4 Wysnewski, R. E. 'Quantitative evaluation of radiographic ima-
weld is marked with two black lines right on its ging variables', Mater Eval (June 1983) pp 826-830
boundaries. The result is shown in Figure 7 (similar to 5 Domanus, J. C. 'Testing of sensitometric properties and image
Figure 3, the negative of the original is shown for the quality of radiographic film and paper by a fast and simple
method', Mater Eval (February 1987) pp 80-85
sake of better visualization). The weld location informa- 6 Becker, G. 'Radiographic image interpretation', Mater Eval
tion obtained from processing the decimated images can (October 1988) pp 1390-1392
be utilized to extract welds from the corresponding 7 Perrone, T. F. 'Principles of radiographic film interpretation of
pipeline welds', Mater Eval (November 1992) pp 1268-1273
fullsize images. The methodology is found to have 100% 8 Graeme, W. A., Jr, Eizember, A C and Douglass, J. 'Digital image
accuracy after being tested with all the 25 POD tapes analysis of nondestructive testing radiographs', Mater Eval
made available for this study. (February 1990) pp 117-120
9 Gayer, A., Saya, A. and Shiloh, A. 'Automatic recognition of
The methodology presented in this paper is designed welding defects in real-time radiography', N D T Intern 23, 3
(1990) pp 131-136
to extract welds having relatively straight line shapes. 10 Rokhlin, S. I., Cho, K. and Gun, A. C. 'Closed-loop process control
Modification of the methodology is needed before it can of weld penetration using real-time radiography', Mater Eval
be applied to extract curved welds, which might be the (March 1989) pp 363-369
11 Yeh, P.-S., Le Moigne, J. and Fong, C. B. 'Computer-aided X-ray
case in some applications. Based on the results of this film interpretation', M M L TR 88-76, Martin Marietta Labora-
study, further research has been planned to develop a tories, Baltimore, Maryland, 1988
more general methodology for the extraction of linear as 12 Basart, J. P. and Xu, J. 'Automatic Detection of Flaws in Welds',
Final Report to Martin Marietta M S S Contract #A71445, Center
well as curved welds. for NDE, Iowa State University, 1991
13 Cloud, E., Fraser, K. and Krywick, S. 'Automated examination
of X-ray welds', Final Report to Martin Marietta MSS, Martin
Conclusions Marietta Electronics, Information & Missiles Group, Orlando,
Florida, 1992
14 Konstantinides, K. and Rasure, J. R. 'The Khoros software devel-
A methodology developed for the extraction of welds opment environment for image and signal processing', IEEE
from digitized radiographic images was discussed. The Trans. Image Proc 3, 3 (1994) pp 243-252
162