Oct 142011

A couple of days ago, I saw a conversation thread on Twitter about geometric modeling kernels. It wasn’t much of a thread—just a few comments back and forth to the effect that modeling kernels are like car engines, and that if you can’t tell the difference without looking under the hood, it doesn’t matter which one you have.

CAD users don’t think too much about what kernel their software uses. I suppose most of them can’t tell anyway. But that doesn’t mean kernels don’t matter.

There are all kinds of potential problems that can crop up with modeling kernels. A while back, I published a couple of articles about interoperability problems (which are inherently related to kernels), one from an academic perspective, and one from the perspective of a kernel guru.

About a month ago, I wrote a series of articles on configuration modeling, pointing out that no modern CAD systems can really do this. A couple of days ago, I made an off-hand comment in an article that a picture I showed (of a sphere) was really a cube that had its edges blended (e.g., start with a 2” cube, and fillet all the edges at 1”.) I learned that trick 15 years ago with SolidWorks. Several readers wrote or commented that they were unable to do it with their modern CAD systems.

The most common sign of a kernel-based problem happens when a CAD user tries to create a geometric feature, and the result is a failure.

Think about that for a moment.  You’re working on a CAD system, trying to create a feature, and the system does something unexpected.  That’s a big red flag saying the modeling kernel can’t handle what you’re asking it to do.

As an aside, I think it’s mighty interesting that one of the signs of an expert CAD user is their ability to work around limitations in the kernels of their CAD programs that would otherwise create modeling failures.

So, yes, geometric modeling kernels matter. Even to CAD users who don’t realize it.

Yet, there is no best alternative when it comes to geometric modeling kernels. ACIS, Parasolid, CGM, Granite and the proprietary kernels out there each have their own kinks. None is so much better than its competitors that I want to jump up and down and say “everybody look at this!”

The spark that set off the Twitter thread that inspired this article was an announcement, from Siemens PLM, of a webinar, to be held on November 8. Here’s the description from the Siemens website:

At the core of your mechanical CAD software is the modeling kernel, an often overlooked tool. The kernel is key to your ability to compute 3D shapes and models and output 2D drawings from 3D geometry. In this webcast, learn the basics about kernels and what impacts a change in this core code can have on your company’s existing and future design data. Dan Staples, development director for Solid Edge at Siemens PLM Software, is joined by medical device designer Billy Oliver from Helena Laboratories to explore the issues facing hundreds of thousands designers and millions of CAD files.

    • The math inside your feature tree
    • Real-world lessons learned in changing kernels
    • Modeling loss, data protection, and reuse risks
    • Impact on hundreds of thousands designers and millions of CAD files
    • Case study: Helena Laboratories ensures data protection

You can register for the webinar here.

While I expect the webinar will be, by its nature, slanted towards Siemens PLM and its Parasolid kernel, I suspect that quite a lot of what will be discussed will be interesting to people who have no intention of changing their CAD tools. I’m planning on listening in.

I doubt that most CAD users will ever spend much energy thinking about their CAD programs’ modeling kernels. But CAD users should spend some energy thinking about broader issues, such as usability and interoperability, which are affected by modeling kernels.

  • Ken

    I think the only time a CAD user will care about which kernel their CAD system uses is if/when the CAD vendor changes it and the result to the user is corrupt models afterwards. If it is transparent and all the old processes still work, then it’s not really an issue.

  • http://jonbanquer.blogspot.com/ Jon Banquer
  • 1SteveJepson1

    See Nemetschek Vectorworks and Alplan websites for examples of how their programs are using the Siemens PLM kernel.