InstantCAD Promises Faster Iterative Design

The design process for any product is necessarily an iterative one. Often, a prototype is modelled or built, and changes are made to overcome problems and improve the design. This can be a tedious process, and it’s one that MIT’s CSAIL has sought to speed up with InstantCAD.

The basic idea is integrating analysis tools as a plugin within already existing CAD software. A design can be created, and then parametrically modified, while the analysis updates on screen in a near-live fashion. Imagine modelling a spanner, and then dragging sliders to change things like length and width while watching the stress concentrations change in real time. The tool appears to primarily be using some sort of finite element analysis, though the paper also shows examples of analyzing fluid flows as well.

The software is impressive, however there are caveats. Like any computer analysis, serious verification work must be undertaken to ensure its validity. We suspect that there may be issues with more complex geometries that lead to inaccurate simulation. It’s not the sort of tool you’d use for anything that puts life and limb at risk, but we can see it having great uses for designing basic objects when you want to quickly gain an idea of what sort of effect certain parameter changes will have.

The other main disappointment is that while this tool looks great, it doesn’t appear to be publicly available in any form. Whether this is due to universities and complicated IP requirements or the potential for future commercialization is anyone’s guess. Regardless, you can read the conference paper here or check out the video below. Or you could read up on the applications of finite element analysis to 3D printer slicers, too.

20 thoughts on “InstantCAD Promises Faster Iterative Design

      1. What’s the point of patenting it, but not having any plans to make it available or using it yourself? The video says ‘The goal is to help designers create better products in less time’, but by locking something out of reach behind a patent, the exact opposite is achieved.

        Don’t get me wrong, I understand the real world is sometimes about protecting your assets and less about ideals.

        1. It’s to patent toll Autodesk. Hopefully they don’t bite and in a few years after these guys figure out how hard it is to market something it’ll go open source. But in the meantime:


        2. What’s the point of patenting it? The United States moved to a first to file. You can invent something, somebody can go pay the patent fees and argue it is now their patent. Then sue you, the original inventor.

          1. The patent system is, as it stands, thoroughly broken. In its current state, it breeds huge dominant parties that consolidate their position by accumulating huge patent portfolios. This, in turn, makes the entry fee for new companies almost insurmountable, them being tripped up by dozens of important and trivial patents, no deep pockets to defend themselves and not enough resources to file many patents themselves.

            Currently, the risk of being mauled in court seems to be a much bigger for technology start-ups than the traditional modes of failure. If you attract the attention of one of the giants, that’s it. It can even be used as a way of cheaply buying smaller companies.

        1. You might want to research that further if you are talking about USA patent laws. You can patent something and literally just prevent others from using your patent (within the jurisdiction of the patent in question and assuming the patent is held to be valid, etc).

  1. “The tool appears to primarily be using some sort of finite element analysis, though the paper also shows examples of analyzing fluid flows as well.”

    I imagine the race towards more powerful GPUs have helped in this regard.

    1. Probably, but not necessarily. The article suggests the actual analysis is done by a server farm, and the UI sliders either select which data set to show, or run an extremely simplified model locally.

      (PS. If any of these statements are NOT what they’re doing, well, they’re my IP and I’m gonna patent it!! Haha!)

  2. If the stiffness matrix is parametric they can do all the solving parametrically as well, so that scaling changes to the part mesh become scaling changes to the results. None of the examples show how long it takes to generate the initial matrix and it doesn’t show how long it took to solve it. No changes were made that affect the topographic relationships.

    It’s interesting, but generalized stress fields are not the biggest driver of component design.

  3. OpenSCAD flame:

    They list OpenSCAD as parametric CAD where parametric is also used as a synonymous of constraint-based CAD, which is not what OpenSCAD does. Quite frankly, I spent 3 hours to model a piece of furniture in OpenSCAD once when I already knew quite well how to use the software, the same task would take 20 minutes in Solvespace, Solidworks or Inventor. OpenSCAD is interesting but it is based on an outdated concept of CAD, more effort from the open source community should be invested in the development of Solvespace and friends.

    Ok all this was totally offtopic. Now, the paper is interesting and if I understood well they propose a method that allows to vary certain geometry parameters and observe the effect on the results without having to remesh or recalculate everything. That is an interesting concept indeed but I’d be surprised if they were the first to come up with this idea.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.