Materials Characterization Techniques-Principles and Applications

Eds : G. Sridhar. S. Ghosh Chowdhunv & N. G. Goswanri

N) NML. Jamshedpur-831007 (1999) pp. 135-150

Image analysis in quantitative metallography

GOUTAM DAS
National Metallurgical Laborator y, Jamshedpur-831007
E-mail : gd @ csnml . ren.nic.in

:ABSTRACT

Quantitative metallography (QM) deals with the relationships be-

tween multi-dimensional features of a metallo,graphic sample and its
projection/intersection on a two dimensional plane.Subsequently
QM can be related to other properties of materials. Number of pa-
rameters to be measured for QM are numerous and time consuming
and this has promoted the development of automatic image analyser
by which, even , the routine quantitative measurements can he done
easily and quickly. So, for quality control of industrial products, it
plays an important role. With the availability of reliable automatic
image analyser equipment, the quantitative characterisation of a
metallurgical samples becomes easier and it takes less amount of
time. At the same time reliable data can be generated with suitable
software and hardware attached with modern microscope. The
analysis modules of a metallurgical sample are broad and they are
- grain size analysis, inclusion rating, percentage, volume & area
fraction, porosity, particle size, image features etc. In this review
article the importance of image analyser for the quantification of a
polished and etched surface has been discussed and some of the
applications for quantitative metallography of metallurgical sample
have also been highlighted.

INTRODUCTION
The aim of quantitative metallography (QM) is to get features of three dimen-
sional effect from the measurement of two dimensional objects of a polished
and etched metallic or alloy surface. The three-dimensional features are shown
to he related with physical and mechanical properties of materials. In practice,
these expected properties are never attained but may be approached with any
desired degree of accuracy by simply increasing the number of measure-
ments Care must be taken that the observed samples must be representa-
tive. Modern image analysis system has become an important tool for quan-
titative measurement of microstructural aspects.

135
 C. DAS

The application limits of a material are becoming to narrow and more specific.
so it is necessary to specify and control the microstructure quantitatively,
though the improvement of metallurgical smelting processes have led to the
production of very pure metals and alloys. The visual evaluation of a polished
section of such materials has become increasingly more difficult and time
consuming. As a result, efforts have been made both to improve the methods
of specimen preparation, as well as to automate the process of evaluation. The
QM with modern image analysis plays an active role in the industry for con-
trolling the quality of their products. The statistical out-put from an image
analysis system becomes essential for total quality control (TQC).

In the year 1964 the first instrument Quantimet B was available for Q\1.
Though. it had some problems relating to light source, insufficient automation
and poor linearity that prevented the instrument from being used for routine
checking in the industry. In 1970 the third generation image analyser came
into the market which was also mainly employed for research purposes. Even.
these were not fast enough to carry out routine test work. For day to day
quality control, the image analysis system was employed only after modern
economical computer became available. Towards the end of 1977 a special
version of Quantimet was available for the first time for the routine determi-
nation of inclusions in steels. In the year of 1983, a modern image analyser
TAS-Plus with modern auto-focusing system came to the market.. In additions
to measuring the purity and grain sizes of specimen. it also measures others
parameters like percentage, volume and area fraction and distribution of
phases, image features. porosity, particle size, distribution analysis etcl'-`.

BASIC MEASUREMENT FOR QUANTITATIVE \ ETALLOGRAPHY

The basic measurement can he done by using areal ratio, linear ratio and point
ratio of selected objects on a microstructure. The measurement of volume
fraction can he obtained through some basic equations (mentioned in the sec-
(ion 3) from any of the above mentioned measurement ratio.

The point count method refers to the measurement of number of points that
fall on a specific features of a microstructure. The ratio (Pr) of number of
points that fall on a specific phase (P.) by the number of total points (P.,.) gives
the quantitative measurement. An application of point-count measurement to
a graphite nodules in ferrite matrix is shown in figure I1". Again, the number
of intersection per unit length of a line (drawn on a microstructurc) also gives
the quantitative measurement of phases. Here a test line or linear array is
drawn randomly on a microstructure containing linear features. The points of
intersection along the test line are counted.

