Future of CAD

[matt]

@burhop started to ask a question about what we all thought the future of CAD held for us. He had a link to a podcast he starred in, and talked a lot about additive viewtopic.php?f=30&t=4&p=74&sid=7ac359e909b2f4561a597b3c2e6c3339#p74

Anyway, I just wanted to write a little about what I think CAD is going to become.

First, I think history-based stuff is going to become a footnote. It will still be there, but it will be a nuance you add when you need it. Direct edit really needs to be an available option, as well as some cross between Tspline-ish tools and subD, or even NX Synchronous Technology for complex surfaces. CAD is going to become this cross roads of data types and methods and tools.

Mesh data is going to continue to grow with the generative design tools and mesh editing tools. Mesh is going to become a CAD input as well as an output. Plus, some of the tools that have been exclusive to 3dprint prep software are going to have to become basic CAD functions. Stuff like support structures, figuring the best orientation for printing, media drainage, and other additive specific tools.

I think it's clear by now that cloud CAD really isn't a universal answer. Onshape is great, but it hasn't taken over. There are just too many security variables that can't be controlled. The cloud idea is great, but a public cloud is not. Private cloud installations solve some of those problems. Public cloud is maybe still the answer for groups with no resources who just need to get going fast, but I think more sophisticated customers will make better use of a situation that gives them more control.

Plus, the whole CAD-in-a-database thing has to gain some traction. file management should not be a separate issue from CAD. I think this is something Onshape has demonstrated nicely.

I think people have dreamed about stuff like functional design that actually does some of the engineering work as you create the geometry. You have to tell it what kind of structure it is, where it's supported, and it optimizes a shape for you with the appropriate joints, and assembly parts. I expected knowledge based CAD to be a bigger deal. Every thing we do now is like a custom one-off. Maybe for some kinds of CAD there should be specialized tools that just do machine design using AI elements to help you make the engineering decisions.

I worked a little with a group of guys who were trying to make a new CAD system that actually implemented some best practice stuff right into the way the software works. This included starting the design from the assembly with a layout structure, and breaking parts out from there. But not in the clunky ways we are all used to using. They had some ideas that were pretty cool about top-down that didn't run into all of these external reference issues, and dealt with the best practice headaches right up front.

Of course interoperability is something the market screams for, but vendors are evil, sniveling grovelers, and keep trying to lock customers into to a proprietary format. The only way to get this to stop is to kill off some of the too-big-for-their-own-good vendors and start over with several smaller developers who are willing to work together. But then the cycle starts over again.

Anyway, CAD is gonna get more complex, and it's going to bring more disciplines together, and operators will have to be more specialists than generalists. At least that's what I think. Some is happening now, and some is 15 years out.

[mike miller]

Me: @matt , can I hijack your thread?"

Matt: "Sure".

Me. "Okay thanks!"



I was talking with a CAD admin from a fairly large company yesterday that does mostly electrical and routing. He mentioned that they are in the process of migrating from Solid Edge to Inventor. After the lights stopped flashing and I regained my senses, I asked him why?

According to him;
-Inventor is much more intuitive and easier for interns to learn and it has a much better import tool than SE.
-IV has a better library of electrical components than SE.
-Support is better and faster.
-It meshes better with their Auto-Cad schematics.
-SE has a history-free tree and doesn't have to be ordered. (not sure why this is a drawback???)


Matt, what is your take on this? Is AD going to take the top seat? Is synchronous really just a fad?

[MJuric]

-Support is better and faster.

Wow, that really shocks me. When I was on IV one of the worst aspects was support. VARS didn't respond unless you had 50 seats or more and it was like dealing with an appliance company. "If we come out and it's something under warranty it's free, if it's something you did we will have to charge you"...and somehow it always ends up being something you did.

I actually think IV is a bit less intuitive than SW but I've never used SE.

[MJuric]

@burhop started to ask a question about what we all thought the future of CAD held for us.

To me the biggest future advances in CAD will be in the HMI. Our brains and the software(s) we use can can move at speeds FAR faster than the by comparison, TERRIBLE, interface we are stuck with.

