SENS 
6 AUTOMATIC IMAGE TREATMENT MINIMIZING ERROR IN THE GRAIN SIZE 
DETERMINATION 
F. Pires!, J. Nascimento’, L.Abud‘, I. Caminha‘ 
! Instituto Nacional de Tecnologia - Rio de Janeiro — RJ, Brasil 
ABSTRACT 
In the present work, low alloy steel and austenitic stainless steel samples were evaluated in order to 
identify the more usual type of errors encountered in the grain size measurements by automatic image 
treatment. The analyses were carried out in an analyser image system with software based on ASTM 
E112 and ASTM E 1382 standards. The results obtained in both materials, after the application of 
filters and complementary logical operations, showed a substantial reduction in the discontinuties of 
the grain boundaries, leading to a final result of average grain size very close to the actual size. In the 
low alloy steel, the ASTM average grain size before the image treatment was 6.37 and after the 
treatment 8.43. For the austenitic stainless steel, the values before and after the image treatment were, 
8.91 and 6.30, respectively. The true values for the low alloy steel and the austenitic stainless steel, 
A obtained by interlaboratorial comparisons, are 8.50 and 6.00. 
1. INTRODUCTION 
The materials’ grain size is directly related to the resulting microstructure of the manufacturing process. 
As known, for applications where the microstructural characteristics are relevant, for example hardness 
standard blocks and surgical implants, the mean grain size is responsible for the product acceptance or 
not (1-2). 
The automatic image treatment has been an important tool for the microstructures’ analysis and 
interpretation, allowing measurements that would be impossible to be made by manual methods, with 
higher analysis speed and consequent increasing of the statistics’ quality of the results, besides the 
possibility of developing automatic procedures using artificial intelligence. 
This work presents an experimental methodology to maximize the contrast between matrix and 
boundary, minimizing the detection of scratches and artifacts as part of the boundary, as well as the 
influence of the metallographic etching in the grain size determination. 
2. MATERIALS AND METHODS 
Two kinds of steels were studied in the present work, which the grain size measurements were 
optimized by automatic image analysis: a low alloy steel with low carbon, sulfur and phosphorus’ 
content, used in the manufacturing of Brinell and Rockwell B hardness standard blocks and an 
austenitic stainless steel, used in the manufacturing of temporary orthopedic implants like plates, 
screws. profiles and wires 
175