Tool condition diagnosis
- 01.04.2019 to 31.03.2021
- Organizational Unit:
- Chair of Manufacturing Technology, Cutting Technology
- Current: Pre-competitive research Scheduled: German Federation of Industrial Research Associations AiF and Federal Ministry of Education and Research
- Machining industry
- Manufacturer of cutting tools, coating and coolant lubricants
- Machine tool manufacturer
The current change of the tool’s cutting edge as a result of tribological stress is for manufacturers of crucial interest when it comes to questions of process safety and productivity. Our research and development activities therefore aim for an in-process tool condition diagnosis, which decides on the usability of the current cutting tool with high transparency and reliability. In order to achieve this objective, our research team is working for you on a holistic approach that combines the possibilities and advantages of direct and indirect methods for the analysis of tool condition.
Based on process and machine integrated sensor signals, a continuous quantitative estimation of the tool condition is realized, independent of the engagement situation. Existing restrictions resulting from the limited quality of the data sources are compensated by using additional knowledge from a direct and objective measurement of wear characteristics. Thus, the estimation error is reduced and therewith the quality and robustness of the in-process tool condition diagnosis is significantly increased compared to state-of-the-art approaches. The direct measurement is implemented via an optical detection of the tool cutting edge by means of a machine-integrated camera system.
In addition to the question of how high-quality images of the tool cutting edge can be automatically acquired in machine environments, our research focus on the reliable extraction of the wear zone from the images. In this context, we investigate and evaluate different image processing methods ranging from classical filtering and segmentation methods to new approaches of deep learning.