Along with that I think we will start to see "Predictive" software that can start to see your end goal and will aid you to that goal. Gone will be the days of picking the right bearings, tolerances, materials, heat treats etc. You will start with a goal, IE I want to put a hole in this in this amount of time and the spindle will largely design itself.

Of course shortly after that Skynet is activated and all hell breaks loose.

[mike miller]

@burhop started to ask a question about what we all thought the future of CAD held for us.

To me the biggest future advances in CAD will be in the HMI. Our brains and the software(s) we use can can move at speeds FAR faster than the by comparison, TERRIBLE, interface we are stuck with.

Along with that I think we will start to see "Predictive" software that can start to see your end goal and will aid you to that goal. Gone will be the days of picking the right bearings, tolerances, materials, heat treats etc. You will start with a goal, IE I want to put a hole in this in this amount of time and the spindle will largely design itself.

Of course shortly after that Skynet is activated and all hell breaks loose.

Solid Edge already has some of that. They have a special toolbar that uses AI to decide what command you are likely to pick next based on your history. (I haven't used it yet, just saw it on demos.)

[MJuric]

Solid Edge already has some of that. They have a special toolbar that uses AI to decide what command you are likely to pick next based on your history. (I haven't used it yet, just saw it on demos.)

If it was in a demo it probably doesn't work :-)

I'm just starting to learn CAMWorks, haven't gone to any training yet. One of the things they tout is "Automatic feature recognition". Supposedly it is supposed to look at the part and say "Yeah that's a hole, apply this strategy for machining it" and so on.

I tried it couple times and it just seem to create a huge mess. I finally asked a friend of mine what I was doing wrong and his answer was "Well you're using that feature of the software....don't do that".

Apparently it is common knowledge that for most things it just doesn't work so most people don't even use it. So essentially you shouldn't use what is essentially the main selling point of the software....go figure.

[burhop]

Darn, late to the party. My wife had had me cleaning the basement the last 4 days and just let me out.

@matt Additive ends up in my conversations a lot mainly because it is driving change in design tools. Let me say upfront that I realize it is just another manufacturing technology like sheet metal bending, use of molds (injection and others), CNC machining and so on. Also, in comparison, it is used very rarely compared to other manufacturing.

I agree with you on the new types of geometries. SubD type stuff and others works great for lots of things already and becomes more useful with additive. The ability to use meshes is almost critical to take real advantage of additive. Otherwise, you are constrained to geometries meant for other manufacturing technologies. Plus, you will be able to use it in some other areas (e.g. injection molding, CNC).

The ability to use lattices is almost a requirement for true design for additive manufacturing (DfAM) but its not so useful in other areas (can you imagine trying to CNC a lattice!?!). However, there are some related areas for surfacing textures, surface structures and such that still transfer to other manufacturing types (imagine your super complex pattern feature recalculating instantaneously!)

I am curious what everyone's opinion of "Generative Design" is or even what we think the definition is. I don't mean the overly hyped marketing version of this but a more humble practical version of software creating useful geometry for you. Where do we think this will go?

[matt]

Darn, late to the party. My wife had had me cleaning the basement the last 4 days and just let me out.

I am curious what everyone's opinion of "Generative Design" is or even what we think the definition is. I don't mean the overly hyped marketing version of this but a more humble practical version of software creating useful geometry for you. Where do we think this will go?

Personally, this is one of the things I'm most excited about. You always see these stress analysis results using parts made from rectilinear shapes. Real stress doesn't like sharp edges and straight lines. Did you ever visit the Sagrada Familia or any of Gaudi's architecture? Way before computers he designed arches that take stress optimally. Real structure is more bone-like. If I understand Generative Design, it is basically shape optimized for stress, and these shapes are complex, and not rectilinear.

I think we should be inputting locations for connections or pivots between parts, and the software will minimize material/weight while maximizing stiffness, and that will result in swoopy bony structures that can be made best by additive processes. We need to be able to CAD that way.

Taking it a step further, these days we still focus on part design a lot because that's what our tools force us to do. We should be able to focus more on the finished assembly or final product, or the intended function or motion for mechanical assemblies.

