Who will the Engineer of the Future be? 4

[Ian Portsmouth]

"Twenty-five years from now, engineering will be a lot different from what it is today." - Captain Obvious, Jr.

Predicting the work and makeup of engineers that far into the future is a fool's errand. But what about 5-6 years from now? What will the job of an engineer look like in 2030, and what should a young or mid-career engineer start doing now to prepare for their future role?

 

[SWComposites]

Probably highly depends on the specific field and industry.
Young engineer should do what they should always have done, keep learning new skills in ones field, both in depth of knowledge and breadth of experience. Learn how things are actually built (hint, they are not what is shown in the ideal CAD model) by visiting sites, factories, etc.

 

[IRstuff]

It's unclear whether anything will really change at the core of math/physics related engineering. AI in the guise of large language models (LLMs) can't do math or real physics, because it has no conception of math, or physics; they're essentially really good at guessing, since that's their design, to guess the answer to a prompt.

I don't think that engineering will be efficiently accomplished by guess (and check).

That said, various incarnations of LLMs could improve engineering design efficiency by doing literature searches more efficiently. Finding papers on specific engineering designs/analyses, while easier with the internet, could potentially be even easier with something that can do all the searches and summarize the results for you. However, caveat emptor, since LLMs are prone to hallucinations, but it's potentially still easier to back-check the search results and summaries from an LLM.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[GregLocock]

I suspect in the future engineering reports will have better grammar and spelling, but will include logical looking analyses that are wrong. Aren't LLMs basically markov chain generators writ large?

What I see happening in my little world is that prototypes are fast becoming a thing of the past, so we are more reliant on simulations than ever, yet have less relevant correlation data to work with. Somebody (and it won't be me) is going to have to come up with robust modelling techniques that produce reliable results in the absence of correlation data. The problem there is that the level of modelling needed to eliminate fudge factors/physics based adjustments/ PBAs is an order of magnitude - a factor of 10, more than we currently have. Just a simple example - rubber bushing has 6 degrees of freedom. We measure or estimate the rate curve in each direction (it is non linear). Yet we do not account for the interaction of loads in different directions on those rate curves. In other words we've gone from needing 6 tests to 6^6/2 tests (or whatever it is). And then there's frequency effects. An alternative is not to model complex systems in all their glory, but to design systems that we can accurately analyse.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[geesaman.d]

Five years ago, the news media was convinced that "3D printing" was going to revolutionize manufacturing. For those of us who have been using rapid prototyping for a couple decades, that seemed rash. It was. Desktop 3D printing at consumer cost points is a thing now, yes, but for manufacturing the materials are the limitation and progress remains slow and steady.

Likewise with AI. It will probably overhaul low-grade marketing and copywriting content, but where accuracy (and liability) is paramount, I do not think it will change much. Like any automated tools we already have, you can just generate more junk that needs to be manually edited to remove the errors and oversights.

Another goodie is a 'digital twin'. I think most of us have fantasized about analytical methods that include 'everything' about a real system. Make up a pretty name that implies an unrealistically simple image of a wildly (impossibly) complicated reality, and you make money.

What I've already seen is the increasing gap between management expectations and actual engineering reality. It will continue to be more difficult for high-value 'Real Engineers' doing 'Real Engineering' when corporate is being buzz-washed about how the next new software tool will do it all for you, for so much less. Business plans that underbudget the technical cost or overlook technical problems that need solutions might still call in the engineers to solve the problems, but that's not a sustainable employment path for entry-level and stability-seeking engineers. So there will be even more churn, panic, and turnover as companies experiment with more reliance on the tools and then either failing or calling in actual engineers to solve the problems that the automated / AI tools simply can't resolve.

If I could imbue any skill on tomorrow's engineers, it's to study and verify the results of the tools they will increasingly rely upon.

 

[dvd]

Engineer of the future -

drafter
BOM clerk
purchasing agent
scheduler
expediter
typist
stenographer

 

[JohnRBaker]

Heck, when I was working as a machine designer/project engineer back in the 70's, I was already doing most of those tasks. Maybe not 'stenographer', but I often did my own detailing and typed my own reports, etc. As for purchasing agents, many times I already had supplier quotes which I would just pass-along to our purchasing department. And once a job was released to production, even if I was starting a new project, I was still responsible for seeing to it that the last project got manufactured, tested and shipped to the customer, and sometimes, if it was a first version of a new class of machines or it was a special customer, I would have to be there at the customer site for start-up.

John R. Baker, P.E. (ret)
Irvine, CA
Siemens PLM:

The secret of life is not finding someone to live with
It's finding someone you can't live without

 

[AZPete]

dvd,

It's officially classified as Engineering Support Support.

 

[geesaman.d]

Engineer of the future -
drafter
BOM clerk
purchasing agent
scheduler
expediter
typist
stenographer

I get what you're meaning. I think this is heavily driven by the size of the organization. Large companies put engineers in smaller boxes.
But smaller and mid-size companies are expecting engineers to do more with less surrounding staff to support those ancillary tasks. So there will always be a shift in that direction to push an engineer to handle wider responsibilities.

