The rapid development of AI-based methods has significantly increased the importance of data usage and data interpretation in all phases of development processes. Modern methods of data science enable development engineers to discover hidden insights in existing datasets or to create new datasets suitable for machine learning. This talk focuses on two applications for filter media and element development where AI-based methods provide additional insights and improve efficiency.
The first use case highlights significant enhancements for the segmentation and microstructure characterization of synthetic filter media. Traditional methods, such as manual and OTSU-based threshold techniques, face challenges e.g. with low-density fiber materials and artifacts in high-resolution µCT scans. These challenges can be overcome by using AI-based methods for the segmentation. We will apply this method on different microstructures and compare the results with OTSU threshold-based segmentation with subsequent morphological operations. Comparison of virtual microstructural properties of both segmentation methods with material testing results serves to validate the new method and make its advantages comprehensible.
The second use shows how to apply machine-learning methods on hybrid data sets for optimized filter element layouts. One goal is to enable development engineers to find optimal filter designs within seconds instead of hours. Therefore, the optimal filter element design needs to consider the design space and customer specifications, amongst others. Looking at the workflow, we will also show how to seamlessly integrate this novel design approach into the whole development process.