Direct vs History based modelling

[Anonymous]

There is another thread going with this sort of conversation but that thread is more to do with history based modelling than direct modelling.

I find this topic very interesting and have had this conversation a number of times with no real conclusion.

I would be forever grateful if somebody could give an example of how history based modelling is needed from a mechanical engineering point of view.

Ideally what I would like is a solid example, such as if a person was designing a mechanical device how changing something downstream would not be possible with direct modelling, or some other example please?

[Beyondforce]

Well, let's assume they will do something about it after the Sheet Metal.

[PhilProcarioJr]

@Beyondforce

The only reason I say I don't think they will get rid of the DM environment is because of discussions I have had with the dev team. I am pretty much in constant contact with them and they have explained a lot of why things are the way they are and what was done and where they are headed. Most of the stuff I can't talk about openly as I am under NDA, but either way from my perspective and what I know I don't see them dropping it, but I could be wrong.

As far as people go you basically have 3 types:

1) People new that haven't learned enough yet.

2) People that have learned but want Fusion to be the "App" they are already used to.

3) People that are dedicated to what they do and will learn all the ins and outs before committing to a big project.

1) People need to be taught the rules and all the ins and outs.

2) People I don't help them because they don't listen and just want to rant and have their way, or they think they know everything which means they are smarter then me and I can't help them because I don't know everything.

3) People just need the answer to their exact question.

When I answer forum posts I usually know which type of person is asking and answer accordingly. If I don't know then I give them a answer to their exact question but tell them there is other ways if they want me to explain. If they want more ways they will ask. Sometimes people have to fail to really learn and holding their hand doesn't guarantee success.

Components and bodies are all through the help, if someone wants to take the stance that they don't need to read and documentation then I say sucks to be them.

In the old days the typical forum response for the people like that was RTFM. I'm sorry to say I miss those days....

Remember you can lead a horse to water but you can't make it drink....

I mean seriously if you want to fly a plane you don't just jump in and figure it out as you go and if you do then you deserve the outcome that awaits you.

[PhilProcarioJr]

@Beyondforce

"Well, let's assume they will do something about it after the Sheet Metal."

They wont, it's not in their immediate 6 month to 1 year plan...unless something changes.

[Anonymous]

I have been reading the replies on the timeline and this thread; I have to say that Fusion 360’s DM capabilities are not getting a fair deal.

I think they are fantastic and have worked to two Keycreator videos to see how Fusion 360 would stand up, the program more than held its own.

As far as documentation and improving the function, it’s just so easy to use and pretty complete so maybe those are the reasons?

I’ve been working in direct mode for about 8 weeks now and have no complaints, everything runs very smoothly and the adjustment capabilities are superb.

[daniel_lyall]

@Anonymous It all done to what you make dead object it does not matter.

Cabinet making I would not do it in DM at all, a cabinet, bookcase is a living object you need to be able to change sizes on the fly have a nest and made object in the same file.

Modeling shed and basic buildings DM is not the best way to go.

There are a few other thing a timeline helps to make.

But if you like doing stuff in DM have at it, It's up to you in the long run.

[Anonymous]

In DM mode you can and do change everything on the fly, in fact that is the whole benefit of DM; the model is a living organism.

Whereas in history based modelling the model is a living organism held back by copious amounts of information.  

It’s not only that I like it; it’s also that I see no benefit in a history based system (yet).  

[etfrench]

If you're building a model from which multiple objects of differing sizes need to be made, then the Timeline mode would be best.  If the model is one where only one object will be made, then Direct Modeling may be the better choice.  Timeline mode is like a garden path.  You can wander down it until completely lost and you have to cry for help.  Direct Modeling exposes all (or most of the changes) to a model, so if you wander down the wrong path, it's easy to change the model without losing any previous changes or additions.

[daniel_lyall]

I know you can do single objects, with cabinet making one paramater or dimension may drive 20 other bits what can include user parameters with dimensions pulled from other parts. and include captured positions. it's the captured positions that needs the timeline so I can go back and forward from built to nested and update as needed.

I tried in DM it was a fail, I tried a lot of ways to do it.

