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?

[kb9ydn]

Alright, so enough dead horse beating.  Lets get to some specifics.  I've have some time today to mess with DM and I've already run into some questions.  These are probably very basic so hopefully they aren't too annoying.  All of this assumes no design history capture.

1)  Lets say I create a box that's 3x4x5.  I do some other stuff and then later find out that the box needs to 3.34243 x 4.3455345 x 5.  Editing the sketch doesn't work because it's no longer linked to the box in any way so to change its size I have to move one of the faces.  That's fine but in order to do that I need to know how far to move the face, which means measuring the current size of the box (because it's 3 days later and I don't remember what size the box was to start with) and subtracting that from the size it needs to be.  This works but seems rather clunky.  Wouldn't it be faster to be able to assign dimensions directly to the box somehow?  Or maybe instead, be able to select the face to move and then have it move to where it's X distance from another face?  I feel like the move command should be able to do this, but either it doesn't or I'm just not seeing how to do it.

2)  In DM mode it seems you can't click and drag on a component to move it around.  Everything is basically fixed unless you move it with the move command.  In this mode it seems like joints are kind of useless.  They limit what you can do with the move command but since the move command requires such deliberate action to use anyway, what does adding joints gain you?  In DM mode I'm getting the impression that the move command is going to be used VERY frequently, which makes me think that it would be SO much faster if things could be moved by click-and-drag.

3)  Patterns are completely non-associative and can't be edited later.  If I need to change the number or the spacing of items I can't.  I tend to make heavy use of patterns and mirrors so this is not so good.

4)  I've tried searching the learning stuff, the tip and tricks forum, and the regular design forum and there doesn't seem to be much guidance out there about direct modelling.  There is quite a bit of material about history modelling but DM seems to be pretty sparse.

C|

[cekuhnen]

@TrippyLighting

All other cad system because of their age let"s be fair also had a lot more time to add tools.

I think it is also fair to say that the project I demonstrated quite well sho the power of Fusion.

there is a reason why Apple uses alias for surfacing and not NX or GM uses Alias for car building and then later for manufacturing the Alias model will be brought into NX and finished 9and Not redone).

i bring this up because I think it is important to understand that for the type of work you might have to select the right type of application.

it is not only about the app or model but also about the process of designing and managing parts and designs.

[TrippyLighting]

@cekuhnen

My comment was not meant as criticism, merely as a description of the current state. Now that we all have worked with the timeline for a while, have grown somewhat more comfortable with it and the projects we trust Fusion 360 with get more complex we can simply imagine more advanced needs to manage that complexity. It will come in time!

[etfrench]

---

@kb9ydn wrote:

1)  Lets say I create a box that's 3x4x5.  I do some other stuff and then later find out that the box needs to 3.34243 x 4.3455345 x 5.  Editing the sketch doesn't work because it's no longer linked to the box in any way so to change its size I have to move one of the faces.  That's fine but in order to do that I need to know how far to move the face, which means measuring the current size of the box (because it's 3 days later and I don't remember what size the box was to start with) and subtracting that from the size it needs to be.  C|

---

On the other hand, changing the box in DM doesn't break anything in the model, whereas in TL it can and does break the model.  The more complex the model, the higher chance of breakage.

[cekuhnen]

@TrippyLighting

Oh hahah you missunderstood me - i did not understand your point in a negative way.

i was more trying to provide examples, to others who are confused with this discussion.

[PhilProcarioJr]

@cekuhnen

"Also because T-Splines is part of a feature in the timeline I can go into the TS body adjust it, leave TS Fusion regenerates the NURBS surfaces for

me and the rest of the design try is updated.

I have a hard time Phill to understand why you see a problem here."

And you will never understand because you work on "Simple" models. Don't take that statement the wrong way, what is meant by it is you work on models that have less then say 30 sketches. There are a lot of reasons I make that statement about not using the timeline and T-Splines.

1) The amount of overhead created when using the two together is huge. File sizes get very large very fast.

2) There are major issue when the design gets complex with meshes exploding or surfaces disappearing.

