Cutting Tool Wear: Data Acquisition and Automation of Wear Measurements on End Mills
dc.contributor.author | Andersson, Pontus | |
dc.contributor.department | Chalmers tekniska högskola / Institutionen för data och informationsteknik | sv |
dc.contributor.department | Chalmers University of Technology / Department of Computer Science and Engineering | en |
dc.contributor.examiner | Ranta, Aarne | |
dc.contributor.supervisor | Yu, Yinan | |
dc.date.accessioned | 2022-12-02T13:54:01Z | |
dc.date.available | 2022-12-02T13:54:01Z | |
dc.date.issued | 2022 | |
dc.date.submitted | 2022 | |
dc.description.abstract | Tool management has an important impact on machining operations. As cutting tools are used, they are worn out and thus they need to be changed. The cost of cutting tools is a significant cost of a machining operation. Likewise using a tool too long can damage the produced part. To balance these two factors the wear on cutting tools must be analyzed and measured. Manual measurements take a long time and cost a lot of money, thus an automatic process for measuring cutting tool wear would be beneficial. Many different approaches for automatic tool wear measurement have been researched. This work proposes a method for automatic wear measurement of end mills using computer vision. The method uses a line scan camera and rotates the end mill to collect an image of the entire tool circumference. A two-step approach was then used to analyze the images, the first stage used a Mask R-CNN model to identify the worn areas. After that, the worn area was quantified by measuring the maximum and average depth of flank wear and the maximum fracture dept. Four Mask R-CNN models with different hyperparameters were trained. The methods were tested on three types of end mills all used for machining super-alloys. The best model resulted in a mean absolute measurement error of 13 % for average flank wear depth, 16 % for maximum flank wear depth, and 24 % for maximum fracture depth. To investigate the inclusion of new tools, experiments were also done on one end mill type that was not used when training the Mask R-CNN model. This resulted in an average error of 48 % for average flank wear depth, 40 % for maximum flank wear depth, and 48 % for maximum fracture depth. The results are promising, however, more research is needed to evaluate how this could improve tool utilization and machining processes in an industrial environment. | |
dc.identifier.coursecode | DATX05 | |
dc.identifier.uri | https://odr.chalmers.se/handle/20.500.12380/305870 | |
dc.language.iso | eng | |
dc.setspec.uppsok | Technology | |
dc.subject | computer science | |
dc.subject | cutting tool wear | |
dc.subject | end mill wear | |
dc.subject | deep learning | |
dc.subject | neural network | |
dc.subject | image analysis | |
dc.title | Cutting Tool Wear: Data Acquisition and Automation of Wear Measurements on End Mills | |
dc.type.degree | Examensarbete för masterexamen | sv |
dc.type.degree | Master's Thesis | en |
dc.type.uppsok | H | |
local.programme | Computer science – algorithms, languages and logic (MPALG), MSc |