Combining some of these ideas, making hollow parts where the empty air volume is replaced with lattice could help us design all sorts of things that could only be manufactured by processes like rotational or blow molding or gas assist/structural foam. Think tires or even tanks or bottles or areas of plastic parts where you have big thick sections.

[bnemec]

Darn, late to the party. My wife had had me cleaning the basement the last 4 days and just let me out.

I am curious what everyone's opinion of "Generative Design" is or even what we think the definition is. I don't mean the overly hyped marketing version of this but a more humble practical version of software creating useful geometry for you. Where do we think this will go?

Personally, this is one of the things I'm most excited about. You always see these stress analysis results using parts made from rectilinear shapes. Real stress doesn't like sharp edges and straight lines. Did you ever visit the Sagrada Familia or any of Gaudi's architecture? Way before computers he designed arches that take stress optimally. Real structure is more bone-like. If I understand Generative Design, it is basically shape optimized for stress, and these shapes are complex, and not rectilinear.

I think we should be inputting locations for connections or pivots between parts, and the software will minimize material/weight while maximizing stiffness, and that will result in swoopy bony structures that can be made best by additive processes. We need to be able to CAD that way.

Taking it a step further, these days we still focus on part design a lot because that's what our tools force us to do. We should be able to focus more on the finished assembly or final product, or the intended function or motion for mechanical assemblies.

Combining some of these ideas, making hollow parts where the empty air volume is replaced with lattice could help us design all sorts of things that could only be manufactured by processes like rotational or blow molding or gas assist/structural foam. Think tires or even tanks or bottles or areas of plastic parts where you have big thick sections.

I though what Solid Edge was doing with Generative Design was cool, as you stated using FEA outputs to optimize geometry is cool. I just don't know what markets will get use out of it's raw outputs as it's shapes are so organic they rule out many manufacturing processes. I only saw it used in stress analysis, but I wonder about fluid flow analysis to shape plenums, manifolds, nozzles, impellers, etc. The work needs done on how to take the raw output from GD and rework it into something manufacturable on a large scale IMO. I think that area is moving though.


The other area in CAD I wondered about while attending a couple Additive Manufacturing classes. It's been a couple years ago now, but the main file format in use at the time was .stl which was not meant to be used for that. I think VRML was quickly gaining popularity allowing more than one "color". I think there are new ones now that allow multiple colors/materials and use xml syntax. CAD packages were scrambling to add file types to the export list and add-in functionality to send your model to "printing services in the cloud" Which seems to me to be low hanging marketing fruit. Since there are so many AM technologies now that allow multiple materials (I'm thinking metal sintering various alloys) when will be able to model a part that has gradients from one material to another? This could even be done to some degree with FDM (although at a very low "resolution") but I think you can get the idea. Was it GE that received FAA cert for using AM fuel injectors that are in use on commercial jet engines? In a way the manufacturing capabilities are out in front of what we can model. I don't know anything about sintering metals with AM processes, but imagine where we could go if a single part could be made of different materials with different expansion, conduction or elastic properties. Almost like doping silicon.

[Frederick_Law]

I am curious what everyone's opinion of "Generative Design" is or even what we think the definition is. I don't mean the overly hyped marketing version of this but a more humble practical version of software creating useful geometry for you. Where do we think this will go?

Generative put material where it is required.
It will help in lots of area.
Much more important in additive.
Conventional production remove material so generative won't help with raw material use. It'll help the final product.
Additive benefit from both.

Generative need to advance to produce "simplified" profile with line and arc.
Not those "eroded" shape.
Right now operator need to draw the profile around generated profile.

And yes, we need more lattice and geometric pattern support. Easy way to fill with hex, circle, triangle, to cut or add.

[Andy_Sanders]

This may have been mentioned but I always thought it would be awesome to have instant "live" FEA results as you design. Setup your FEA parameters and the FEA will be able to be turned on as an option to see where you are at all times as you design. Make a tweak, and instantly see what is affected.
Obviously the tech and compute power isn't there for this to be always on and shown, but it would make the iterative steps a whole lot easier.

[jcapriotti]

This may have been mentioned but I always thought it would be awesome to have instant "live" FEA results as you design. Setup your FEA parameters and the FEA will be able to be turned on as an option to see where you are at all times as you design. Make a tweak, and instantly see what is affected.
Obviously the tech and compute power isn't there for this to be always on and shown, but it would make the iterative steps a whole lot easier.