3) Fusion becomes very unstable fast in complex designs

4) You are severely limited in what you can create. My skeleton, Human Body, Topogun Head would be completely impossible to create if the timeline is use...I know I tried.

5) Because of the fact that all T-Spline work can be done in a separate file and the final model can easily be dropped right back into the timeline file it's best to keep them separate.

So yes if your goal is a simple surface, simple design using the timeline is fine. You guys don't get the PM's from people working on complex T-Spline work like I do asking why they can't do anything after a certain point. They hit the performance limits and capabilities in Fusion using Timeline and T-Splines.

Either way it doesn't matter, I base everything I say on these forums based on hard data I have collected with thousands of hours on very complex work in Fusion yet people still want to argue when they haven't even tested things past simple operations. I have said it before many times, I use the timeline, I like the timeline, but not for everything. The problem here is opinions get everyone worked up. I say when it's my opinion and when its based on facts, there is a difference....When I see someone create forms with T-Splines that are complex enough for them to show me different that's great and I will listen to them all day long, but no one has proven to me that my data is wrong so I stand by what I said.

In the end everyone can do whatever they want, but when they argue with me over things I know to be one way and they say they are not I just don't have time or interest in that drama.

If T-Splines and the timeline work for you that's great, but when a Fusion user listens to you and mixes the two on their client work and they keep crashing or lose all their work and can't get a job done I will let you explain it to them.

Cheers

[cekuhnen]

@PhilProcarioJr

Oh don't worry I am not taking it the wrong way - I stated already that we work on very different objects.

That is why I pointed out that based on what you want to work on you might have to adjust your workflow.

There is a reason why in car design they use Alias most times without the construction history. With a simple

object I do with some surfaces I can keep it. But with a car it becomes to complex to manage and to slow to computer.

So in Alias then switching into a DM mode is just more practical.

Over the weekend with an engineering student we build an automated lawn leaf vacuum. And with something like

that were you deal with many parts and components you need to make some modeling strategy decisions. For example

how much does everything have to be in a relationship. The ability of creating relationships can also later become way

to complex because then you have to jump into components and the right timeline position.

I guess that is still why when starting a project we sketch first make a model or a rough cad model to explore what we

might need to do before really getting serious with CAD.

And trust me - I am not against what you are saying. I can see the usefulness of a DM workflow like I can see in SE videos.

Thing is it is just now Fusion.

I am simply approaching this topic with years of experience that sofar I never found a CAD app that works the best.

In the end you still need to adjust your workflow somewhat.

And this is in the responsibility of the CAD software user to make the right decision.

[kb9ydn]

---

@etfrench wrote:

---

@kb9ydn wrote:

1)  Lets say I create a box that's 3x4x5.  I do some other stuff and then later find out that the box needs to 3.34243 x 4.3455345 x 5.  Editing the sketch doesn't work because it's no longer linked to the box in any way so to change its size I have to move one of the faces.  That's fine but in order to do that I need to know how far to move the face, which means measuring the current size of the box (because it's 3 days later and I don't remember what size the box was to start with) and subtracting that from the size it needs to be.  C|

---

On the other hand, changing the box in DM doesn't break anything in the model, whereas in TL it can and does break the model.  The more complex the model, the higher chance of breakage.

---

I think this is overstated a bit.  When you get good at history based modelling, breaking a model happens less often than you might think because you learn what things to avoid.  Also, you usually know ahead of time if a change is likely to break something or not, so when it does break you already know what to do to fix it.

I will also say though that the single time line and position tracking in Fusion add extra complication to keeping an error free model.  And Fusion seems less friendly in regards to fixing broken models than Solidworks for example. Solidworks has excellent tools for relinking missing references.  But Fusion will get better at this as it matures.

C|

[daniel_lyall]

---

@kb9ydn wrote:

---

@etfrench wrote:

---

@kb9ydn wrote:

1)  Lets say I create a box that's 3x4x5.  I do some other stuff and then later find out that the box needs to 3.34243 x 4.3455345 x 5.  Editing the sketch doesn't work because it's no longer linked to the box in any way so to change its size I have to move one of the faces.  That's fine but in order to do that I need to know how far to move the face, which means measuring the current size of the box (because it's 3 days later and I don't remember what size the box was to start with) and subtracting that from the size it needs to be.  C|

---

On the other hand, changing the box in DM doesn't break anything in the model, whereas in TL it can and does break the model.  The more complex the model, the higher chance of breakage.