136
 G. DA S

N r.n,

Fig. 1 : Application of'point-court grids to graphite nodules in the ferrite matrix

gf,Qrade 60-45-12 ductile iron. (a) 2% nital 100X (b) 2% vital, lightly etched 140 X

To measure the surface to volume ratio of discrete particles (or phases) on a

specimen. the equation can he used as following"):
S 4P
(1)
Va hl
where, S. and V. are mean particle surface area and volume, P = point of in-
tersection of the boundary, I = Tenth of test line drawn randomly on the micro-
structure and h = the end point that hit within the 2nd phase. A schematic
presentation is shown in figure 2 which explain how a surface area can be
converted to volume fraction. Again, using superimposed square grid on a mi-
crostructure the following equation may be employed for surface to volume
conversion'''.
Sa L
(2)
V P1,
where, Pr and Pc are point count and intersection count respectively. Example
is shown in figure 3" -s'.

For a system, with various particles in a matrix the equation will he in the
form of
(SA 2PI
(3)
(V,.), P,,
where , S, = surface or interface area divided by total test volume (surface to
volume ratio ) and Vv = sum of volumes of structural features divided by total
test volume.

137
 G. DAS

Fig. Chalklev method for de terntining the surface-to-vohune

ratio nfdiscrete particles.

o Points counted for

Pp (6 points)
• Points counted for
PL (12 points)

Fig. 3: Superimposed square grid used in the Saltvkov method for detennining the
surface- to-volume ratio of discrete particles. P. represents a point count and
P, represents on intersection count.

BASIC EQUATIONS FOR QUANTITATIVE METALLOGRAPHY

The three basic equations which are mainly employed for the quantitative
measurement of a metallic samples, are as follows' 6.3.91

V`,=AA=Lr =PE, (4)

S^,=(4/tt )LA=2PL (5)

L =2P, (6)

where, AA = sum of areas of intercepted features divided by total test area.

LL = sum of linear intercepted lengths divided by total test line length,
LA = linear traces on the plane of polish surface, L4. = length of features per
test volume and PA = number of point feature divided by total test area.

138
 G. IM S

Equation (4) show the equality of volume fraction to areal, linear and point
ratio of the selected phases on a ma ^ rix. When surface area per unit volume
and linear traces on microstructure are of interest the equation (5) can be
used. For example. inter-phase boundary area (S,,), length of grain boundary
traces can he measured by this equation. In case of estimating linear elements
on microstructure (like dislocation lines, grain edges where three adjacent
grains contact. needle like precipitate particles and slag or oxide stringers) per
unit volume is necessary to measure from points counted on the microstruc-
ture, then equation (6) can he employed. A typical example is shown in figure
4 where, to get dislocation density the number of dislocation etch pits is used.

' , ti ^' '

•• 'I.. 1 •

^,^ , r 2 ore
• !^ ^.•r s
^ rr •- r•
r ' + , • IF

Fig. 4: Etch pits due to dislocation lines in copper

\1'11:1'1' IS AN IMAGE ANALYSER AND HOW IT W ORKS?

Image analyser is primarily concerned with the extraction of useful informa-

tion^and data from an image. which is the ultimate product of numerical out-
put. In an image analyser. the images are digitised which provides description
of complex scene in simpler way from a very large and often ill - defined ob-
jects . Image manipulation can he done by which the representation of images
are more appealing and more understandable . It consists of a microcomputer
to process images. It is user friendly and can he operated manually or auto-
matically by using suitable software . Its performance also depends on
operator's knowledge and how the operator relates the objects to one another
and to the back -ground"'.

