SEM Mages yy 
“Urn yi 3 | 
Hon fey § 
x hi and| 
- Simple Vow 
00). In on 
trough a sim 
<togone Of size) 
A three times), [| 
i Solution of 10 
rong dark cop ; 
1 usual indus : 
fication 500), 7 
he topology of y 
amed with a Pay 
Fig. 3: Illustration of the automatic discrimination 
between carbides and & phase: a) initial 
microstructure ; b) binary image including & phase 
and carbides ; c) suppression of the carbides 
detected by erosion then partial reconstruction 
fiön Of” typ 
mee EM 
Results and discussion 
The present paper does not focus on the relations between Vv(8) and the process parameters, which have 
been investigated in the frame of an industrial contract. It is limited to a preliminary discussion about the 
reliability of the image analysis procedure to measure Vv(8). From this point of view, our main result is 
illustrated in Figure 4. It shows that a good agreement is found between the Vv(3) values measured by X- 
ray diffraction and by BSEM image analysis, while the volume fraction is systematically strongly 
overestimated by the usual LM image analysis procedure. Moreover, no systematic trend can be found for 
this estimation bias. This hinders an even qualitative comparison of Vv(8) values measured by the second 
image analysis technique. 
$= Sn Fig. 4 : Comparison of the values of & 
croscopy in ba "i phase volume fraction measured 
SOP) i ; ZA either by X-ray diffraction (y-axis) 
Lr a , ns - and by image analysis of light 
In SOURIS microscopy and scanning microscopy 
WH £3 ~ Tok 3 r 0 Scanning mic] in back-scattered electron mode (x- 
3 . > Le Light mic. , axis) 
= 10 15 20 
delta volume fraction measured by image analysis (%) 
The very large bias pointed out in the case of LM image analysis have at least two origins : 1) the limited 
resolution of light microscopy to detect thin 8 lamellae and ii) the possible enlargement of the 0 phase due 
to the etching. Observation of polished/etched surfaces by scanning microscopy in secondary electron 
97