---

I think this is overstated a bit.  When you get good at history based modelling, breaking a model happens less often than you might think because you learn what things to avoid.  Also, you usually know ahead of time if a change is likely to break something or not, so when it does break you already know what to do to fix it.

I will also say though that the single time line and position tracking in Fusion add extra complication to keeping an error free model.  And Fusion seems less friendly in regards to fixing broken models than Solidworks for example. Solidworks has excellent tools for relinking missing references.  But Fusion will get better at this as it matures.

C|

---

That straight out how it was done that's what parameters are for

[O.Tan]

 @jeff_strater, no problem. I know Fusion Devs has a lot going on right now 😄

As mentioned previously, I just wanted the devs to know that there's still many room for improvement in terms of advanced direct modelling and that there's still people who prefers to use DM (we have our reasons, so lets not get into another round of debate pls). IMHO, SolidEdge is the benchmark or gold standard when it comes to Hybrid Modelling (and Drawings) and it'll be wise for the devs to have a look at it.

Yes, every year brings some kind of improvements in Synchronous Tech and they even did a drastic UI change in 3D Constraints in ST8 which is a bold move from my point of view as everyone has gotten use to how 3D constraint works in past versions. Admittedly the new version is better. It's only with ST9 that they managed to do some kind of "ghost" so that when you make edits in DM environment, you're able to see how your actions will affect in TL environment.

---

"All designs were inherently parametric.  However, if you switched to DM mode, we just would not show the timeline.  That way, you could toggle back-and-forth between direct and parametric.  If you started in DM, then decided later you wanted the ability to edit an early sketch, just switch to parametric, and voila:  all the stuff you did in DM was recorded"

Interesting point you brought up, but now that you've mentioned it, I guess the issue lies at the very end where "all the stuff you did in DM was recorded" and how your concerns about the timeline getting ungodly long due to each little tweak will be recorded is perfectly valid.

In someways, I guess the user action in SE while in DM is recorded (explains why they're able to implement mirroring and makes sense as it's build on top of TL), but it's never shown to the user, even when the user switch back to TL, so any "feature editing" is locked into the respective environment, features created in DM only can be edited in DM and vice versa. 

For example, the user might have 10 various features in DM, when the user switch to TL, it'll be a blank canvas, user is able to see the model created in DM and create sketch referencing it (reference will break if the feature in DM is removed/modified but it'll not affect TL as it'll simply break the link, similar to how X-REF is implemented in Fusion where the user is able to "break the chain"), but can't access its features unless user goes back to DM, any features made in TL stays in TL and switching to DM will cause those features created in TL to disappear hence why in ST9 they implemented this "ghost" feature to solve this inherited problem. Does this makes sense? I don't have ST9 but I should be able to create a simple video to demonstrate what I meant with ST8. 

As a summary, it seems that SE isolates DM and TL features which prevents the issues/concerns that some of the Fusion dev had.

---

Admittedly for Fusion to implement what SE did is a lil different as SE only had to implement this on a part level whereas in Fusion, this needs to be implemented in a part-assembly level since it's all called components in Fusion. 

One idea I have but not sure if Fusion devs will like this as it diverts from the common workflow but it will be better for people who works in Machine Design as oppose to Industrial Design. I'll post it up here once I thought of how to phrase it properly. Brb.

[cekuhnen]

@O.Tan

I watched some more SE SyncEdit videos and I start getting the impression that this is not just DM as stated here.

SolidEdge totally records all the features and values. I would not be able to other wise update such edits.

In Fusion you would select the feature in the timeline and edit it.

In SE you select the object in the 3D view and manipulate it and this will update the feature for you.

But in both apps after the edit the design is refreshed.

It is quite nice to build something and then later add dimensions to the 3D model

I am curious what the app does when you draw on faces, does it glue them onto the face?

workflow wise it seems faster because I don't edit exit sketches - I just draw

The ability to further manipulate an ordered design is quite impressive.

the traditional feature approach generates the data / model and then to make it fit into the design or adjust later I can adjust

it by treating it like an object move parts around and when done the object is refit.