139
 G. DAS

The basic operations are as follows"'':

• image input or captured,

• Digitisation of images.
• Selection of different grey levels,
• Segmentation of images into binary regions of interest.
• Reprocessing of binary images,
• Defining and extraction of specific features.
• Mode of extraction.
• Measurement of selected parameters and
• Data output/stored.

SPECIMEN PREPARATION AND MIETH()I)' OF MEASUREMENT

The most optimum specimen preparation is required to evaluate a specimen by

using automatic image analyser. To investigate the lower number of inclusions
in the materials, the inclusions should remain securely embedded in metal
surface with higher degree of flatness. Normal specimen preparation technique
is that grinding on various carborundum papers followed by polishing on felt
clothes using chromium oxide or alumina. For better result silicon carbide
papers are used for grinding and then cloth polishing on synthetic clothes by
using diamond paste. Excessive heat should not he generated while grinding
the specimen. Specimen flatness is necessary for image analysis observations.
Pressure on the specimen while polishing and speed of polishing wheels
should be optimum.

The actual configuration and arrangement of an incase analyser is not de-

scribed here. Some points which are of practical importance for the analysis
have been described. Time requirement for the entire operation on an image
analyser is high. So, the stress is also given to reduce the time obligation for
the analysis in addition to improvement of the techniques and quality. Entire
time requirement for a measurement can be divided into three stages.

• Operation of specimen table,

• Focusing of the images and
• Determination of the parameters to be measured and data analysis.
Specimen table must operate quickly and smoothly and remain free of all
kinds of vibration. The movements may be programmntable. As numerous pa-
rameters to he measured and a large amount of data to be generated. proper
auto-focusing helps to minimise the time requirement of over all operation.

140
 G. /)AS

Higher memory capacity of a powerful microprocessor can he introduced to an

image analyser, by which the time needed for data analysis can be reduced
considerably. At present grey scale and shape are main parameters for quan-
tification. In future the colour differentiation may he an additional parameter
to distinguish various microstructural constituents.

APPLIC :AT1()NS
Grain Size Measurement

Measurement of grain size is one of the most oldest and important techniques
for QM as it influences many properties of metals and alloys. There are sev-
eral methods which have been for the measurement of grain sizes or diameters
though the definitions of diameter is usually arbitrary. These methods are
intercept or plainmatric or by chart comparison.

In the intercept method, developed by E.Heyn in 1903, is based upon the

number of counts by a line (the line may be either straight or curved or cir-
cular) of either the grain or grain boundaries. The plainmatric method, devel-
oped in the year of 1916, is based upon on the determination of the number
grains per unit area (n) which can he directly converted to average ASTM
grain size by"

logn
N = + 1.0000 (7)
log2

where, n = number of grains per square inch area at IOOX magnification.

Normally, to obtain a reasonable ASTM grain sizes, 50 grains in each of a 3
areas must he counted at IOOX magnifications. In the chart comparison
method, the experimental microscopic images are compared to a set of stan-
dard ASTM chart [ASTM E 1382-91 / ASTM E 112 ). A standard ASTM grain
size chart is shown in figure 5' `).

For many random planes, of course, the average intersection values are true
representative of a 3-diamensional parameters. The mean intercept length is
defined as :

L1.
(8)
NL (or Pr) PM

where, NL = number of intercept per unit length, LT = total test length,

P = number of grains or number of grain boundary intersection and
M = magnification.

141
 G. DAS

Fig. 5 : S7rnuiard classilred grain sizes for steel (grain size numbers are
indicated ). X 100 ( reduced to 3/5 size in printing)

Now-a-days, in most of the advanced laboratories, image analysis systems are

in use for quantitative metallography. First, it transfers the incoming images in
to a form which can he easily analysed and this process is known as -trans-
formations" or "convolutions" or convolvcrs" and present grey level opera-
tions are employed for the quantification of multiphases in a polycrystalline
materials. For a greater degree of accuracy a better input images are required.
A modern image analysis system equipped with better hardware and software,
an image like figure 6 can be transformed to a image like figure 7(a,h)"-'
through convolution. Using various analytical techniques, each grain size
groups can he sorted out by using a modern IA system. Furthermore, once
grains have been sorted in grey range or colour groupings by their grain size
values, they can he further classified as to the number and area percentage of
each grain size range in the field of view. These detailed analysis provide a
better characterisation of microstructures than was ever possible before.

