The first interview on the revamped Dezignstuff blog is Matt Sederberg, one of the guys responsible for creating T-Splines. T-Splines, you may remember, was the company that started to bridge the gap between mesh and NURBS type models.
Matt now has a new venture, disclosing new technology, and, well, he tells the story better, so let me just allow him to do it. Green text is Sederberg, black is Lombard.
1) Some of our readers will recognize you as part of the team behind
T-Splines. Can you remind us what T-Splines is and where it is today?
Sure, Matt. T-Splines is a CAD surface, compatible with NURBS and subds, invented back in 2003 by my dad, Dr. Tom Sederberg. Our company, T-Splines, Inc. commercialized this technology as plugins for Rhino and SolidWorks. Users like it because it lets them easily model even complex organic shapes as a single watertight model, which you can push and pull on to refine the shape. This is a similar workflow to what you’d find in animation products like Maya or modo, but T-Splines has unique capabilities that make it ideal for integration with CAD. T-Splines has technical advantages over subds in that you can refine parts of the model to fit to a tolerance; T-Splines are also forward and backward compatible with NURBS. These qualities made for a very nice integration into Rhino and SolidWorks; in practice most users create mixed models, part T-splines, part NURBS, with smooth blends in between. The Rhino plugin in particular was used heavily in industries which created aesthetic surfaces for manufacturing, such as industrial design, architecture, jewelry, and toys.
In late 2011, Autodesk acquired T-Splines and made it part of Fusion 360 (T-Splines remains a key differentiator of Fusion from other cloud-based CAD software, which don’t have similar freeform modeling capabilities). Later, T-Splines was added to other Autodesk products such as Inventor, Alias SpeedForm, and Dynamo (for generative modeling).
One unanticipated application of T-Splines has been its use in computer-aided engineering (CAE). An emerging field in CAE called isogeometric analysis (IGA) started to emerge around 2005, pioneered by one of my Coreform co-founders, Dr. Thomas Hughes. IGA uses CAD data directly in simulation instead of first requiring a conversion to a mesh. IGA began with a focus on untrimmed NURBS, which are incapable of representing complex CAD objects used in industry. Another Coreform co-founder, Dr. Michael Scott, discovered that T-Splines have the ability to also be used as the basis of IGA, which gave the potential to transition IGA from a research idea to something usable in industry. Many research papers have been written about using T-Splines directly in simulation without meshing them.
However, T-Splines were never invented with analysis concerns in mind, and some limitations have been identified over the years. A number of ideas have been proposed to address these. Recently, another Coreform co-founder, Dr. Derek Thomas, invented a new surface type called U-splines (unstructured splines), which we feel are a significant breakthrough. In many ways, U-splines are similar to T-Splines, but they were invented from the ground up to be suitable for analysis. U-splines are new – we are publicly disclosing them this summer for the first time at a series of conferences – but they have very promising properties. They have a straightforward extension to volumes, which is crucial for CAE as well as intriguing as a smooth, volumetric CAD basis for additive manufacturing. They are mathematically backwards compatible with NURBS, T-splines, and Bézier patches, and can be comprised of patches of arbitrary degree, which offers interesting avenues to pursue for class-A or near-class A surfacing.
2) So we’re really talking about smoother, more accurate geometry for simulation as well as better interoperability between CAD and FEA?
Yes, that’s the vision. When you take a step back, it really doesn’t make sense that while CAD data is the trusted source of truth in the manufacturing process, it is replaced with a less accurate, faceted approximation when sent to CAE. We believe that in the future CAE will be run directly on CAD. It will be faster and more accurate and tedious model prep time will be a thing of the past. (By the way, we’ve compiled a list of some of the key academic benchmarks that IGA has been subjected to over the past decade on our website at www.coreform.com/iga).
3) When in the CAD-FEA process will this technology be implemented? Will there be some transfer after the CAD is done from NURBS to a U-Spline model, or is this just native CAD data so that you can FEA and CAD without translation? Do you envision this as being a separate tool, part of CAD, part of FEA…
That’s the big question, is how to fit this into production workflows. There have been many new simulation methods proposed over the years; most haven’t succeeded because either they don’t fit well into production workflows, or because they are less accurate than traditional CAE. The first product we are commercializing is a preprocessor, Coreform Prep, that will take in an existing CAE mesh, smooth it to a U-spline, then export the smooth basis for use in a simulation code that can read in smooth geometry, like LS-DYNA. We’re also working on developing IGA-native tools that leverage properties of the CAD-CAE integration to improve the quality of analysis results.
The reason we’re focusing our development here first is because that will give customers the ability to start experiencing and benchmarking results of simulations on a smooth surface instead of a mesh – without needing to change other parts of their existing pipeline.
4) How far along are you in your development process? When will you have software people can test on their own data?
Commercializing new CAD technology is a significant endeavor. Though my co-founders have been developing this tech for several years, we’re still at the beginning. We’re funded by a number of industrial and government partners and we’re currently building out our capabilities in response to problems they have given us that aren’t adequately addressed with traditional CAE software. We haven’t yet announced when our software will be ready for public release, but welcome additional partners who would like us to consider their requirements today.
5) Can you tell us how your product aims to fit in the automotive design workflow?
Matt, there has been a strong trend in automotive design for several years to move away from traditional NURBS modeling and towards more flexible, form-finding technologies like subds, T-splines, and polygons. This cuts a significant amount of the modeling time during the initial form exploration stage. However, as the shape is refined to class A, it always requires being completely and tediously remodeled by Bezier patches, because the form-finding technologies lack the precision and smoothness to achieve class A quality for automotive manufacturing.
At Coreform, our vision for the future is that the same geometry (potentially U-splines) would be used for both CAD and CAE. For that to happen, there must be a compelling reason for both industries to change, beyond the benefit of interoperability. We believe the reason for CAE users to switch to IGA will be time savings and higher accuracy.
It would be most compelling for CAD users to also eventually use analysis-suitable geometry if it also improves their own CAD work. Changing CAD workflows is not a trivial process and requires significant benefits to be considered. One possible benefit that U-splines might provide to CAD would be high quality surfacing – potentially even sufficient to provide class-A or near-class-A quality in automotive design, along with tighter integration with the emerging form-finding technologies used early in the design process. Another potential driver for adoption could be the possibility of U-splines to represent interior elements of a model for additive manufacturing. It will be exciting to see which industry emerges as the first to embrace a fully-integrated, isogeometric design-through-analysis workflow!