This is pretty genius.

So yeah you cannot do it in Fusion the same way.

You either go back in time and edit the feature correctly or you add features to manipulate it at the end.

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

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

I can for sure see the benefit of SyEd here since in the timeline when I edit something based on when something

was created it can trigger the visibility. This is currently a big problem in a utility project with an engineering student because

we blocked out the model and add details and change parts and you need to know dimensions and positions well to make things

line up when using timeline mode.

But that was also because we just jumped out of time issues right into CAD - instead of exploring the design first through other means

to get a good idea.

We solved this later by simply deciding were we start and how much we will only use relationships to model parts and what the relationships should only

be between parts. Then it became a lot easier to manage in TL mode.

[cekuhnen]

Another thing I want to point out is that I am curious about how many have a background in pure rhino surface modeling.

the problem in how y9u can only work in those programs is that if something has to be changed chances are very high that we have

to undo a lot of steps (fillets trimming etc) and rework that area.

i find it in this case a lot easier to study the timeline (selecting a face highlights the feature) and quickly understand how I could adjust the design.

[kb9ydn]

---

@cekuhnen wrote:

@O.Tan

It is quite nice to build something and then later add dimensions to the 3D model

I am curious what the app does when you draw on faces, does it glue them onto the face?

workflow wise it seems faster because I don't edit exit sketches - I just draw

The ability to further manipulate an ordered design is quite impressive.

the traditional feature approach generates the data / model and then to make it fit into the design or adjust later I can adjust

it by treating it like an object move parts around and when done the object is refit.

This is pretty genius.

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

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

---

Those videos are fantastic!  And indeed I have the same questions:  Why can't we just dimension the 3D model directly?  Why are we limited to drawing inside sketches instead of just in free space?  For DM mode these would make perfect sense since sketches are essentially disposable anyway.

C|

[cekuhnen]

Based on what I see is that it does the same as all do it just does not jump into sketch mode when drawing a line but makes the sketch on the face for you on the fly.

this is rather a UI workflow question than fusion not being able to do it

[kb9ydn]

---

@cekuhnen wrote:

Based on what I see is that it does the same as all do it just does not jump into sketch mode when drawing a line but makes the sketch on the face for you on the fly.

this is rather a UI workflow question than fusion not being able to do it

---

For the sketching thing I would mostly agree.  The question though is would it need to create a separate sketch for every line that is drawn or would it somehow group them?  Maybe just have a single 3D sketch that ALL drawn lines go into?  I don't know what the best way to do this would be.

As for dimensioning of models directly, I think that would be a little more difficult.  But man would it be awesome!

C|

[cekuhnen]

@kb9ydn well a sketch that does something to a face like slicing it should best be glued to it specifically when something changes to the face (orientation) the rest will follow.

otherwise free form sketches so one could work like in rhino or alias Woodman be highly welcome 

2d sketches can do a lot but sometimes their workflow is too labor intensive.

freeform sketches are often more flexible for organic shapes but then you cannot make use of constraints either lol ...

[O.Tan]

@cekuhnen, in SolidEdge:

Fusion DM = Synchronous

Fusion Timeline = Ordered (History)

All the images you attached is SolidEdge DM mode and you're probably right that it records all features and values (hence why I mentioned earlier that SolidEdge DM sits on top of their History)

But being DM, means I can reorder stuffs whenever I like without the model breaking, so even though it records, it doesn't break easily (unless you do some operations that remove or partially remove the feature (holes is a common culprit here))

I am curious what the app does when you draw on faces, does it glue them onto the face?

workflow wise it seems faster because I don't edit exit sketches - I just draw

Yes, it glues them onto the face, so if you resize or move the features, the dimension follows along. My guess in the background (hidden from the user) what the software did was to automatically create a sketch on the selected surface.

And if you look in the browser tree, there's this thing called PMI (aka 3D PMI), that's where all the dimensions you see here are stored. SolidEdge allows you to lock the said dimension so let say you have a hole that is 20mm away from the edge that you would like it to be fixed, you can do it and whenever you adjust the position of the hole, it'll maintain that 20mm away from the edge. 