Obvious it wouldn't work in real time. but what's to stop them from calculating in the background and use those extra cores we pay for?

[Ry-guy]

I only saw it used in stress analysis, but I wonder about fluid flow analysis to shape plenums, manifolds, nozzles, impellers, etc. The work needs done on how to take the raw output from GD and rework it into something manufacturable on a large scale IMO. I think that area is moving though.

I recall Solid Edge has the ability to take an output from Generative Design process and give you a more prismatic version that could be made using traditional "subtractive" methods (aka milling and machining)??

[Ry-guy]

My only concern with generative modeling is that you have to input what you think are going to be the stress location and the amount of stress that is going to occur. Then maybe apply some region rules for "keep out zones". Then you get your funky shape.
My concern with all that is that it describes a perfect world. Not a world full of accidents. Say you drop a heavy tool or something on one of those skinny connecting web-like parts of your gd part. THe part was not designed for that. Or say someone slides a large machine or steel case into the side of a GD part. It then fractures several of the spider legs and you can't tell. Then the real load happens and the part shreds.
GD is only as good as the original thoughts that go into it. Same thing applies to FEA and Simulation!

As for the future of design and CAD, well I feel we are moving, like someone already mention to a more CAID system. I'm copyrighting that now! Computer AI Design. CAID or better yet C-AID! or Computer AI Design & Engineering which gives you C-AIDE!
Where you sketch out a basic design using something like that Siemens 2D sketching app...ah...can't think of the name of thing anymore..we you could sketch out a basic framework and then start dropping in "connections", define a gear, have it go online search a db somewhere and find a motor to meet your requirements and then drop it into place. Then it looks for mounting holes and makes recommendations for mounting plate thickness and connecting hardware.
Think about letting the AI do the grunt work of design and then come back in to the CAD and run the simulations to verify the design results. Then move to the next step and let the machine do a full system optimization and allow it to iterate the full design of the product- not just the part. Llok for things like biggest constraints to the design- like to tight or loose tolerances, etc.

[KQuigley]

I kind of avoided this one because I don’t want to seem like a grumpy git. I do think the CAD industry is in a bit of a crisis right now-from the customer’s viewpoint. The state of the global economy means investors are looking for places to dump cash, and CAD vendors are a prime target. Investors love cloud. They love subscription. They love scalability.

One of my customers once said to me, “I love mortgages. When a good employee buys a property, they are tied into me, because we fund the mortgage by paying them”. Cad vendors want, above all else, to tie you in.

I say this because over the last year I’ve had calls from no less than 4 investment companies asking about opinions on the CAD market. Investors absolutely love 3DEXPERIENCE btw. Dassault are very very good at investor relationships.

As for the actual future I predict two things I am 110% sure about.

1. All the main vendors will offer free or very low cost subscriptions to their platforms aimed at the education and hobby market.
2. The running cost of using commercial systems will increase significantly.

As for actual feature sets I don’t think we will see much difference from what we do right now. I’ll chip in on @burhop comments on additive and generative as these are two areas I’ve worked in for years.

Firstly additive. We typically buy in probably 10k of additive parts a year. We have a Stratsys uPrint in house for design work, but the volume is SLS end use parts. Make no mistake, additive is chipping away at those niche high value applications. The infrastructure is growing, and issues with finishing are now irrelevant. I’m not talking about desktop FDM here, I’m talking about MJF or SLS industrial platforms with automated finishing.

As Mark said, lattice is important for additive, but more important is designing the lattice or the geometry for rapid and reliable building and post print clean up. This is particularly relevant in resin based or metal systems requiring supports. The issue has been that to do lattice work you typically offloaded it to the additive vendor. But that is changing. Tools like nTopology or Carbon’s new design tool will enable designers to design for these systems.

All this links into generative. To truly exploit additive, you need to minimise material use and create parts that only additive can do...otherwise what is the point? We’ve all seen the generative stuff Autodesk knocks out right? Skateboard bits, VW camper bits, seating etc. This stuff is all PR nonsense designed for press releases and trade shows. If you want to see actual generative design that considers use and manufacturing process look at the Carbon helmets or the Altair Inspire product case studies.