142
 Pig. 6 : Iantalnrn sumplr pc'lrslu •d said etched. Note lack of
high/' defined groan /,oundarics 6(K)X

Fig. 7(a) : Same field of view as shown in Fig. 6 after application of

Kirsch filter which highlights discontinuities, in this case , the grain boundaries

Fig. 7( b) : Scone field of riew as figures 6 and 7a inte r application of a series

of image processing steps to reconstruct grain boundaries , compare with Figure 6.

143
 G. 1).4 S

Shape and Size Distribution of (:rains

In some circumstance,, of industrial practice, it can he observed that the t\%o

specimens of a given material can behave in different ways though their
equivalent mean grain sizes are same. That difference may frequently be quan-
titatively related to grain si/e distribution around the mean value. The impor-
tance of size distribution is felt where there are existing heterogeneous distri-
bution of grains resulting in the inconsistency of mechanical properties. Cri-
teria are available for admissible degree of heterogeneity to avoid the delete-
rious influence of the presence of coarse grains in a finer microstructure. In
case of hot rolling there may he, in some cases, partial recrystalliz ation which
results the formation of heterogeneous grains' .

Inclusions

Except few cases (like oxide dispersion strengthened alloys and free machin-
ing steels), it is well established that inclusions play a detrimental role in the
performance of a material depending on their type, shape, size and distribu-
tion. It is defined as particles of foreign materials in a metallic matrix. The
particles are usually compounds such as oxides, sulphides or silicates and
insoluble in the matrix. Inclusions are relatively small, their volume fraction
are low, they are not randomly distributed, and they deform at different rates
than the matrix during working..

The oldest procedure for rating of inclusions are based on standard charts e.g.
ASTM E45. JK (Jernkontoret. Swedish, E 1122). A standard JK chart for
comparative inclusions studies of various type is shown in figure 8'"'. For
general purposes this method is adequate but has it number of deficiencies.
The main problem is related to the reproducibility, time consuming, difficult
for easy handling and availability of standard charts are limited in numbers.
Because of this, some other methods were developed.Sulphur prints was also
developed. The development for colour metallography was also thought of for
their identification. To make the quantification easier and faster, an image
analyser with automatic focusing system plays an important role. In the begin-
ning. the image analyser was employed only for rating of inclusions. The
inclusions present are differentiated between by their reflecting grey values.
Attempt was also made to classify the inclusions by measuring their lengths.
But length measurements alone however proved to he insufficiently character-
istic criterion. In modern image analyser. to quantify inclusions. the most
important parameters used are"':

• Inclusion type.
• The volume fraction,

144
 G. /)AS

^OCOO0
0000

^ Q000
^U. 145
 G.I)AS

• The number of inclusions that present

• The site. & shape and
• The disc ibution of sites.
Generally. inclusions are assigned to a categoery and based on their similari-
ties in morphology and not necessarily on their chemical identity . Automatic
image analyser can he used %kith ASTM1 standards which coyer JK rating
(practice El 122) and E 1245''''. The main non - metallic inclusions are sulphide.
alumina. silicate and ,.lobular oxide. Constituents such as carbides , nitrides,
horides. intermetallic phases may he rated using the ahoye mentioned method.

Spatial Distribution of ('articles ( or Inclusions)

It is well known'" that the tendency for detects to initiate can he lowered by
insuring that inclusions are well distributed and do not cluster together. Im-
proyement of defects initiation tendency can sometimes he achieved by reduc-
ing the clustering. rather than by lowering the overall inclusions volume frac-
tion. Several methods are available for obtaining spatial distribution of sphere
shaped particles. Procedures are also available for convex particles of arbitrary
shapes. ellipsoids etc. Although. the exact solution for regular shapes are
available, but for irregular shaped particles the situation is not so simple.
Assumptions are required with a loss of accuracy. From the true two-dimen-
sional size distribution, the 3-dimensional pictures can he drawn.

