Home Publications Research Grants Inventions & Patents Awards Additional Research Activities Faculties & Institutions Research Areas

Deep learning based drill wear segmentation and analysis of the wear progress

Thamm A, Thamm F, Wiedemann M, Bretschneider J, Sikorska M, Maier A (2024)

---

Publication Type: Journal article

Publication year: 2024

Journal

DOI: 10.1007/s12008-024-02045-0

Abstract

Cutting tools and their condition monitoring play a major role in the world of manufacturing. The ultimate goal is to strategically prolong the intervals between tool replacements, optimizing cutting times and consequently minimizing costs. Achieving this requires precise quantification of tool wear, a process traditionally marked by its time-intensive and laborious nature. An automatic drill wear detection for two types of a damage (flank and corner wear) has been developed. This approach utilizes deep learning, specifically the nnU-Net framework. The model has been tested using a five-fold cross-validation approach, and the results are promising. This approach has been assessed by comparing the ground truths, which were manually annotated by experts, with the network’s predictions. For flank wear, a Dice Similarity Coefficient (DSC) of 0.93 and an Intersection over Union (IoU) of 0.87 has been achived. For outer corner wear, a DSC of 0.95 and an IoU of 0.90 has been attained. The whole wear segmentation yielded a DSC of 0.93 and an IoU of 0.87. With this presented method, the user automatically receives information about the degree of the two types of wear. Measurement of the wear width is no longer necessary. This method can assists in the development of wear parameters over time, commonly referred to as the wear-tool life curve. This curve optimization enables precise adjustment of cutting parameters, encompassing machine settings, environmental conditions, workpiece materials, and machine tool characteristics for future batches of workpieces. Additionally, this information serves as a foundation for empirical modeling of wear progression.

Authors with CRIS profile

Involved external institutions

How to cite

APA:

Thamm, A., Thamm, F., Wiedemann, M., Bretschneider, J., Sikorska, M., & Maier, A. (2024). Deep learning based drill wear segmentation and analysis of the wear progress. International Journal on Interactive Design and Manufacturing. https://doi.org/10.1007/s12008-024-02045-0

MLA:

Thamm, Aleksandra, et al. "Deep learning based drill wear segmentation and analysis of the wear progress." International Journal on Interactive Design and Manufacturing (2024).

BibTeX: Download