The example shown here, is a part of the imported µCT image of a carbon paper gas diffusion layer (GDL) used for a proton-exchange membrane (PEM) fuel cell. The fibers and binder contained in it are undistinguishable by their gray value in the FIB/SEM image.

The input data used, as well as a detailed description of the necessary processing steps, can be found in the tutorial Digital analysis of fibers and binder content of a carbon paper GDL.

Some exemplary slices of the structure in Z and X-direction, imported into GeoDict, are shown here.

To start, select Analyze → FiberFind from the menu bar and, in the FiberFind section, select Identify Binder (AI) from the pull-down menu. Click the Options’ Edit... button.

The Binder Identification dialog that opens includes the Binder Identification, the GPU Options and the Equations & References tabs. The GPU Options tab is similar to the GPU Options tab for Identify Fiber (AI).

Two choices are available as Identification Mode: Use Current Structure and Load Neural Network Output.

The default Use Current Structure applies the binder identification to the structure currently available in GeoDict memory and is used for most of the cases. If a binder identification step has already been done successfully and you want to re-run the postprocessing step with another threshold, Load Neural Network Output may be chosen.

Browse to a .npz file from the result folder of a previous run of FiberFind - Identify Binder (AI). Since already the output of the neural network is used in that case, no selection of Neural Network or Material to Analyze is possible and the loaded structure needs to be the same.

On the Binder Identification tab, select the neural networks available under Neural Network. GeoDict provides one neural network to identify binder. The application constraints, like the fiber cross-section and diameter of the structures with which the neural network was trained, are listed in the Description field.

Neural networks trained with GeoDict-AI can be used by copying the corresponding .gnn-file into the installation directory for Identify Binder (AI) (e.g. in C:\Program Files\Math2Market GmbH\GeoDict 2025\FiberFind\IdentifyBinder). Then, these neural networks appear as further options in the Neural Network field.

The Material to Analyze panel offers the choice of whether the binder identification should be applied to all solid and porous materials in the structure or only on a subset, defined either by choosing Material IDs or Materials.

The neural network provides probabilities about whether a solid voxel is binder or not. The Threshold parameter defines the limit for this labelling and can therefore be defined by the user between 0 and 1 (default 0.5).

The Binder Material specifies the material to label the binder. The material chosen should be different from the material of any solid or pore material present in the original structure. The ID assigned to the binder material is the first ID that is not present in the current structure.

Select one or several boxes on the Domain-Boundary Options panel if the underlying structure for the binder identification is periodic in one or several directions. However, periodicity is unusual for most 3D-scans and this setting is rarely used.