The significant increase in productivity and resource efficiency is made possible by automated type classification. In the conventional process, particles are automatically segmented and classified by particle type on the basis of a gray scale determination. However, this method often requires manual rechecking and manual reclassification by the operator. By combining this with a pre-trained machine learning based object classification model, the process now becomes much more efficient. In this process, the particle measurement results obtained in the classical method are combined and analyzed by using the pre-trained model to form a large number of uncorrelated decision trees. The result is a reproducible majority decision. If particles are detected that were previously misclassified as non-metallic, the machine learning-based object classification overwrites the results of the classical gray scale determination, resulting in a more selective classification of metallic particles. The ZEISS TCA module with machine learning extension was pre-trained using correctly classified particles and can be further trained individually by the user with little time expenditure.
Modern production processes and quality requirements are intensifying the demands on technical cleanliness – across all industries. With the TCA module, productivity does not fall by the wayside. Rather, the ZEISS solution enables industrial users to obtain quickly, easily, and efficiently certainty about the number, nature, and origin of process-critical particles. The ZEISS development thus has the potential, as the Fraunhofer Institute for Manufacturing Engineering and Automation IPA put it on the occasion of the REINER! 2023 award, “to drive the economy and cleanliness industry forward”.
For more information: www.zeiss.com/metrology