Digital Prediction of Nonwoven Media Performance Using GeoDict

This work presents a digital twin workflow for predicting the performance of nonwoven filter media, developed at MAHLE, using GeoDict simulation software. Starting from high-resolution Nano CT scan data of real nonwoven structures, a systematic image processing pipeline is applied to import raw scan data and perform segmentation, isolating fiber structures from background noise to obtain a clean, representative 3D dataset.

The segmented structure is reconstructed into a fully parametrized 3D digital twin of the nonwoven media, capturing key morphological features such as fiber diameter distribution, porosity, and layer architecture. Building on this digital twin, an automated parameter variation study is performed using a Python-based scripting framework, enabling efficient exploration of structural variants without requiring additional physical prototypes or CT scans.

For each generated variant, key filtration performance indicators are simulated: pressure drop (ΔP), filtration efficiency (η), and dust holding capacity (DHC). Simulation results are systematically compared against experimental measurements on physical filter samples to assess model accuracy and validate the digital prediction approach.

The proposed methodology demonstrates how combining Nano CT imaging, digital twin reconstruction, and automated GeoDict-based parametric simulation can accelerate nonwoven filter media development, reduce reliance on physical prototyping, and provide deeper insight into the structure-performance relationship governing pressure drop, efficiency, and dust holding capacity.