I've been doing academic research at my university and I've stumbled across an algorithm that could be incredibly...

I've been doing academic research at my university and I've stumbled across an algorithm that could be incredibly valuable and revolutionize engineering. The tl:dr is that I've devised a way for computer to automatically design mechanical parts and machinery that's many times more effective than the current state of the art.

Since I'm lucky and not bound by any contractual obligations in this case, I can keep the IP. The question is, what do I do with it? I haven't told anyone else about this yet, but I've made a prototype implementation and it works very well. I'm just not sure how to profit off of this since it would need to be integrated into a much larger, very expensive to develop CAD program if it were to be useful to end users. I was thinking perhaps offer use of the algorithm as part of a consulting package? It's very good at designing things, and a good deal better than a human, so maybe offer it for things like ICE and gas turbines where a few percent efficiency increase is a really big deal?

Pls advise

dude, we already have this. Supercomputers calculate even more efficient designs.


good that you wasted your time on tho, props and few internets for being a total loser kek.

Surely the best way to profit from this is to create software that provides the service on a bespoke platform... If its as good as you say word will catch on ultra quick and profit will be just as swift... Obviously easier said than done... If you can get Patent to the idea... Pitching to already bespoke package providers (AutoDesk) comes to mind wwould the be viable... If no patent approved/ or not appropriate, then make them sign a non-disclosure agreement, but this is a fairly week safety net... Obviously your profits will take a hit, but even if your getting >30% of the profits and it >could be incredibly valuable and revolutionize engineering then you would still become a millionaire from it... Not being greedy is the key here i feel and take a hit to make it widely available to the end users that it is most beneficial for!

MechE here. Elaborate a little bit more. I can help point you in the right direction.

Can this be specialized for a certain industry (like oil and gas for example)?

Also answer these questions (heilmeier's catechism)

>what are you trying to do? Articulate the objectives using absolutely no jargon

>how is it done today and what are the limitations to the current practice?

>what is new in your approach and why do you think it will be successful?

>who cares? If you succeed what difference will it make?

>what are the risks?

>how much will it cost?

>how long will it take?

>what are some mid term and final "exams" to check for success

You don't have to answer those here but if you do (please well thought out sentences, more detailed the better) we can help you. If you want to protect your IP then more power to you.

If you have well thought out answers to these questions, you can think of a reasonable approach to capitalize. But with your vague descriptions I'm not sure if what you're saying is useful/practical or been done before (I'm assuming you don't know since you are an undergrad)

zzzzzzzz

You didn't create anything. You're not going to "revolutionize engineering". You're just another neet mongoloid posting on the cryptoboard.

Eat fat amounts of dick post haste.

The whole point is that this is efficient enough to be done on a regular workstation, and doesn't experience exponential blow-up with increasing problem size.

I've considered patents, but I'm not sure how I would market them to these huge companies when I'm just an individual.

Well, without spilling the beans, I guess the best way to describe it is as an improvement to SIMP optimization methods.

It lets computers design optimal mechanical structures very efficiently.

Current implementations are slow, unwieldy, and limited to a narrow range of very specific problem types, whereas this is generalized to arbitrary objective functions.

I'm not sure who cares. That's the part that I'm trying to figure out by making this thread.

Also it's just an algorithm, so costs and risks don't really apply.

Just because you push shopping carts at walmart doesn't mean we all do :^)

So this is a way to have designed parts undergo small changes to bring for favorable results in Finite Element Analysis and do useful things like reducing stress concentration?

Would this algorithm change part dimensions to accomplish this?

One small note
>Current implementations are slow, unwieldy, and limited to a narrow range of very specific problem types, whereas this is generalized to arbitrary objective functions
People who do this all day are good with this stuff. I've worked with many CAD all stars in my career and they are pretty good at minimizing this stuff

Idea may have potential.

Best thing you should do is write a report and take it to one of your trusted professors. Maybe write up an NDA. They should give you better guidance.

are you sure this isn't some wicked combination of built in features (albeit obscure) of your CAD suite?

Also to elaborate on your report. Make it as official looking as possible.

>cover sheet
>abstract
>background/description
>procedure
>data/graphs
>pictures
>discussion of results
>conclusions
>real world applications

This report if well done can be a standalone component you have to explain your idea. While you make it you will learn your idea more as well.

Implement it as a paid plugin for Autocad/other CAD software. Yes you definitely must get a patent on this as well, google "rocket lawyers" and ask there, affordable price.

>So this is a way to have designed parts undergo small changes to bring for favorable results in Finite Element Analysis and do useful things like reducing stress concentration?
That's more or less what it does, but it's capable of much larger scale design synthesis and topology optimization. You give it blundary conditions on some control volume that you want to have the structure optimized inside of, and also a series of objectives like maximize the strength-weight ratio, minimize heat transfer, minimize drag, etc. It then designs the optimal structure for that part.

