Scan and Solve offers meshing-less FEA

19 March 2012: Recently I had a chance to sit in on a demo of Scan&Solve™, software promising to (virtually) automatically solve parts for linear stress FEA analysis without any concern about meshing the part. Used in conjunction with solid models of parts developed with Rhino, the demo did just that. To verify the accuracy of the results, the demoer adjusted something called the resolution of the “geometry scan” of the part. Adjusting the resolution showed that the accuracy was converging. Wow! I thought. Time to find out more about how this worked, how extensible it was, how it differed from traditional FEA, and its cost. I went to the company website and soon located the founder of Intact Solutions LLC, the company that authored the software – Vadim Shapiro, Professor of Mechanical Engineering and Computer Sciences at the University of Wisconsin-Madison.

Right away, I figured uh oh, an academic. They are not always known for providing crisp answers. Nevertheless I requested and was granted an interview with Dr. Shapiro. He turned out to be very open and enthusiastic about the product.

Vadim Shapiro, Intact Solutions founder

Here is what Scan&Solve does differently than traditional FEA meshers and related solvers:

  • Scan&Solve is not a replacement for FEA; it is an extension of FEA, which aims specifically to solve the problem of CAD/CAE interoperability. Any reasonable geometric kernel and any reasonable FEA package can be interfaced with great benefits. The goals of the product are simplicity, universality, and complete automation.
  • The current version analyzes parts only, not assemblies.
  • Instead of meshing, the software assigns an analysis space (a grid) surrounding the part to be worked on, as shown below:

How Scan&Solve works

  • Scan&Solve needs to interface with the CAD system to supply coordinates of the model to Scan&Solve for each point on the grid. Given this interface between the CAD system and S&S, there is no need for a mesh to be created. Instead the software works with the precise model geometry. Scan&Solve directly modifies the basis functions, sometimes called “shape functions” — functions that approximate the solution of the problem. In the current implementation, these basis functions are associated not with vertices of the mesh, but with cells in the mesh (of the space, not of geometry). “Modify functions” means that they are modified to satisfy the applied restraints everywhere — not just at vertices. Scan&Solve™ can be applied to any geometric model and used within any geometric modeling system that supports two fundamental queries: point membership testing and distance to boundary computation.
  • No simplification or de-featuring of the model is needed.
  • Increasing the resolution of the grid can test convergence of the results. If a higher resolution produces large changes in the results, keep increasing the resolution. Shapiro noted, “The issue is essentially the same as with standard FEA. One can estimate the error and refine the mesh (or increase density in our case), but it is more or less the same for all techniques. We do not do anything automatically right now. We advise the users to run at different resolutions (which requires NO WORK from the user) and compare the results. If results are significantly different, increase the resolution. In principle, this can and will be automated in the future.”
  • Can work directly with polygonal models. Scan&Solve performs all analysis related computations on the native geometry (whether polygonal, NURBS, or other form of geometry). Shapiro stated that “This eliminates the need for preprocessing: no healing, smoothing, de-featuring, or meshing is needed. This drastically reduces preparation/set up time.” However, the commercial product in Rhino works only with NURBS solids.
  • It always produces results. Shapiro stated “The solution procedure is deterministic, does not use heuristics, and always produces a result. (In other words, failure means a bug in the code: not inability to handle some geometry.) The advantages of S&S are full automation, complete integration and interoperability. Use it at any stage of the design process: from concept creation to detailed geometry.”
  • Prices are very reasonable. Scan&Solve for Rhino commercial licenses are $695 for a node locked version and $1295 for a floating license. Academic, trial and rental licenses are also available. Scan&Solve for Rhino also requires a Rhino license.
  • Interfaces are available currently for a limited number of CAD systems. Scan&Solve can be applied to any geometric model and used within any geometric modeling system that supports two fundamental queries: point membership testing and distance to boundary computation.

References: http://www.intact-solutions.com/

http://www.scan-and-solve.com/

Disclosure: No remuneration of any kind was paid for this article.

Conclusion: Both CAD and FEA vendors should check out the possibility of offering this technology as an option for users. With trial copies available from both Rhino and Intact Solutions, users wanting to extend FEA analysis beyond the traditional analysis experts should consider the benefits and urge their CAD partners to investigate this alternative.

6 responses

  1. Hi ray,
    Very interesting. Do you know what the actual solver is doing/using? i.e., can it use an existing “off the shelf” solver like Nastran, Ansys, Abaqus? Or does it do it’s own solving of the elasticity equations using an implementation of “Isogeometric Analysis”?

    • Dennis, Scan&Solve for Rhino uses its own solver that uses
      uniform grid of B-splines as “finite elements”. To combine Scan&Solve with other types of elements/analysis, we need to modify solver’s integration and post-processing routines. We would need to change how numerical integration is performed over individual elements in the code,and modify the way the results are post-processed. A typical FEA assumes that geometry is represented by a conforming mesh, and all their computations (integration and visualization in particular) use mesh as geometry. We would modify these routine to refer to the native geometric model instead and adapt their integration and visualization routines to work on the original geometry.

      There is nothing special about iso-geometric analysis — just a different set of basis functions, and it can also be used with Scan&Solve. In fact, it would help quite a bit with imposing boundary conditions (restraints). But the main claim of iso-geometric analysis is it ability to conform to the original geometry — which Scan&Solve does not really require.

      Hope this answers your questions.
      -Vadim

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

Join 784 other followers

%d bloggers like this: