Presenting SesamX - Questions about open sourcing

[Ali baba Sx]

Hi everyone,

I have been working on a finite element project for a few months: SesamX. To introduce SesamX briefly:
[ul]
[li]It focuses on mechanical simulation.[/li]
[li]It provides the usual elements (truss, beam, shell and 3d solid).[/li]
[li]It can import mesh data from Abaqus input file or Salome .med file.[/li]
[li]Post-processing is done with Salome.[/li]
[li]The SesamX data storage relies on h5 file, which allows for model manipulation even outside SesamX.[/li]
[/ul]
Do not hesitate to check the website for more information.

Currently, I feel the software is mature to consider its business model more thoroughly. I am considering open sourcing SesamX. I am educating myself on the subject and I have a few questions that I would like to discuss with you:

[ul]
[li]According yo you, what is the biggest advantage of going for open source?[/li]
[li]Do you think that providing access to the underlying implementation brings more credibility and confidence to users?[/li]
[li]I think that, for open sourcing to make sense, you need an active community. Maybe I am getting it wrong but it should take energy and efforts to create and maintain a community? Or is it more self sustaining than I imagine?[/li]
[li]From a business point of view, I am still lacking confidence about opening the code. The usual open source business model (paid support, paid training, software integration services, ...) does not prevent a private company from "stealing" the code and embed it in a fully commercial solution. What is puzzling me is the fact that opening the code will create more competition making it harder to become profitable. What is your thought about my reasoning?[/li]
[/ul]

My plan is to focus the paid service on "taylormade" finite element software. The value proposition would be to develop and sell to our business customers a customized version of SesamX perfectly suited to their needs.

Thanks for sharing your thoughts

 

[FEA way]

It would be great to make this software open-source since we are still missing good FEA solutions that can be further developed by other enthusiasts. And that’s the biggest advantage of going open-source. You don’t have to work on this software alone - other people will be glad to help you with future releases. Of course, you can offer additional services and customized versions for a fee, as you said.

Also, to be honest, most people will be interested in a software like that only if it’s generally free and doesn’t have mesh size limits. That’s because of large amount of commercial FEA programs that have wider range of capabilities and sometimes don’t cost that much. Especially that nowadays all CAD programs have built-in FEA modules (usually quite advanced) and again there are low-cost offers for that.

Finally, your software relies on open-source pre/post Salome and that’s another reason for publishing its code.

 

[rb1957]

Yeah, that's an eternal question ... standard or personalised software.

everyone knows what NASTRAN delivers, the modelling may be in question but we understand how a CQUAD4 works (for better, for worse) but then you're limited to what NASTRAN has to offer.

A tailored FEA engine would do what you want, but nobody other than you would have much confidence in it (maybe even after you show the validation).
My problem with an open source FEA engine would be it's validation (even day-to-day)

I prefer to do crack growth using my own cycle counting excel s/sheet, rather than using published software (like CRAGRO) 'cause I know exactly what it's doing.

another day in paradise, or is paradise one day closer ?

 

[Ali baba Sx]

Thanks for your answer. When I mention tailored FEA, I do not imagine going as deep as modifying elements formulations. Instead, I imagine focusing on operations that can become time consuming and that can be automated. For instance, properties definition, loading case definition and so on.

Basically you more keen to trust MSC than an open source solution? Even if the open source validation procedure is freely exposed and consistent?

I agree with your last point about the excel sheet. But I guess you still have to manage manually the connection between the solver and your excel sheet. Even though I do not have enough data to tell if it would be relevant in your case, the solver could directly output the results in the excel format you desired. And then you are free to run your macro and analyze the data. This is typically what I imagine by tailored FEA.

 

[rb1957]

"Basically you more keen to trust MSC than an open source solution? Even if the open source validation procedure is freely exposed and consistent?"

Absolutely, NASTRAN is controlled software, an industry standard, used for decades, validated by many.

As I understand "open source" means anyone can change anything at anytime. So something validated yesterday may be invalid (due to some overnight code change).


another day in paradise, or is paradise one day closer ?

 

[Ali baba Sx]

Even though anyone can change anything anytime, there is still someone responsible for integrating the work of others. Besides, a strong validation procedure can be applied at each new release. Similarly to commercial software, a new release is not expected every night.

 

[rb1957]

so not really open source ?

another day in paradise, or is paradise one day closer ?

 

[Ali baba Sx]

Other people still have access to your code and can start parallel project. I guess it is still open source.

 

[Coreform_Greg]

I think it's fantastic that you've written a FEM code - regardless of whether it ever catches on. I also think it would be fine for you to open-source the project, if that's what you want. I do think, however, that you need to do some more consideration if you expect your code to form the foundation of a business. For example, it's not clear to me why a business (large, small, or self-employed consultant) should consider working with you or use SesamX. You claim that SesamX is "engineer friendly" - but I'm not sure I see that yet. Is it more "engineer friendly" than Solidworks Simulate, or than Abaqus? Those are very engineer friendly in my opinion - moreso than SesamX currently.

How will SesamX provide a better solution than Abaqus for your target customers? Who are your target customers? Recall that for businesses, the main cost is labor (as was mentioned on your Reddit thread). What benefits does SesamX provide over FEniCS? Over MOOSE? OpenCFS? ONELAB? ANSYS? StressCheck? ElmerFEM? Deform3D? Calculix? Simit? Firedrake? FreeFEM? Midas? My point being - there are a ton of options out there, and it's not (yet) clear to me why a prospective customer should choose SesamX over any of these other codes/paradigms. This is a challenge for every FEM developer, mind you, not just for you. FWIW, I wouldn't worry about a big code like Abaqus or ANSYS copying what you're doing - those codes won't ever be able to radically rewrite themselves.

A few funding options that you might explore are SBIR/STTR funding sources (if US citizen) or perhaps NUMFOCUS. You could also explore working as a FEM consultant, developing your code as needed to solve customer problems. Look at how groups like Kitware (devs of many open-source projects, including VTK and ParaView) operate. Or you could simply open-source it without the expectation that it will ever be commercially viable, but instead to serve as a fun project and build a community of like-minded developers & users.

I hope this doesn't come off as pessimistic, I just think it's a bit premature to think about building a business around your code -- open-source or not. I think you need more market research, and a clearer strategy. But I'm interested to follow your progress, and kudos for putting yourself "out there"!

 

[Ali baba Sx]

Thank you for your answer Greg.

You have put the finger where it hurts currently. The value proposition and the target customers are what I am trying to define. My assumption about it right now is that the target customer is the medium or small company that needs simulation software (which cannot afford to spend thousands of $ on licenses). The value proposition is not clear on my website, but it is along the lines I tried to explain on my previous replies: a tailored FEA solution that embeds all the "plumbing" with the whole simulation software chain (pre-processing, excel sheet, post-processing, ...). It may seem a bit vague and I need to refine and validate/check these assumptions.

Thanks for the funding options. I am not a US citizen but I'll have a look at NUMFOCUS.

And thanks also for your support. I'll keep you informed of the novelties

 

[rb1957]

Yeah, Greg's questions are key !

Who do you expect to use this, at least initially. You can't say "everyone, anyone" 'cause you don't know how to talk to "everyone" ... everyone are all different !! Think of an industry or an application, eg small UAV drones ? (I suspect they use SolidWorks or similar) civil eng'g ? concrete beams ??

There are many good (and some not so good) pre- and post-processors. You don't need your own FEA engine to do this ... and this is a real weakness IMO.

Good Luck.

another day in paradise, or is paradise one day closer ?