The beauty of this method is that you can choose your objective function to be any of the results of an FEA simulation, such as distortional energy, factor of safety, etc. as opposed to just one specific analytically derived objective like minimum compliance. You can also do multi-objective optimization where you can optimize for both structural and aerodynamic properties at the same time, for example. And it also lets you do nonlocal optimization, where your objective function can just focus on a certain region of the part, so far example you could maximize stiffness in one region but then minimize it in another.

There are existing structural topology optimization solutions used in industry, but they're pretty much just toys at this point.

Has potential. Take this to some professors you know that specialize in this stuff in the report I described.

They can have a unique insight based on their experience.

Not sure if this can be a plugin or part of a design service tbqh

I remember this thread from last year.

Dude, I don't speak engineering so I'm not quite clear. With that being said, I have a company in the solar power space and we're close to securing deals to go international. Would this be beneficial to us? Do you have any throw-away contact?

>Also it's just an algorithm, so costs and risks don't really apply.

Cost: development, maintainance (how often do you think cad updates break packages?)

Risks: piracy

me too

I'm currently working with two engineers on a project of topology optimization for mechanical parts of various rolling stock systems (metros, tramways). I would definitely be interested in your product, especially if you can prove its performance on some examples.
If you want a first trial with a client to see if it could be sold and used by direct customers we can arrange something (no problem to sign an NDA).

Thanks for the imput, we will be implementing your idea very soon in our production process.
Best regards,
Chang Chen

Thanks, I'll look into this.

I have no idea. I don't reay know anything about photovoltaics, but if it might be helpful if there's some kind of thermal system involved.

[email protected]

Be warned that it's still mostly a proof of concept at this point. More development work will need to be done.

user to be honest... hire me as sales manager. Forge a contract, vow to share 40% of sales profits with me and i won't put the phone down until we can buy our own football team

Hey OP if you're still here, i know an interested party in aerospace.

Feel free to send me an email Also sorry for all the shitty typos: I'm posting on mobile

I just sent you a message on the e-mail address

So you found an algorithm that uses an optimization function?

Whoop-dee-doo.

Scam

I'd check to see if your Uni has any profit sharing with patents. If you truly want to be independent, then DOCUMENT your improvements on ANY possible patents (especially those that don't have any plausible ties to the University), and hope for the best. Want to be rich? Create a company designed to be bought out by another (preferably local) company.

>ITT newfags that don't know this shit already exists and is probably thousands of times better than OP's algorithm
>muh silicon valley

Hell this is already baked into industry software like solidworks and revit. Nicememe

Way to hedge your reply bets.

>gottem, he doesn't know about post ID
It was an afterthought
Nice contribution to the thread there friendo

Your two replies essentially said the same thing.

Prove me wrong.

>probability
>commonly known specifics

Way to hedge your reply bets.

>Still hasn't provided any evidence to prove me wrong
>commonly known specifics
>>fing prove it
I'll take any reliable peer-reviewed journal as evidence.

Thanks, I'll see what their policies are.

I've largely stopped posting on Veeky Forums because whenever I do it tends to rile up armchair expert NEETs who can't possibly imagine someone being more successful with their lives. I'm a published researcher in this area, what are you?

google.com faggot

I actually do cad production for a living.
Someone on Veeky Forums said their way is best, seems legit, nice research there boss.

usually unis have you sign off any intellectual rights to anything you make there.
so maybe wait till you finish uni if it's really something.
altho "algorithm" that "designs better stuff" is fishy snake oil sounding shit op as a programmer i have to say you sound like a nigerian prince that never saw a computer up close.

I don't owe you anything. If you want to be butthurt over the idea of someone making something then go ahead and be butthurt: it doesn't affect me personally. I am very tired of having this same conversation every time I post here, though.

It's a spinoff of some other research I'm doing which is currently unfunded so I'm free to do whatever I want with this particular project.

Also this is more numerical methods and applied math than computer science. They don't teach computer scientists this stuff.

>Also this is more numerical methods and applied math than computer science

exactly what do you think computer science is about? if really CS, you sound like first semester.

very often you can make a deal with your technology transfer office beforehand. it's a bitch if you try to negotiate afterwards.

I mean real, heavy duty, non-analytic numerical methods, not that wimpy discrete math stuff. I know for a fact that these subjects are not taught

>I mean real, heavy duty, non-analytic numerical methods, not that wimpy discrete math stuff. I know for a fact that these subjects are not taught

what, like gradient descent?

PDE domain discretization methods, boolean relaxations, nonlinear programming, etc.

you're pretty good at making up concepts that don't exist, nor do they even make sense

Alright, I guess the finite element method is just a figment of my imagination then. I really wonder why I still post here.

usually you get to that in your master's, in courses like machine learning, etc.

wrong these are all first&second year stuff.
at least here.