Broaching Tool Degradation Characterization Based on Functional Descriptors

With rapid advancements in sensing technologies and computation capabilities, high-resolution machine vision data have become available for various manufacturing processes. For machining, the use of machine vision data has shown great promise in machining tool condition monitoring, a critical factor for final product quality. Extensive research has been performed on wear characterization using intensity-based methods, but limited work has made use of process knowledge for image processing phases. Additionally, previous work focuses on single cutting edge machining tools, but no methods have been proposed for multiple cutting edge machining tools, such as broaches. In this paper, a process knowledge-based image filtering method is proposed to eliminate within-image and between-image noise to obtain effective wear region(s) for each cutting edge on a broach. In addition, these wear regions across multiple cutting edges are jointly described by fitting their relationship with each cutting edge’s respective chip load. Finally, the extracted model parameters are used for unsupervised learning to determine the entire tool’s degradation levels from a training dataset. A case study is introduced to show the effectiveness of the proposed methodology using a hexagonal broach.