Downsampling reduces the structure size by combining multiple voxels. By default, the Downsampling Factor is set to two – this means that 2x2x2 voxels are combined to one voxel. The material model used for each voxel in the downsampled structure is a combination of the material model (e.g., a Composite Voxel) of the original voxels. Since the simulated structure size is much smaller than the original structure, Downsampling allows to speed up the simulation and to quickly obtain results even on computers with little memory. This option should not be used if highest accuracy is required.

In most cases, a Downsampling Factor of 2 is the best choice, but it can also be set to 4 if needed. Then, the computation will be faster, but the results could be more inaccurate.

To be able to use downsampling, the features (e.g., fibers or grains) in each material ID must have a diameter of at least the Downsampling factor in voxels (e.g., 2 voxels for a Downsampling Factor of 2). Otherwise, the computation of the material properties might not work properly. Since GeoDict 2021, an error message is shown if all features in the structure are smaller than this diameter, and the simulation cannot be started.

Composite Voxels

Two different methods to compute the Composite Voxels: Laminate Theory (the default), and Arithmetic Mean. When using Laminate Theory, the composite material properties are computed based on the laminate theory of composite materials. With this option, also direction dependent properties are represented correctly. With Arithmetic Mean, the material properties of the composite voxel are the average of the material properties of the original voxels. This method is less accurate, but the simulation might be more stable, especially if the structure contains many pores (as e.g., in foams). Nevertheless, it is recommended to keep the default Laminate Theory in most use cases for the higher accuracy.