Another piece of information worth mentioning is in SE, they also implemented things like 3D constraints and "live rules" which allows the operations that you see in the video to work as user intended. It does however has a learning curve which my guess is why many people are not familiar or aware that such advanced direct modelling existed at the price point of Inventor or SolidWorks (this is the main reason why I got the company I'm working with to go for SolidEdge rather then Inventor or SolidWorks as to me, both of them has similar approach to 3D modelling).

Yup, the engineers at Siemens really implemented something impressive with their take on the whole DM and TL, something I wished Fusion would do but from my understanding, there's no plans to make DM parametrically smarter at the moment. 

But that was also because we just jumped out of time issues right into CAD - instead of exploring the design first through other means

to get a good idea.

We solved this later by simply deciding were we start and how much we will only use relationships to model parts and what the relationships should only

be between parts. Then it became a lot easier to manage in TL mode.

That's true, TL will work well if it's planned but reality wise, in some industries. Detailed planning of relationships and how the part should look/be built doesn't really work cause the environment requires iterations to be made and unforeseen problems popping up that requires quick changes. The last thing the designer/engineer would want is for the CAD software to throw errors and having to spend time "fixing" those errors.

Another thing I want to point out is that I am curious about how many have a background in pure rhino surface modeling.

I don't but I did try learning Blender and 3DS Max before and it's really different coming from a mechanical 3D CAD background. I guess this is why our opinion when it comes to things like this differs greatly. 

[O.Tan]

Those videos are fantastic!  And indeed I have the same questions:  Why can't we just dimension the 3D model directly?  Why are we limited to drawing inside sketches instead of just in free space?  For DM mode these would make perfect sense since sketches are essentially disposable anyway.

If you're referring to why we can't do it in Fusion, I guess cause additional features needs to be implemented and the last I check, AD is not interested to improve DM at the moment.

Cause you have to take account that the software needs to be able to:

1. Take account that if the user moves the surface, the dimension shown needs to be updated

2. If the dimension is edited to a different size, it needs to notify the user, which direction the user wants the change to happen (in the case of a box, do you want it to expand symmetrically, to the left or to the right)

So just the 2 things alone mentioned would require the team to do quite a number of changes on the software and UI side.

For the sketching thing I would mostly agree.  The question though is would it need to create a separate sketch for every line that is drawn or would it somehow group them?  Maybe just have a single 3D sketch that ALL drawn lines go into?  I don't know what the best way to do this would be.

As for dimensioning of models directly, I think that would be a little more difficult.  But man would it be awesome!

In SE, if I recall correctly, simply attached Dimension directly will place it in the PMI folder, and if you want to "sketch", you can straight away select the sketch tools you want, select the plane (F2) that you want to create your sketch on and just draw. Your sketches will be placed inside a Sketch folder and the software is smart enough to know that if you were to add additional sketch to that same surface, it'll update the sketch in the sketch folder with your additional sketch.

In Fusion, since you have to create a sketch, you can always edit the sketch or if you're really lazy, reuse the sketch on other similar areas by moving the sketch. Haha

[O.Tan]

@jeff_strater, as mentioned this is my 3 takes on how Fusion can improve the DM and or TL experience. Though as I don't know how exactly Fusion DM and TL is programmed, I'm limited to my observations and use experience. Sorry that it took very long to reply cause I was busy with work and as you can see, I put lots of thoughts behind this. If you feel this post should be moved to a new thread, please do so.

Option 1: Hybrid Modelling 

As mentioned in this thread earlier, SolidEdge is the gold standard when it comes to hybrid modelling or DM implementation.

