K-MEANS AND AGGLOMERATIVE HIERARCHY CLUSTERING ANALYSIS ON THE STAINLESS STEEL CORROSION PROBLEM

Abstract

Stainless Steel (SS) is a material that is widely used in various fields because it is resistant to corrosion. However, if SS is exposed to heat at high temperatures for a long period of time, a sigma phase, namely the Fe-Cr compound, will form, which indicates that corrosion has begun. The appearance of this corrosion can be detected through color changes on the SS surface, ranging from light brown to dark blue. Corrosion events will be observed through the distribution of color on the sample surface at the location selected through the SS microstructure image. Cluster analysis will be used to group the colors on the surface of the SS sample through the images used. The results of cluster analysis can be used to identify SS color which indicates the appearance of corrosion in the sample. In this research, we will examine the determination of many clusters for K-Means and Agglomerative Hierarchy with Ward's Criterion, Single, Average, and Complete Linkages. In addition, the model quality measure was tested with Silhouette Coeficient. Single linkage gives the worst results because it gives the impression that only one dominant color appears so it can be said that it is unable to distribute each color to the specified cluster. Likewise with Average because the number of clusters cannot be determined with certainty. On the other hand, the K-Means results are similar to Ward's results, this is reasonable because the basic idea of both is to find the minimum distance between each object and its center, in this case the average is used as the measure of the center, while the results that are most similar to the original image are clustering uses complete linkage. These results can be used as recommendations for academics and practitioners in the fields of Statistics, Mathematics and Materials Engineering in the subsequent analysis process to solve SS corrosion problems.