The three main type of measurements are as tolloyys

• Diameters,
• Section areas and
• Section chords.

Three main parameters : mean diameter (D). the standard deviation [a(D)J and
the number of particles per unit volume (Ny,) are necessary for representing
a site distribution curve. These parameters can he obtained from the appropri-
ate experimental data or from the analysis of particle size distribution. Figure
9 shows a schematic representation of three type of measurements.

Porosity

Porosity is a common feature for ordinary cast products. While melting metals
and alloys - oxvaen. nitrogen and hydrogen get trapped and during solidifica-
tion they are released leaving behind considerable amounts of pores of differ-
ent sites. Deoxidising process by simply adding deoxydiser during melting
these gases can he removed. However, in practice for many cases some remain

146
 G. RAS

Areas Chords

Fig. 9 : Sc•hen,atir p,z sen:(mon of three mail t yes of measurements

made on planar section

trapped in the ingot. So, quantification of such porosity in terms of sire dis-
tribution are essential for assessing the health of materials. A modern image
analyser can he employed for exactly quantify the pores present in the mate-
rials. Care must he taken to differentiate inclusions and pores.

Multiple Phase Area percentage

In view of enormous development of processing technology the application of

materials are more and more specific, so it has become important to quantify
microstructure consisting of several phases. The mean free path or mean phase
intercept of individually dispersed phase or particles in a matrix is given by''":

4Vv (9)

SV
Where. V,, = the volume fraction of the obtained phase and S, = the interface
area. The above equation can also be extended for multiphase phases.

An additional parameter, similar to the mean phase intercept, is the mean ran-
dom spacing and is defined as'="

6 = - (10)
N1
Where. N, = particle (or clusters) intercept counts. 6 may he interpreted as a
mean centre to centre distance between dispersed particles or clusters of a par-
ticular phase.

The fundamental parameters used for mean free path relations are V and S'
and. the volume fraction of phase "i" and the surface area per unit volume of

147
 G. DAS

boundary shared by phase 'i' and 'j'. Vcan he found by point count procedure
which can he related to the fraction of points on a grid occupying the phase
1,. P', by

V'=1" (I1)

Similarly. S" can he related by (S'= 2P`'r ), where. P'L = boundaries intercept
per unit test line length.

Modern image analyser is often utilise to measure area percentage of a

multiphase system. Measurement of multiphase in a materials is done by mea-
suring the intensity of grey levels. The grey range of Ist phase (minimum grey
level) would he detected and its area percentage is measured over several
fields of view. By this way the detection of 2nd and 3rd phases can he done
and areas are measured. This technique is requires a certain amount of dexter-
ity and adjustment of thresholds for detection of individual phases. A video
level histogram can show distinct peaks relating to the grey shades ranges of
different phases. A typical video level histogram of four phases materials is
shown in figure 10"''. Again, after detection the various phases of the images
can he coloured correspondingly to the colour assigned to the grey ranges.

Fig. 10(a) : Typical AZS re fra( t,n nuurrial shat, rn,' lour diswict phases : P orosit.%,
Silica, Alumina (reed Zhrnnia 400X.

