Partitioned abrasive belt condition monitoring based on a unified coefficient and image processing

[Abstract] Abrasive belt condition (BC) monitoring is significant for achieving profile finishing precision and quality in grinding of difficult-to-machine materials like Inconel 718. While indirect signal-based BC monitoring methods are ineffective when varying grinding parameters, existing image-based direct monitoring methods currently suffer from a lack of: (i) a unified and quantitative definition of the belt condition; (ii) in situ tool-surface image capture and relevant feature extraction; and (iii) continuous monitoring of the entire belt conditions. This paper proposes a partitioned BC monitoring method that is adaptable to ever-changing grinding conditions. Based on the belt surface analysis, a unified BC coefficient is quantitatively defined by using two critical BC-dependent features, the average area and number of worn flats of abrasive grains per unit area. The belt surface image is in-situ captured from moving belts and is preprocessed to eliminate image defects in a unified form, then the entire belt is partitioned, and finally the image features are extracted by Gabor filter and K-means clustering. The proposed robust method which has a maximum relative repeatability error of 9.33%, and less computation was validated by the experimental results. This study provides an adaptable and efficient way for continuously monitoring the conditions of the entire belt and the grinding area.



友情链接: 上海交通大学 | 上海交大材料学院 | 上海市激光制造与材料改性重点实验室 | 智能化焊接与精密制造研究所 | 机器人焊接智能化技术实验室 

版权所有 © 2020-2023 上海交通大学先进材料智能制造实验室  All Rights Reserved     沪交ICP备20200283号