Visualization Quality and Mesh Settings

Learn how to use the Visualization Quality and Meshing Settings to manage compute times, the visual appearance, and the geometric accuracy of implicit and mesh geometry when working with Inspire Implicit Modeling. This includes working in Implicit Modeling, converting other geometry formats in Implicit Modeling, and meshing implicit geometry for export or other downstream processes.

Introduction to Visualization Quality

In Implicit Modeling, most of the geometry is created from implicit functions, which, theoretically, are analytical and therefore perfectly describe the geometry. However, this "perfect" geometry has to be rendered and also combined with other geometry that may not have a perfectly accurate definition, such as a triangle mesh (e.g., .stl). For these reasons, all implicit geometry is sampled on a 3D grid of points, which is analogous to the pixels in a digital image. In 3D, we call these pixels "voxels." In much the same way as we talk about the resolution of digital images, we have a similar notion of resolution in Implicit Modeling. This resolution is controlled by the Visualization Quality dropdown that you can see in the top-left corner of the images below. The first image shows a lattice structure, computed and visualized using the Very Low setting. It is clear that sharp edges have become rounded, and some of the diameters of the struts differ from the specified value.

If you look at the next image, the sharp edges are much sharper, although not truly perfect, and the diameter of the struts is very close to the specified value. You should use the Visualization settings to balance the trade-off between compute times (which depend on your hardware, e.g., GPU memory) and the desired Visualization Quality and geometry accuracy.

Advice for Managing the Visualization Quality

A good habit to form is to change the Visualization Quality to suit what you are doing at any given time. For example, computationally intensive operations, such as using Convert in Implicit Modeling, might be best attempted at a lower Visualization Quality first, then increasing the quality when you are satisfied that the operation is running correctly. Conversely, when creating models for export or render, it is recommended to try and use the highest feasible Visualization Quality for your computer.

Another consideration is that each body in an implicit part has its own voxel grid. In the image below, we have a cylinder in the foreground and a cube, positioned at a distance, in the background. You should note that both geometries are captured accurately, with sharp features.

If we use a Boolean Combine operation to join these two objects, they will be treated as a single entity. This means they will each be captured in the same voxel grid. Because many of the voxels are now being used to model the empty space between the two objects, there are fewer voxels dedicated to modeling each shape. In the image below, you can see the clear reduction in geometric accuracy, where edges that were previously sharp are now rounded. This is despite the fact that the Visualization Quality is the same in both images. You should keep this in mind when combining objects that are physically separated by a large distance.

The final piece of advice is to consider the alignment of the geometry within the voxel grid. In every case, the voxel grid is aligned to the global x-, y-, and z-directions. We call this an axis-aligned grid. Also, the voxels in the grid are always cubic elements. Features with curvature will be captured more or less accurately depending on how they cut through the voxels. As an illustrative thought exercise, imagine trying to represent a 2D straight line using a digital image with pixels. If the line is horizontal or vertical within the image, the representation will be sharp and accurate. However, if the line is diagonal in the image, you can expect to see a staircase-like effect along the lines as the it passes through pixels in different rows and columns.

The same is true in 3D with voxels. To illustrate this effect, we have created a very long and thin cylinder in Implicit Modeling. The image below shows such a cylinder aligned to the global z-direction. The first thing to note is that the sharp circular edge at the top of the cylinder (similarly, the bottom) has been rounded. This is a result of how the voxel size is calculated. Firstly, the axis-aligned bounding box for the object is computed. The longest dimension, in this case the z-direction, is used to set the voxel size. For example, the Very High setting in the Visualization Quality dropdown divides the longest dimension of the bounding box into 512 equal sections, thereby setting the voxel size. This voxel size is used in the other two directions, which will clearly result in fewer voxels across the circular face. As such, we have many more voxels capturing details along the length of the cylinder as compared to its cross-section. There is therefore a slight loss of accuracy in the circular profile.

There is a further loss of accuracy if we align the cylinder along the body diagonal of the axis-aligned bounding box (from, say, the bottom-front-left to the top-rear-right corner). In the image below, we can see further loss of accuracy as a result of the aforementioned staircase effect where the features of the cylinder do not align favorably with the discrete voxel grid. Although this is an inherent property of modeling this way, being aware of this effect can help you to effectively minimize its impact.

