Mesh Modeling, Part 2: Siemens Convergent Modeling for Meshes and NURBS-based Geometry

See also Part 1 of the Mesh Modeling series: 3D Systems, Best Known for 3D Printers, Offers Robust Tools for Meshes.

Developers of Siemens NX are working to integrate mesh data for everyday CAD users. Siemens, for both NX and Solid Edge, calls its marriage of meshes—referred to as facet, or polygonal models—and NURBS-based models—known as B-rep models—convergent modeling. Siemens takes “convergent” to mean multiple types of models are coming together in a single application interface. “Hybrid” is the other term applied to mesh and NURBS geometry in the same model, although hybrid has also been applied to many combinations, such as wireframe and surfaces, surfaces and solids, direct and history-based, and now mesh and NURBS data.

The future that vendors hope to address with hybrid modeling is already here. The addition of non-NURBS data to a NURBS-only CAD world is more and more common. Tools to work with mesh data are being increasingly incorporated into engineering software, from the high-end PLM suites to the mid-range and lower levels of CAD.

This image demonstrates the difference between facet data, a set of points, and NURBS or B-rep, driven by math. (Image courtesy of Siemens.)

What kinds of items do we find in a mesh format that we might want to use in a NURBS/B-rep CAD modeler?

Mesh models use analytical surfaces for mating
FEA uses mesh for analysis
3D scans yield mesh (point cloud) results
Medical MRI and CT scans
Textures can be applied via mesh
Generative design produces mesh data
3D prints can be made from mesh data
Display data uses shaders working with tessellated point clouds
NX CAM can machine from double accuracy mesh data
AEC applications need scanned data of terrain for civil engineering and architectural planning

Convergent modeling was first announced by Siemens in 2016. It is meant to make working with different data types seamless and transparent to the CAD user.

A great example of this is shown in the part below. The part was created by generative design, one of the technologies that convergent modeling was meant to originally support. But the mesh model generated by the generative process was just a stand-alone part. You couldn’t use it anywhere because none of the holes were really holes, and none of the planes were perfectly flat. You couldn’t mate this into an assembly. The model needed to contain two different kinds of data: point-based data for shape and analytical surfaces for function and mating.

The generative geometry is mesh, but the mating surfaces are B-rep. Both types are needed. (Image courtesy of Siemens.)

To add analytical geometry to a mesh model is probably one of the most basic requirements of convergent modeling. Very few products are actually stand-alone parts. Almost everything that is manufactured is put together with something else in a fashion that requires functional surfaces, such as planes and cylinders. These analytical surfaces might represent secondary processes, including machining, grinding, drilling, broaching, reaming or tapping.

So, we have to move on to assembling parts built using non-traditional means with traditional parts. At this stage, we can use mesh in the form of finite element analysis.

Simcenter 3D Convergent FE Analysis. (Image courtesy of Siemens.)

Integrated downstream Siemens applications, such as Simcenter for simulation and analysis, are also convergent enabled and can take in parts from various sources.

NX also allows data types that are as much properties or instructions as geometry. For example, a turbine blade, shown below, that might have been machined previously might now be printed in a metal material. The printing process allows interior spaces to be filled with a supportive lattice instead of solid material. The lattice is not explicit in the modeled geometry and is generated by the printer.

NX CAD variable rod diameter lattice on a turbine blade. (Image courtesy of Siemens.)

Another type of modeling that has plagued traditional NURBS-based CAD for years is applying a texture. Textures are sometimes applied chemically, photochemically or through some process to erode a surface of a mold. With mesh modeling, you can actually model a texture onto a surface. You can also take this a step further by using an image to drive the texture. This is typical product-design work that most CAD systems either struggle with or hand off to downstream applications and processes.

NX CAD Texture Modeling. (Image courtesy of Siemens.)

CAM makes use of convergent modeling as well, which enables it to machine from faceted (STL type) data. This can save a lot of conversion or re-surfacing work that you might otherwise have to do when machining from faceted data, which is happening more and more as this kind of information becomes more prevalent. With the drastic drop in the cost of 3D scanners and 3D printers, we are seeing more facet data making its way into the product development process.

Convergent models in NX CAM. (Image courtesy of Siemens.)

Alongside working with faceted data comes other types of functions, such as smoothing, hole repair and adding thickness to faceted sheets. STL data, whether from scans or other sources, tends to leave holes or create errant island bodies that require powerful editing functions to fix.

NX CAD Convergent mesh cleanup can smoothly fill holes in scanned data. (Image courtesy of Siemens.)

Architects and civil engineers require information about the existing terrain. These days, that information comes from scans from drones, aircraft and satellites, and it comes in the form of meshes. Engineers build roads, railroad grades, tunnels, bridges, piping systems, reservoirs and drainage fields around this information.

NX AEC market data. (Image courtesy of Siemens.)

A form of polygonal data that is used constantly in CAD, which we have possibly come to take for granted, is the tessellated display that gives us Gouraud shading. It is familiar to CAD users from the rectangular or triangular artifacts often seen on shaded curved surfaces. Notice in the screen capture below that the edges in the shape on the left are well rounded while the edges in the shape on the right are segmented, or tessellated. Both images are taken of the same model. The underlying geometry is smooth, but the display is created from a collection of flat triangles that approximate the curvature of the real model. A courser mesh gives a faster but less accurate display. The same concept is at work in all of these mesh/facet/polygonal/point cloud methods discussed in this article and has been used extensively in all visual computing for decades.

Tessellated display shows the underlying polygonal nature of the display data, even if the actual geometry is mathematically smooth.

Finally, the big prize at the bottom of the box is that from a mesh mode you can produce a subdivision model. If you have spent many hours deep inside a history-based surface model, this is something you’ll be able to appreciate. You can deal with the entire shape as a shape and not as a series of instructions for making a shape. You get instant visual feedback and no error messages.

You no longer have to be envious of 3DS max or Rhino users who produce organic shapes in minutes compared to the days or weeks it takes you with your NURBS CAD.

Summary

CAD using non-NURBS data is no pie-in-the-sky dream. It’s happening, and it’s fueled by 3D printing and 3D scanning, as well as developments in medical applications. There is so much polygonal data out there that this data type cannot be ignored. Techniques needed to edit and manipulate mesh data have existed for a long time and can be directly applied to organic shape design. Working in sketch and feature driven history-based NURBS models to produce organic shapes is error-prone and painstakingly slow. You can achieve better organic shapes in less time using the mesh modeling applications. If you still want NURBS and meshes in the same app, there’s Siemens convergent modeling.