It's not complex it's simple math driving from 1 to a lot of parts what are moved around.

useing the timeline I can do a save as, then if I need to remove a part I can move the timeline back change then step forward to rebuild it, then just step forward one at a time updating what needs updated.

doing it this way I may spend a few hours building a cabinet the first time, but then if it needs a feature needs added, removed or a dimension changed  that's all I have to do.

otherwise it's defeature do some sketches delete some bit re extrude parts that needed defeatured do each dimension change on every dimension that's the same, an hour or so latter finished after haveing to do the same to the nested part. and if it has a lot of hole very rude word F that in DM mode one at a time no way.

time is money.

it's the same for framing you do some stuff first then do stuff that needs moved. in DM you have to do a new component project onto its origin then hope you got it correct, time line you just move it back and project it.

or if in DM you have to do ref sketches for angles and positions for stuff.

that gets quite complex.

the disability stuff I do most of it is assembly's and To keep costs down I have to use the same size bolts screws so on so there are a lot of holes the same size haveing to do one at a time hell no.

As i said depends on what you are doing.

so as a test do 30 holes in a 10 x 10 x 3 inch plate see how long it takes to change all the sizes, then do it in time line modeand it will be seconds

[Anonymous]

One of the Autodesk employees mentioned that they introduced history because users were having trouble with long parts (not sure but I think that’s what was said).

I find that interesting but unfortunately it was not expanded upon.

[Anonymous]

That’s a good idea; they should set a task and see how the two different systems handle that task.

[daniel_lyall]

you would not need to test the hole feature it is minutes versus seconds just knowing how it works is enough 

[O.Tan]

@Anonymous, I use Fusion primarily in DM (you can ask other people here if you need verification about this), and yes. DM is powerful but as with all tools, some tools are better at completing some jobs then the other.

To keep it short and simple, History Based modelling works best when the user have an end goal of how the final product should look like. For e.g.: this cabinet has 4 partitions and etc., reason being that a properly modelled part will allow a seamless update from upstream to downstream,

but for those who isn't sure how they would like the final product to look like or from a mechanical engineer perspective, they're more interested to model a part that'll solve the issues they encounter, then having to care about modelling it the correct way, Direct Based Modelling will be better.

DM could be more powerful then what it's today as you can see from my post in the other thread.

Also I posted an IdeaStation here which will improve DM which would greatly appreciate your votes.

[daniel_lyall]

@O.Tan if we had access to parameters for DM and could link parameters to many objects DM would would be all we need that is the only hold back for me.

Just working out how to make something look DM is quicker for that.

[O.Tan]

Problem is, at the moment Fusion doesn't store parameters in DM, at least that's the impression I got speaking to one of the employees awhile back.

From the way I see it, timeline is AD answer to that which basically just "stores" user created features on top of DM. Remember how at the beginning it records all our "moves" and we all know how messy it was. IMO, AD created timeline as a quick way out so they can go about developing other things but it basically inherited all the issues with History based model.

Sorry to do this comparison again, but hear me out, since SolidEdge Direct Modelling sits on top of History model (which will explain why creating a feature in DM will be seen in History but not the other way round, they solved this in ST9 by having some kind of "ghost" preview), to put it in simple words, SE DM is simply a real fancy and flexible History solver.

Saying that, there's some areas in Fusion DM that it "stores" parameters, case being that we're able to edit holes and thread and even patterns after it being created.

[TrippyLighting]

---

@O.Tan wrote:

Also I posted an IdeaStation here which will improve DM which would greatly appreciate your votes.

---

Done!

[daniel_lyall]

There is nothing wrong saying what program the idea comes from, it gives the devs somewhere to look. 

[jeff_strater · Accepted solution]

Direct Modeling vs Parametric Modeling... I have some reservations about engaging in such a religious topic, which is why I haven't chimed in sooner. It might be safer to comment on the US elections... As you can tell from the discussion in these threads, there are strong beliefs here.

The answer, of course, has been stated many times in both of these threads: Each technique has its own benefits and costs, and places where each is useful. Fusion offers both, so you can choose whichever is better for your own use, or even mix them in the same design. This is, in fact, where the "Fusion" name comes from. We envisioned Fusion as a "fusion" of the parametric and direct worlds.