148
 G. D .4 S

I PK I

Fig. I0(b) : Typical video level hislograun (4".1o ur-phase / IZS material seen in
Fig. /0( a). Each peak corresponds to a separate phase

CONCLUSIONS

Quantitative data. extracted from polished and etched samples (if required),
can he used for various purposes like quality control, process control, failure
analysis or research purposes . A systematic and positive approach should he
taken for precise measurement of various parameters . From sample prepara-
tion to data analysis - all steps must he properly and carefully performed to get
best results. Sampling or identification of test samples from a large piece of
materials or from a big component is most important. Specimen must he pre-
pared properly, both during polishing and etching (if required). The best pos-
sible image should he fed into image analyser for proper quantification of
various parameters of an image . Manual method can also be employed for
image analysis but it is less sensitive and time consuming . Accuracy can also
he improved simply by increasing the number of test measurements of fields
or features . When all the steps will he fulfilled the final results will be precise
and reproducible.

REFERENCE:
I. O.P. Mcxli , Metallography. Work - shop on Characterisation of Engg. Materials. RRL,
Bhopal, 1997.
2. D.D'Angelo. V.Rcgis, U.Franszoni and M.A.Rossi, Creep and Fracture of Engg. Mate-
rials & Structure, Conf. Proc.. Swanesa. U.K., 1984.
3. W.Phol & A. Fischer. Pract. Met., 27,101. 1990.
4. M.Goken & H.Vehrff, Scr. Mat.. 35. 9.983. 1996.
5. A. Chadda . V.K.Joshi & D.J.Murthy, Bull. Mat . Sci.. (India ). 8.2. 231, 1986.

149
 G, AA A
6. Metals Handbook, 9th Edition, Metallography and Microscopy. V.9, ASTM Publica-
tion. Metals Park. Ohio. 1988.
7. Eds. R.FBunshan, Techniques for the direct observation of structure and imperfection,
Part 1. Interscience Publication, 1968.
8. C.M.Chermont, Pract. Met., 21. 12. 602. 1984.
9. V.S.Kovalenko, G.I.Levistakava, V.B.Enushenko and V.P.Cherkosava. Metal Gronontdn
Promnt., 2, 18, 1985.
10. A.Morris. Recent Advances in Image analysistech.. Quantitative Microscopy and Im-
age Analysis. Ed. David J.Diaj. ASM International. Metals Park. Ohio. 1994.
11. G.S.Cole, Theory and Concepts of Computer Aided microscopy, Quantitative Micros-
copy and Image Analysis. Ed. David J.Diaj. ASM International, Metals Park, Ohio.
1994.
12. Reid]. Metallography. 14. 2, 119, 1981.
13. B.F.Shepard. The P-F characteristic of steel, TASM. 22, 39.1934.
14. E.Exner. Analysis of Grain and particle size distribution in metallic materials. Ent..
Met., Rev.. Review 159. 17.25, 1972.
15. F.Vander Voort, grain Size measurement. Practical application of quantitative metal-
lography. ASTM STP 839. J.L.McCall and J. H. Steele Jr. Eds., ASTM Philla. PA.
85,1984.
16. Yu. Lakhtin, Engg. Phy. Met. and Heat Treatment, Mir Publisher Moscow, 1990.
17. I.C.Oppenticim. Pratical application of advance image analysis techniques, Quantita-
tive Microscopy and Image Analysis. Ed. David J.Diaj, ASM International , Metals
Park. Ohio. 1994.
18. H.Mathy, Common Eur. Commision, Report. 1987.
19. Non-metallic inclusion in steel , Course organized by NML and SINTEF, 1988.
20. M.V. Kuhtin . V.P.Cheremnikh. IZV. AKAD. NALK. SSSR. Met. 5, 123. 1979.
21. Non-metallic Inclusion in steel , Part-I to IV, R.Kisseling, Institute of Metals, London.
1989.
22. Bearing Steel-The Rating of Non -metallic Inclusions. Symposium Proc., ASTM,
Philladelphia, 1974.
23. Inclusion rating in steel. ASTM, E45/El 122, ASTM(03.01) 1997.
24. Inclusions and their influence on materials behaviour. Eds., Ravi Rungta. ASTM pub-
lication . Chicago. 1988.
25. P.N. Crepeau, A. M. Gokhalc and C. W. Meyers. JOM. 41.2. 16, 1989.

1 50