Altair are one of the companies doing it right in my view. They are a simulation led business and co-developed design modelling tools to generate design solutions. But the difference is the designer is in control and dictates. Other generative systems try to take the push button approach...define boundary/loads/materials and ping...out spits 1001 spiders web design solutions.

Latest incarnations of generative tools DO try to factor in actual manufacturing processes, but a lot of it is still automated and you basically pick from a series of results.

The most exciting stuff I’ve seen in the last 5 years has come from analysis based businesses not CAD vendors (who are pre-occupied with changing delivery platform rather than expanding the technology. Companies like Ansys, with Discovery Live (factoid...this is a great deal on subscription as it comes with Spaceclaim). SimSolid now owned by Altair. I mentioned Altair Inspire already but that has grown to include many more use cases.

But here’s the thing. We don’t use much of this stuff.very few of our customers use any of this stuff. You go to trade shows and you see the same case studies wheeled out year after year. Because the people that do use this stuff are the biggest players, because they can afford it. The fact is analysis generated design is complex. For most smaller businesses in most markets, saving a few grammes here or there is irrelevant to the product. Saving £50 off the prototype costs is far more relevant.

Which leads onto what I think will be a big change in the next 5 years. CAD linked to the supply chain.

Forget geometry. That stuff is far too hard to develop. But linking part data to production processes to advise on costing and part design...that is doable, and already happening. Try out real time quoting from Fractory or Xometry. Try a quote from Protolabs to see design advice and live costing. These companies work with the CAD vendors and some build add ins for the CAD tools. I think it is only a question of time before we see vendors doing this for themselves. Sure there are already costing tools and design advice tools inside many systems, but these are, well, a bit crap.

So, 5 years down the line, cloud enabled CAD live linked to numerous manufacturing platforms with design advice like the live spell checker in Word. Injection moulding? Highlighting thick spots, likely warp, sink marks, as you design, then giving you indicative part costs and tooling costs with a list of suppliers if you want it (or an API for your own in house overheads and capabilities).

This is a big subject so I’ll shut up now. I haven’t even started on tablet CAD or VR.

5 years down the line, the CAD landscape will be much the same as now, but the direction of travel will be much clearer.

[burhop]

5 years down the line, the CAD landscape will be much the same as now, but the direction of travel will be much clearer.

Yep. I think one issue (and something I do myself) is that if you are in the middle of the the "long tail" things, you don't always realize you are not anywhere close to what real designers do. Frankly, it is why I like hanging out here

[HerrTick]

So much of this "future" stuff has been lurking under the hood n NX for decades. Direct modeling? Did it in the 90s. Sub-D? They call it facet modeling.

Sub-D is becoming a bigger deal as wearables put more demands on CAD for stretchability. Also any capability that aids in processing anatomical scan data.

Textile modeling crosses the line between CAD and FEA. FEA-like algorithms are needed to process textile draping and stretching.

[SolidMuse]

If it was in a demo it probably doesn't work :-)

I'm just starting to learn CAMWorks, haven't gone to any training yet. One of the things they tout is "Automatic feature recognition". Supposedly it is supposed to look at the part and say "Yeah that's a hole, apply this strategy for machining it" and so on.

I tried it couple times and it just seem to create a huge mess. I finally asked a friend of mine what I was doing wrong and his answer was "Well you're using that feature of the software....don't do that".

Apparently it is common knowledge that for most things it just doesn't work so most people don't even use it. So essentially you shouldn't use what is essentially the main selling point of the software....go figure.

Most people do not take the time to really try to make Feature Recognition work. One has to dig very deep into the raw database and get a real good understanding of how CAMWorks works under the hood to make it work. Unfortunately, the VAR's have precious little knowledge on how to support a CAMWorks user. I have one of the very few CAMWorks users in the world that does take advantage of Feature Recognition. We are in the process of changing VAR's to get the level of support we need. We have also engaged HCL directly to get what we need to make it work. It is a powerful tool if one takes the time to make it work for you.

Your mileage may vary....