 

[dvd]

AZPete - I classify most of it as perfectionism by engineers who seem to think that no one else is capable of doing these tasks. Management eats it up and gladly provides tools tailored to that mindset.

 

[cranky108]

May I state that my daughter graduated as a chemical engineer recently, and has started into a PhD program.
She will be the engineer of the future.

Does that help?

 

[JohnRBaker]

I was hoping that our granddaughter would go into engineering, and while she's at least attending my alma mater, she decided to go a different route. Her major is Chemistry, but she said that the Math is so easy that she's decided to get a double major, Chemistry and Math. She says that she should still have time to get a minor in Physics. It helped that she started college with 46 AP credits from high school ;-)

John R. Baker, P.E. (ret)
Irvine, CA
Siemens PLM:

The secret of life is not finding someone to live with
It's finding someone you can't live without

 

[CWB1]

I dont foresee much changing but do expect the last of the luddites who need other engineers to do their CAD, FEA, CFD, and other work to retire or be fired. Patience for "I never learned" common undergrad coursework ran out long ago, along with the endless quality and timing issues caused by multiple engineers doing one's work.

 

[GregLocock]

The problem there is that FEA CFD and MBD are learned skills such that you don't have time to specialize in more than one. Any student can stumble through the 'Getting Started' tutorials in those packages, it takes 6 months before they can even produce a reasonable model, and several years before they can reliably produce trustworthy models. Of the three CFD is most likely to hallucinate, that is produce OK looking results that are at odds with reality, and FEA is the easiest to spot errors in if the observer is an experienced stress analyst or dynamicist.

So, no, Joe Blow pressing the automesh button on Solidworks is not replacing an FEA guy.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[geotechguy1]

In civil and related disciplines, I see AI and the move towards unified 3D models and unified FEM models for everything revamping professions and leading to a re-merging of some disciplines that have split up.

AI will separate the wheat from the chaf.

 

[drawoh]

geotechguy1,

My impression of having used FEA in SolidWorks over the last twenty years or so is that it is easy to solve problems of rigidity and vibration modes. Structural integrity, particularly at the assembly level, is way more tricky, with loads being transferred through all sorts of interesting fasteners.

If an AI operator pushes buttons and asks the AI program for answers to his or her questions, who is responsble if the final product catastrophically falls down or blows up? Are the AI vendors willing to take over from professional engineers?

--
JHG

 

[CWB1]

Complex system analysis like combustion development and crash analysis will always need a specialist, but the generalist FEA/CFD/CAE guy who ran endless bracket modals and laminar mixing analysis disappeared long ago at most companies.

Agreed on AI. It may not be able to solve "real" problems yet but its ability to optimize, identify relationships, and manage constraints is already saving a ton of time and enabling more to be done by fewer staff. Its also enabling powerful CAD-like tools for industrial artists/designers and other staff, saving engineering time/effort.

The responsibility for testing/validating a design will always fall on the releasing engineer regardless of the tools, AI, or humans used.

 

[SWComposites]

AI will just automate the FEA GIGO process. At some point the results will be tragic.

 

[geotechguy1]

At the moment the current form of AI is trained mostly on non-engineering / corporate data. But I'm imagining that, for example, as a geotech, an AI even with the current technology trained on all of a large companies historical geotechnical data, textbooks, calculation sets, and designs might be quite useful.

I see AI as a challenge to raise the bar in terms of individual performance. Alot of us sadly don't have a high level of knowledge that might be implied by our job title or salary - there's alot of answer googling, stalling clients to buy time to go and find an answer to a simple question you should be able to answer on the spot, or overconfident bullshit artists bluffing. I see a future test as this: If you can't answer a question in your supposed domain of expertise on the spot >90% of the time, why should a client pay $250/hr for you instead of just asking CoPilot or a future engineering-data trained version? We have to become the 'I can explain this concept to a child' type of expert rather than the 'I vaguely know that if I open Terzhagi and Peck 1996 I can find the answer to this in a few minutes' kind. We have to get better (which at least with the current iterations of AI is easily achievable IMO). Alternatively you can take a defeatist approach and 'give up', and perhaps many will - and should - give up and become PMs or something, but my observation coming of age in the google generation is that the people who rose to the top with reputations as experts were people who actually learned things deeply.

Also, the world is getting pushed towards tighter and tighter designs with lower factors of safety that are more economical, which means actually understanding things again rather than having implicit factors of safety of 10 or 20 or more built into everything.

 

[SWComposites]

Well, in the aircraft world, if we had factors of safety of 10-20, nothing would fly. We’ve been squeezing the margins out for decades in the name of weight reduction. More or less size everything to an factor of safety of 1.5 on design loads, but also account for fatigue, damage and degradation. It gets quite complicated. Imagining the next management fad being “lets design everything using AI” scares the **** out of me.