Specific to Fusion, I can shed some historical light on the topic. Originally, Fusion was a pure direct modeler. When SpaceClaim was first released, it reignited interest in direct modeling, which was seen as a bit of an "old technology" at that point. But, with new UI, direct modeling was suddently seen as having a valid place in CAD. So, we implemented the first versions as a pure direct modeler. But, soon, we realized that it was easy to get to the limits of what direct modeling can do. Global changes (where you want to change the overall length of a design) are very difficult in pure direct modeling - you sometimes have to select an awful lot of faces in order to change the overall length of a part. So for traditional mechanical parts, it was easy to run into the limits of direct modeling. There are also cases where modification in direct modeling is very difficult to achieve, such as a global finish fillet radius. If you have a bunch of features that all want to share the same fillet radius, it is hard to modify this in a direct model. There are no parameters in a direct model, so you cannot just change one value and have all the fillets update. So, we started a project to bring traditonal parametrics to this product. That project, if you are interested, was called "Polaris". We introduced parametrics and joints at the same time.

From a development point of view, I loved direct modeling. It is so much easier to implement a direct modeler than a parametric modeler. But, as a user, I did quickly hit the limits of direct modeling. I created a few designs that, while they were easy to create, were a big pain to edit later.

Parametric modeling has its own challenges from a user point of view. That parametric power does not come for free. All those relationships have to be understood and managed. In fact, @TrippyLighting and I taught a class on this topic at AU this year. And, as has been pointed out in both threads, unless you really understand your parametric design and how it is built, it is very easy to break a parametric design with what appear to be small changes. So, it's no cakewalk, either.

The topic of SolidEdge/NX has come up in these threads, as well. Yes, I have to admit, this product has achieved what appears to be a nice blend of direct modeling with "parameters and constraints". From a development point of view, this appears to be a very nice implementation. I have a lot of respect for that development team, to be completely honest. @PhilProcarioJr: I usually agree with most of what you say. However, I have to disagree with your statement that "having a direct modeling environment work in a parametric way is actually not that hard to achieve". I know this because we did try to achieve it. In the very first, direct-only incarnation of Fusion, we did try to implement "geometric constraints in a DM model". We were able to get it to work in a few cases, but it actually turns out to be pretty hard to get a good, general, implementation. I have not used NX a lot, but it would be interesting to use it in a large design, to see how it reacts, and what the limits are. Anyway, in the end, we decided to put our resources behind traditional parametrics instead.

On the topic of the mixed modes in Fusion, I admit that this is not as smooth as I had hoped it would be. We had a lot of passionate internal debates about this: What should the "boundary" of a direct modeling "island" in a parametric design be? At the component level? At the Body level? At the feature level? We decided on what you see today: A document can be entirely direct or parametric. Within a parametric design, you can have a "Base Feature" which is a body that is a direct modeling body. When you edit that feature, you are in a little direct modeling mode, within a larger parametric design. You can have a Base Feature with downstream parametric features. You can convert a series of parametric features into a new Base Body. It's pretty flexible, but as mentioned here, can be a bit "clunky", I admit.

@Anonymous mentions the fact that in Fusion direct modeling, patterns are associative. Yes, this is true. We found, fairly early on, that patterns was one weakness in "pure direct modeling". So, we added some very basic associativity to Fusion DM patterns. It works OK. There are limitations to that associativity. But, as we went onward, we discovered that there were more and more areas where "associativity" was needed to smooth the workflow. Edit Feature was another case. In DM, you can do some limited feature editing (assuming that the feature was not modified by other geometry). So, you can "edit" a fillet that was applied recently. You can "edit" a hole, unless some other feature has modified the hole geometry. What we found was that it became a slippery slope, and, for all the work we were putting into this "associativity", we might just as well go all the way to parametrics.

Anyway, this is a fun and healthy debate, but in the end, it comes down to personal preference. I like and use both. My top-level design is almost always parametric. If I am creating a component that is very simple, and which I know is not likely to change, I'll do it as a Base Feature in a parametric design. Or, I'll create it as DM design, and XRef it into my top-level parametric design. I find that this techinque results in a shorter, more stable timeline, and less compute time. But, for the more complex parts of the design I find that parametrics is the way to go. I know how parametrics, work, though, so I can avoid the pitfalls that others run into.

Jeff

[Beyondforce]

@jeff_strater,

I think your post sums it up really nicely!

Ben.

[O.Tan]

@jeff_strater, thanks for the long and detailed reply, appreciate it.

I just hope this doesn't mean it can't happen in the future, as back then Fusion doesn't have parametric so understandably some features is just too hard to implement. But that isn't the case now isn't it?

