Fuzzy Rule Based Classification and Quantification of Graphite Inclusions from Microstructure Images of Cast Iron

Abstract

The quantification of three classes of graphite inclusions in cast iron, namely, nodular, flake, and irregular, is the most important process in the foundry industry. This classification is based on the ISO 945 proposed morphology of graphite inclusions. This work presents a novel solution for automatic quantitative analysis of graphite inclusions into the three mentioned classes. The proposed work comprises three stages, namely, preprocessing of micrographs, classification of graphite inclusions, and then quantification of inclusions in each class. An effort has been made in this work to propose a minimum set of features to represent graphite inclusion morphology. The method employs just two geometric shape descriptors: the diameter ratio and the area ratio. A fuzzy rule based classifier is built using known feature values that are efficient in the classification of the three classes of graphite inclusions. The proposed method is automatic, fast, and provides the basis for determining many more morphological parameters that can be determined with the least effort. The results obtained by the proposed method are compared with the manual method. It is observed that the results obtained from the proposed method are useful in the optimization of cast iron manufacturing in the foundry industry.

Microstructure Image Analysis for Estimating Mechanical Properties of Ductile Cast Iron

“…Hence, the accuracy in estimation of nodularity value of material is important. The prevailing visual based manual technique often produces biased results due to human natural fatigue and visual limitations [2,6,7,9,13]. Manual based results are highly dependent on expert's visual judgment and not repeatable.…”

“…Several investigators have shown a correlation between nodularity and mechanical properties but, again, the correlations are based on visual estimates of nodularity and ferrite quantity [1,4,8,11,2,13]. With the advent of modern computational facilities and analytical tools, it seems appropriate to reconsider this analytical issue.…”

“…It is tolerant of imprecise data and can model nonlinear functions of arbitrary complexity. Fuzzy logic is built on the structures of qualitative description represented by linguistic variables in a natural language [13,15,16].…”

“…The microstructure images of materials ( Fig.1) are processed using digital image processing techniques and the microstructure properties are assessed. The microstructure properties of the material and the mechanical properties are closely related [3, 5,6,7,9,11,13]. The ductile cast iron is produced to have a wide range of properties through control of the microstructure.…”

Texture-Based Metallurgical Phase Identification in Structural Steels: A Supervised Machine Learning Approach

“…Other notable works that employed threshold-based techniques include segmentation of ferritic-martensitic dual phase steel by Burikova et al [22] and identification of bainite in Fe-C-Mo steel by Ontman et al [23]. In addition to these studies, artificial intelligence-based image segmentation is also found in the literature, which includes clustering [24], neural networks [25], fuzzy logic [26], and support vector machine (SVM) [27] methods for identification of microstructures in various metals. A sophisticated image processing technique that accounts for additional distinguishing features is required for accurate phase identification in a microstructure with phases that have overlapping pixel intensities.…”