Wear Debris Classification of Steel Production Equipment using Feature Fusion and Case-based Reasoning

Abstract

Wear debris classification is of great significance for identifying machine wear states. In this paper, a method of wear debris classification using feature fusion and CBR is proposed. The method integrates local feature LBP, global feature FD and Tamura coarseness, and then the fused features are applied in CBR system with different weights and different similarity, which is adaptable, extendable, modular and fast. The results show that the subdivision of wear debris images into size 32*32 when calculating LBP is helpful for improving the classification, the combination of local features and global features can get better results. The comparative experimental results of different classification methods show that the CBR system has the shortest time-consuming while maintaining high classification accuracy.