Professional Documents
Culture Documents
net/publication/291974416
CITATIONS READS
0 1,522
13 authors, including:
Lihong Su Huijun Li
University of Wollongong University of Wollongong
49 PUBLICATIONS 755 CITATIONS 445 PUBLICATIONS 10,386 CITATIONS
Cheng Lu Jintao Li
University of Wollongong University of Wollongong
595 PUBLICATIONS 12,425 CITATIONS 30 PUBLICATIONS 350 CITATIONS
Some of the authors of this publication are also working on these related projects:
Fibre Optic Acoustic Emission Measurement System for Civil Engineering Applications View project
All content following this page was uploaded by Jintao Li on 05 March 2019.
Lihong Su1, Huijun Li1, Cheng Lu1, Jintao Li1, Ian Simpson2, Frank Barbaro1, Leigh
Fletcher3, Lei Zheng4, Mingzhuo Bai4, Jianlan Shen4, Liqiang Fan4, Guodong Xu4 and
Guangzheng Gui4
1
School of Mechanical, Materials and Mechatronic Engineering, University of Wollongong,
Australia
2
The Australian Pipeline Industry Association, Australia
3
Welding and Pipeline Integrity, Australia
4
Baoshan Iron & Steel Co., Ltd., China
Abstract
In the current work, a MATLAB program based on image analysis was developed to evaluate
the images of centreline segregation by automatic measurement of segregation features. The
original grey scale images of the centre line segregation were reduced to black and white
images for automatic measurement. The black features of the image, representing areas of
element segregation, were identified according to predefined criteria. The degree of
segregation was evaluated based on equations which were developed by considering different
parameters of the segregation segments, and also was evaluated by rules gleaned from
experienced steelmakers.
Introduction
575
Fig. 1. Example of a macro etched steel slab image. (Image courtesy of Baosteel)
Centreline segregation develops on a large scale and is one of the most important defects
since it is impossible to reduce subsequently by thermo-mechanical post-treatments [3, 4].
The common methods of centreline segregation evaluation used in scientific research differ
from that used in steel manufacturers. The methods used in laboratories include optical
emission spectrometry with pulse discrimination analysis (OES-PDA) [5], ultrasonic test [6],
emission spectrometry [7], electro microprobe examination [5], X-ray mapping [8], etc.
These methods provide quantitative information such as the exact contents of different
elements and their distribution. However, these methods are costly and time-consuming, and
unsuitable for industry use. Furthermore, these methods usually focus on a localised area, so
the results are sometimes not representative of the whole product.
The most common ways of evaluating centreline segregation in industry are sulphur printing
and macro etching. Sulphur printing reveals the size and the distribution of mainly
manganese sulphide inclusions in steel. This method is insensitive when applied to low
sulphur steels (less than about 0.006 wt%) [9]. Macro etching is a method of revealing
centreline segregation by chemical etching using a range of possible etchants. The image
from macro etching is similar to that of sulphur printing.
In this paper, line pipe steel is used to discuss the importance of centreline segregation and to
assess the evaluation methods. Severe centreline segregation can have adverse effects on the
mechanical properties, weldability and service performance of line pipes. The general
requirements for line pipe steels are high strength and ductility, good fatigue and fracture
resistance. In the meantime, the weldability must be guaranteed. Usually, these properties are
realised by decreasing C contents, increasing Mn and adding alloy elements such as Nb, V
and Ti in carbon steels. S is one of the most harmful elements in line pipe steels. It severely
decreases fracture toughness and the steel’s ability to resist hydrogen induced cracking (HIC)
and sulphide stress corrosion cracking (SCC) [10]. Modern line pipe steels contain ultra-low
S (less than 0.005 wt% and often less than 0.001 wt%), therefore, the traditional sulphur
printing method is not suitable for revealing centreline segregation in line pipe steels, and
macro etching must be used. [8].
Traditionally the sulphur prints and macro etched surfaces were evaluated by experts
comparing them to standard charts, such as the charts developed by Mannesmann [11]. This
is a subjective method and the evaluation results from different experts can vary. The most
pressing problem with the Mannesmann method is its coarse scale. There are only five rating
levels, which can cause disagreements when the segregation pattern lies between two ratings,
e.g. worse than a rating of 2 but better than 3. Besides the direct image comparison methods,
many steelmakers have developed their own standard procedure for the evaluation of
centreline segregation. Several manual, real size-based measuring methods are used by
different steelmakers. These methods are more consistent since there are rules to follow but
576
they are very time consuming. In addition, these procedures are usually considered as internal
information and different steelmakers use different rules.
In order to develop objective evaluation techniques, recently, automatic evaluation tools have
also been developed [12]. The process involves image analysis by computer programs which
allow automatic measurement of segregation features. The current paper reports on a study of
the automatic measurement of centreline segregation in continuously cast line pipe steel slabs
by a computer based image processing technique.
Methods
The images for automatic measurement were obtained by digitising photographic images of
macro etched continuously cast steel slabs. Macro etching was conducted by hot etching
using 50% industrial hydrochloric acid at 70±10 °C for ~20 min. Images of steel slabs with
centreline segregation at various levels were obtained. The five standard Mannesmann
images given in Ref. [11] were used as a reference. The segregation ratings used in the
current study were based on the Mannesmann scales.
