A Cluster-Profile Comparative Study on Machining AlSi7/63% of SiC Hybrid Composite Using Agglomerative Hierarchical Clustering and K-Means

Abstract

Clustering techniques are used to group the data based on the structure or through classification to reduce the mathematical complexity of large datasets. The hierarchical and partitioning approach are two broad clustering/classification techniques in data-mining. An attempt has been made to study the possibilities of taking the advantages of these approaches to machining. AlSi7/63% of SiC hybrid composite prepared by stir casting technique is machined using the Abrasive Water Jet Machine (AWJM) for Taguchi’s L27 Orthogonal Array (OA). Water Pressure, cutting distance, and cutting Speed are taken as independent parameters. Material Removal Rate (MRR), Kerf Angle (KA) and Surface profile Roughness (Ra) are taken as dependent responses. Support Vector Machine (SVM) classifiers with Agglomerative Hierarchical Clustering (AHC) classifies L27 OA into three classes of nine observations each. To compare and explore the difference between the partitional clustering and hierarchical clustering techniques at the same level of class, the study on K-means value is taken as 3 because of AHC group L27 OA into three classes. The value of K is fixed with three and it group into three classes of nine observations each. XLSTAT software is used for the analysis of AHC and K-means. Further, linear regression equations are developed for each class/classification of AHC and K-means and compared with the experimental observations. The analysis reveals that K-means classification based on the partitioning approach fits best with the experimental observations. AHC develops a single equation for all the classes, whereas K-means develops individual equations for all its classes.