I hate to bring this up again but SolidEdge managed to make it happen from a history based modeller, whereas Fusion started of as a direct modeller and now that it do have parametric, I don't see why it can't be done.

I believe the key to how they (SolidEdge) make it work is as follows:
1. In a component, the user is able to switch between DM and TL without losing features created in each environment, they're however this rule to understand:
    a. Features created in DM will be made available in TL but any modifications made in TL will not reflect in DM
    b. This also means that features created in TL will not appear in DM (hence those who want to work only in TL will not have their workflow affected), unless the user chose to want to work those features in DM, the user has to "convert" the TL into DM and accepting the lost of all parametric features
2. The user is able to set by default if they want to model a part in DM or TL.

So it looks like SolidEdge DM sits on top of TL and that'll explain why any features created in TL will not be reflected in DM whereas DM features can be seen and edited in TL. Does this now make sense why something like this is possible to be implemented in Fusion?

I remember in the first iteration of Synchronous Technology (what SolidEdge calls their Hybrid Modelling), you can't really do this back and forth thing as you can only go from History to Direct, but a year or two later they allowed it but only for SheetMetal, and subsequent years open it up to all features. So it's a phase-by-phase implementation, something that I feel is doable with Fusion as well.

Of course the most recent version have a "solution" for the issue where in DM, feature modified in TL goes missing making it hard, to modify features created in DM with respect to features in TL, but it took them 9 years to figure it out and as a pioneer nonetheless!

And Fusion now has the privilege to learn how they do it and implement it their own way (and no, create base feature is not it as it still force the user to work in a TL and not the other way round). Again, I know they're many features that Fusion still need to implement but I hope is that future available resources will be directed towards this work, and if that takes a year or 2, so be it. But don't close the door on hybrid modelling or whatever you'll call it internally.

and let say that in 2 years you'll be working in it, and I'm still around, I'll be willing to help you guys to figure out implementing it. Peace out!

[cekuhnen]

I think that it is time to stop using the term parametric and think about the timeline as a generative modeling approach (similar to how features are stacked in SW and such)

it has 2 useful applications.

1. The timeline shows you clearly how you evolved your design. Step by step you can go forward and backwards

the timeline because it lists features after features as they are created makes reading a design very easy.

Here  you can also drag if possible features to different positions in time to clean up the timeline, organize, or restructure your process.

it is important to now however that later steps never can be send when editing an earlier feature. That is the downside of this process.

2. The second very useful function which is why I recht parametric and rather call this generative is that you can not only edit a feature

but if it allows you switch feature inputs, fezze features, meaning in one design you can have multiple variations included.

Features also also do not only include changing a fillet radius but also moving an object, rotating an object, or for the res of the timeline removing a body.

From @PhilProcarioJr I keep hearing people say that the timeline is so powerful because you can go back and edit a sketch or feature and have everything downstream auto update. I'm sorry but that whole line is a load of crap. 9 times out of 10 changes made to upstream sketches and features breaks most of the model going downstream unless you spend 20x more time prefiguring everything out. Your playing the what-if game when designing. So you spend a ton of time solving all the what-ifs to have a solid stable model. Now I fully understand that for simple designs History is great but if your model gets very complex you can throw that out the window, thus the need for both.

I beg to different here. 3D modeling no matter what workflow requires experience and skill. In the timeline with a feature driven by a sketch I can break the following features that use that sketch when I adjust the sketch in a wrong way. And that is not limited to the timeline but happens in every application that uses geometry that I should driven by a feature like Alias SolidThinking, Cobalt, everybody.

So with the timeline or any design history there are limitations to it how you can use and have to use it.

So if something fails it is not the softwares fault but because the user simply did use it not in a way he should. Period!

Now direct modeling for certain models is a lot faster and more push n pull modeling but then I cannot really use a generative modeling approach either.

And for others a feature based approach is faster because with the sketches and surface tools I set up relationships in a way that I know later can manipulate and adjust and don't have to go into direct modeling and adjust / rebuild work.

For me the biggest draw back in Fusion if there is one is that strict focus on 2D sketches and not being able to offer true 3D sketches at the moment.

This particularly can be noticeable when doing furniture and you have to work with 3D pipes.

I agree that complex feature based models can become complex later on. But I am not convinced that just a DM model would be that much easier or the solution to the problem.

I found that often smart working is key and a complex design is just what it is a complex design.