A MATLAB program was developed to perform the automatic measurements. The images
were first transformed into grey-scale images after digitisation. A histogram of grey-scale
values for each image was then plotted and a suitable grey-scale value was chosen as a
threshold according to distribution of the histogram (Fig. 2). The pixels with grey-scale
values lower than the threshold were defined as black and those higher were defined as white.
By this method, the grey-scale images were transformed into black and white binary images
(Fig. 3).
Fig. 3. An example of grey-scale image (top) and corresponding black and white image
(bottom).
The centreline was defined as the line with maximum number of black dots (one dot was
equal to one pixel). Since most of the areas in the images were white pixels, the region that
577
was used for further analysis was chosen by extending ±20 pixels from above and below the
centreline. Black dots and black segments were defined according to the following criteria:
Choose a black dot as a target point, define it as a black dot if at least one of its eight
neighbouring dots is black; Define a series of continuous black dots as a black segment,
regardless of the length of the segments.
Once the black and white images were ready, the program started to measure various
parameters for centreline segregation: the number of black segments N, the percentage of
black segments over the total area P, the average length AL and standard deviation SL of black
segments.
Two methods were used to evaluate the centreline segregation and give ratings based on the
Mannesmann scales: one using the number, percentage, length or area of the black segments;
the other uses rules according to the size and number of black segments per 100 mm.
Results
Two main factors: number and continuity of the segregation features are generally considered
the most important parameters influencing the severity of centreline segregation. Based on
this rule, the following equations were developed to calculate the segregation degree:
(1)
(2)
(3)
where D is segregation degree, P, N, AL and SL are the same as stated above, SL’ and AL’ are
the standard deviation and average length of white segments.
Eq. (1) is adopted from Ref. [12]. The equation considers the continuity and width of the
segregation features, and considers that big dots of segregation are more dangerous than
small dots even if they give the same percentage. Applying Eq. (1) to the standard
Mannesmann charts, the results show that the segregation degree calculated by the equation
returns a close to linear relationship with the Mannesmann scales, as shown in Fig. 4 (a).
Eq. (2) is developed by considering the percentage of black segments in the centreline as the
main factor, and combined with the number of the segments. The percentage of the black
segments is proportional to segregation degree as more black segments would naturally lead
to worse segregation. The number of black segments, however, is inversely proportional to
segregation degree, because when the percentages are the same, fewer in numbers means
longer in length and more severe in segregation degree. It can be seen from Fig. 4 (b) that
when applying Eq. (2) to the standard Mannesmann charts the relationship of segregation
degree and the Mannesmann scales is also close to linear.
Eq. (3) is developed by considering that the average length of black segments and their
standard deviation also have positive influence upon the segregation degree. Fig. 4 (c) shows
the segregation degree of the five standard Mannesmann charts when using Eq. (3). It shows
an exponential relationship rather than linear, which is in agreement with the results shown in
578
Ref. [13], where the hardness values and the segregation ratios of P and Mn both have a close
to exponential relationship with the segregation scale.
Many steelmakers use a manual method which measures the actual size of segregation
features to rate the severity of segregation. The method measures the size of dots and the
number of dots per 100 mm and then rates the segregation according to certain rules [14].
The current program can relate the pixels of the segregation image to real size and
automatically measure the real size of segregation features. A new assessment rule has been
developed to rate the segregation based on Mannesmann scales. Fig. 5 shows a size
measurement example. The real image is shown in Fig. 1. The pixels in the processed black
and white image are related to real size by the ruler shown in the image. The results of the
measurement are shown in Table 1. The rating is 2.5 based on Mannesmann scales and is
very close to the rating given by experts using the image comparison method.
When the same image is evaluated by the equations, the segregation degree falls on the trend
line of the standard Mannesmann charts shown in Fig. 4 and the rating is also in agreement
with that evaluated by the size measurement rule and experts’ opinion when converted to
Mannesmann scales. However, the number of images that have been evaluated by the
program is still limited, especially the images with segregation rating larger than
Mannesmann scale 3. These images are difficult to get, since most steel slabs have ratings
between Mannesmann scale 2 and 3. Thus it has not been possible to judge which equation
best describes the relationship between segregation features and segregation degree over the
full range of Mannesmann ratings. The self-developed size measurement rule also needs
further adjustment by testing more images.
579
Summary
Centreline segregation can be harmful to the weldability and integrity of fabricated steel
products, especially pipelines. At present, there is no international consensus on evaluation
methods of centreline segregation. In the current work, a MATLAB program based on image
analysis was developed to evaluate the images of centreline segregation by automatic
measurement of segregation features. The segregation severity was evaluated from grey-scale
graded images based on equations which were developed by considering different parameters
of the black segments. These assessment results were also evaluated by rules gleaned from
experienced steelmakers. By applying the two evaluation methods to certain segregation
images, it was shown that both methods can correctly relate the segregation severity and
macrostructural images. Further the calculated ratings are close to that rated by experts using
the traditional image comparison method.
Acknowledgements
The authors would like to thank Baosteel-Australia Joint research and development Centre
(BAJC) for its financial support. The Energy Pipelines CRC (EPCRC), supported through the
Australian Government’s Cooperative Research Centres Program, is also acknowledged for
its in-kind support.
References
580
14. S. Rapp, “Requirements of the MAOP rule and its implications to pipe procurement,”
(Presentation at INGAA Foundation Best Practices in Pipe Procurement and Manufacturing,
Westin Galleria, Houston, Texas, 2010).
581