What does this mean is DM sits on top of TL where DM is not able to see the features created in TL but TL is able to see DM features and is able to edit it as well. Below is some pictures explaining this:

This 2 features is created in DM, I'm now switching to TL

In TL, I added a hole as highlighted in green

When I go back to DM, notice the hole is missing and hover my mouse over the hole in Ordered (TL) shows an orange outline

Back in TL, I'm able to resize the feature created in DM even though it wasn't created in TL

In the case of Fusion, it’s slightly difficult to implement since there’s no part and assembly files. Traditionally, assembly actions like Move isn’t recorded or at least made visible to the user and TL is only available at Part file which makes it easier to manage since if a part has a messed up TL, the user can just either fix it or recreate the part, but in Fusion since the assembly is also involved, that process might not work as the user might break the assembly as well (or create lots of warning icons cause dependencies is broken)

Pros: Truly revolutionary, best of both worlds

Cons: Difficult to implement especially since Fusion doesn’t separate part and assembly files.

---

Option 2: Different Workflow (X-REF)

Using the current X-REF built into Fusion, the user is assumed to manually manage their own part and assemblies using components. For the sake of clarity, a part = a component that only has bodies and assembly is a component that has a minimum of 1 component (means it can both have body and component)

To better explain my points, please see the photos and description below it:

So upon opening an assembly, the user will be greeted with a bunch of linked parts. Since all these parts are linked, each has their own TL or if the user chooses so, DM, it depends on the design intent of the part. And at the assembly level, it’s up to the user to either use DM or TL. This means, TL or DM is constrained at its own component.

At the assembly level, user is not able to access the part TL unless he/she activates/edit-in-place the part. This allows scenarios like: parts is done in TL but assembly is done in DM or parts is done in DM but assembly is done in TL. Or better yet, a mixture of both depending on the design requirement. And guess we need to add an option like "create external component in-place" where it'll first do this as a part, and if the user creates more components in that, it'll become an assembly. 

To make it easier for existing user, a new tool must be created to allow current components in a design file to be separated to part and assembly files

The upside of such method is it’ll allow the data panel to have a folder that’ll automatically keep track of individual parts iterations (each part is now individual version tracked instead of the whole assembly) and collaboration with different users (machinist, SIM and etc) is easier as it’s now possible to add restrictions to what users is able to access (e.g: Machinist only able to access individual parts and SIM only able to access whole assembly).

Pros: Easier to implement, better for collaboration cause of individual parts, improved part change versioning

Cons: Creates a different workflow which will mean added complexity to new users and tutorials has to be created on workflow recommendations based on the type of Fusion user, at the moment, using X-REF in Fusion leads to poorer FPS.

---

Option 3: Improved DM (Hidden Timeline)

Timeline is added to DM, but it’s hidden from the user

Since there’s now a hidden timeline, it’s possible to implement features like editable patterns and mirrors. Some rules probably have to be made on the hidden TL to make this work, mainly that sketches will be placed at the front, followed by bodies, then components and then pattern and mirror tool. Regardless of when the mirror tool is invoked or sketch is done, it’ll place it at the respective assigned areas in this hidden TL.

Probably to aid in making implementation of this easier, sub-assemblies or parts will not be tracked as it’ll live in its own hidden TL, in the parent component, it’ll only list part or assemblies in the TL.

Dynamic pattern(s) or mirror(s) is to allow the user to edit back patterns and for mirrors to be implemented in DM, whenever a pattern or mirror tool is edited, it'll automatically move to to back of the timeline (so it'll have the latest components change and etc).

By compartmentalising features into segments in the hidden timeline, it'll allow components to be reordered without breaking things. 

Pros: DM will be better then what’s today with the addition of further intelligence and reduce the need for a user to switch to TL

Cons: Switching to TL will still likely be the same as Fusion today, means it’ll register everything as a single BaseFeature/Group.

---

Personally I like Option 2 the most, as it balances how Fusion works now and allows those additional benefits (mainly the smart folders and tracking individual parts) but at the expense of a new workflow is created. And I feel it's easier to implement then Option 1, though if Option 1 managed to be implemented, it'll really open up Fusion modelling capabilities. Option 3 is essentially improving DM by adding more intelligence to it so the user doesn't need to switch to TL.

Thank you for spending your time reading this long post and if there's any questions or clarifications, or if I didn't take account of a few scenarios, please ask away and I'm open to be contacted, be it via emails, forum or phone call.

[daniel_lyall]

Man @O.Tan that's some good idea's option 2 looks good to me.