Converting Implicit Models to Triangle Meshes

There are many reasons for converting implicit geometry to a triangle mesh. Some common examples are:

The software responsible for build preparation on your additive manufacturing machine (3D printer) accepts or prefers triangle mesh formats.
You have critical geometrical features that need to be preserved, such as sharp edges, but you also want to incorporate implicit geometry in the design, so you convert everything into a triangle mesh format prior to the Combine operation.
You are working on a model that will subsequently be sent to analysis (e.g., a structural simulation), and the simulation pipeline cannot interpret Implicit Modeling geometry.
You want to use geometry editing tools that do not accept implicit models as an input (e.g., Sculpt in PolyMesh).
You wish to send your design to a colleague or customer who does not have access to Inspire Implicit Modeling.

The following steps walk you through the process of converting an implicit part into a triangle mesh, making the most of the Implicit Meshing options to suit your application.

Complete your edits of the implicit part by clicking the green check mark.
Ensure the Property Editor window is visible by selecting the option under the View tab.
Select the implicit part that you wish to convert to a triangle mesh, either by selecting it in the Model Browser or by clicking the model in the scene.
Scroll to the bottom of the properties in the Property Editor window to find the Implicit Meshing properties.
Select the destination for your mesh. For simulation, select the Analysis option. For export for printing or saving the file, select Print.
Choose your quality settings from the two options: Standard or Custom.
Right-click the model you wish to convert (either in the Model Browser or directly on the model in the scene), and then select Convert to Triangle Mesh.

Choose a conversion setting in the Convert to Triangle Mesh dialog.


Option	Description
Create sharp edges and specify mesh size	Sharp edges are created at the boundaries based on the geometry of the original implicit part. Type a value in the Average mesh size box to set the size of the triangle mesh. A higher average mesh size generates fewer triangles and vice versa.
Specify mesh size	Type a value in the Average mesh size box to set the size of the triangle mesh. The original surface geometry is not preserved. A higher average mesh size generates fewer triangles and vice versa.
Use implicit visualization quality	Triangle mesh is created as visualized. Higher implicit visualization quality generates more triangles and vice versa.

Click Convert.

Guidance for Choosing Implicit Meshing Properties

Firstly, you should note that the quality of the mesh that is produced in the conversion relates directly to the Visualization Quality setting that was used during Implicit Modeling. If the Very High setting was used, the triangle mesh will be very fine, as shown in the image below. Using progressively lower Visualization Quality settings will produce increasingly coarse meshes.

If the destination for your mesh is not analysis, you might wish to reduce the triangle count, where possible. To do this, select the Custom quality setting and make sure that the decimation checkbox is selected. During the conversion, the triangle mesh will merge small triangles into much larger triangles wherever possible. This is usually seen in large, flat expanses. The result using the default Decimation Reduction and Decimation Error settings is shown below.

If the decimation using the default settings is too aggressive, you can reduce the Decimation Reduction property. The Decimation Reduction property should be a number between zero and one. Setting the reduction to, say, 0.9 is the same as asking for a 90% reduction in the total triangle count in the mesh (compared to the original mesh). Conversely, 0.4 is the same as asking for a 40% reduction. The image above uses 0.75 and the image below uses 0.4. You can see that there are more triangles in the image below, and the mesh is a better approximation of the original fine mesh.

The Decimation Error property is a value specified as a percentage of the bounding box size. This specifies the permissible deviation between the original fine mesh and the newly decimated mesh. Setting this property to smaller values generate meshes that reproduce the geometry in the original fine mesh more accurately. As a limit case, this value can be set to zero, which typically means that triangles are only merged if the region is planar and there is no loss of accuracy. The impact of doing this is shown below.

Meshes for Analysis

If the destination for your mesh is analysis, such as a structural simulation, neither the original fine mesh nor the decimated mesh will be appropriate. The fine mesh is almost certainly too fine and will unnecessarily slow the analysis pipeline down. Conversely, decimated meshes contain triangles that are generally too large, covering large expanses with a single triangle.

A better mesh input for analysis is one that contains triangles that are close to equilateral. There should be plenty of smaller triangles across large planar expanses, rather than just a few large ones. Also, triangles should be smaller in areas of high curvature and somewhat larger in flat regions. Also, wherever possible, sharp features such as corners and edges should remain sharp in the mesh. This helps with feature identification for the application of boundary conditions during the simulation phase. There are no rigid guidelines for what makes a mesh ideal for analysis and user discretion is encouraged. However, an illustrative example is shown below, clearly highlighting the triangle size in planar regions and also the smaller triangles near sharp edges.