How to clean up data mess - circular dependency between sketches?

[kristianY3VSG]

Hi,

I´m back again, as another problem occured ;-(

After inserting a new sketch "NEW" just before an older one ("OLD") in the timeline and projecting some elements from NEW to OLD, dimensioning anything in NEW announces a circular dependency. When I delete the projection, everything works fine, but I need this projection, as I don´t want to mix two different things into one sketch.

I´ve  searched here for some information about circular dependencies, but the only common point is "data mess". This might be my problem too, as this design originally had insert-derived sketches, but after occuring massive problems of various types in other designs using this feature, I deleted the insert-derived sketches from this design, hoping, the Autodesk team will find the bug. 

Another source of the actual problem might be, that the the blue elements in OLD were copy-pasted from another sketch before (cleaning up unneccessary sketches). 

As I´m struggling with weird beavior of Fusion since allmost 2 weeks, spending more time on creating screencasts and writing  postings to the forum then for useful work, I´ll ask now:

Isn´t there some "Fix the Design" function, which would  clean up accidental mess in the internal data, following the timeline elements step by step and recreating all the data from scratch? If not, so I´d recommend to create such a function ASAP 😉 I really don´t want to recreate all my weird behaving designs from scratch again!!! The more, as I can´t be sure, the problems will not occure again 😞

Thanks for any supporting infos...

Kristian

PS: Inserting screencast has failed again :-((( 

Screencast will be displayed here after you click Post.

[kristianY3VSG]

Grundhog Screencast Day :-((

[jeff_strater]

Thanks for posting this, @kristianY3VSG , I'll take a look at it.  We try hard to prevent ever getting into a circular dependency case, so a reproducible way to do it is good - we should be able to block whatever you did to get into that state.  I may come back and ask more questions about how you got into this state, later, probably not today.

[kristianY3VSG]

@jeff_strater , thanks for investigating this problem. In case of interest I can send you selected older versions of this design, showing the main steps. Maybe they could help you to track back the history, where this problem began - in my opinion with the insert-derived sketch. Later a forgotten projection appeared in this design, and now the circular dependency.

[TrippyLighting]

@kristianY3VSG I've been following some of your posts here on this forum but had not taken a chance to take a deeper look into any of the designs.

May I ask what prior experience with CAD software you've had ?

You seem to get into workflow issues that look like these were acquired using another CAD software. I've done this a number of times and while most have similar concepts the difficulty is usually where the overlap stops. That's not  different in Fusion 360.

I am wondering if the project you have started is a bit too complex for the first steps in Fusion 360 as I can see your frustration increasing. My recommendation would be to prepare to start over from scratch a few times to get a good workflow down.

Looking at your sketch Umriss Seitenträger you have 10 fillets in that sketch. All of those could have been applied to the solid after extruding. Not only does this substantially reduce the time for sketching but due to the reduced number of sketch elements and constraints it is much faster to compute for Fusion 360. Most importantly, however, it behaves much more stable.

In general many users rely too much on projections into sketches.

Following the BORN technique is a little bit more work, but also creates much more sable designs with fewer unneeded inter dependencies  between geometries.

[kristianY3VSG]

Hi @TrippyLighting , thanks for your interest and suggestions.

I started with Fusion in July 2017, when I realized, that I can use 3D-printing for my project - a live-steam locomotive model 1:8.  Before that moment I just draw some 2D metal sheet parts in Librecad, this was my first CAD experience. Since that time I created designs for wheels (including molds for wax patterns for steel castings) and some other test parts (directly 3D-printed patterns). During that time I learned quite much, what I can do in Fusion and what Fusion doesn´t like, and also, how to optimize my designs = minimize the work I have to do. And of course, I use the BORN technique. I have been working as a software designer & programmer for more than 30 years, so I´m quite familiar with computers, and my logical thinking is not the worst 😉

Now, having succeeded with test castings, the moment has come to design the whole engine. Regarding my skills and experience I have with Fusion so far (a bit too high probability of weird reactions, sorry), I don´t want to push dozens of parts, each representing dozens of timeline steps, in one single design file. If a one-part-design crashes (like  those beeing discussed now), I am willing to recreate this particular design from scratch again, BUT NOT A MACHINE DESIGN REPRESENTING MANY MONTHS OF WORK!!!

As I didn´t found any other way, how to share basic dimensions between many designs automatically (yes, I could write them in an excel sheet and update them manually, or Fusion could provide parameter tables similar to sketches, able to be inserted into other designs), I created a hierarchic structure of single part designs and used insert-derived sketches to share basic point and lines. At that moment the problems now beeing discussed began. Now I just can hope, Autodesk will find and eliminate the bug(s) quite soon.

Why I don´t use fix dimensions: It´s not possible to downscale such an engine strictly by scale, as some parts would become to small, screwdrivers couldn´t reach some screws etc. This problem increases, the smaller the parts are. When I realize after 3 months of constructing, that I need 1,5 milimeters more space for some part on level -4, I´d like to create more space fot it changing some dimension at some higher level, which should automatically be passed through all levels to this part. Of course, I should start with the smallest parts and proceed to bigger and bigger parts (higher levels), but as I´m not an experienced machinery engineer, I´m not able to work in this way - I have to construct downwards, sorry. That´s why I tried to find a technique, how to achieve the flexibility needed.

Of course, I know, that using body fillets is better. So far I prefered sketch arcs, as I´m using original drawings (you may have noticed the canvases). Regarding the last experiences now I´ll think a bit, how to use solid fillets without getting "lost in dimensions".

Ad BORN: In the assumed hierarchic structure of designs (masterplan, main parts, smaller parts etc.), I want to project needed geometries downwards from level to level. To backckeck the whole design, I have to combine all these singular parts to an overall assembly design (using insert or insert-derive). Yes, I could design each part around the system origin and move it to its real position in the assembly file, but projecting e.g. holes downward from part to part is easier, when I place them at their final positions just from beginning. That´s why the backframe (Hinterrahmen) is placed far away from origin (the future projected reference points beeing represented by fix points so far).

Please excuse my probably too long post, but I just wanted to explain my position. I´d be happy to learn some better way, how to structure such a project.

Kristian

[jeff_strater]

thanks for your patience, @kristianY3VSG .  I spent a bit of time with this design today.  There is clearly a bug here.  Here's what I found:

• sketch "Adamsachse Radius" is the second-to-last sketch in the timeline before the timeline
• sketch "Grundriss allgemein" is the last sketch in the timeline before the timeline
• the earlier sketch has a later timestamp than the later sketch.  This implies that the sketches have been reordered at some point.  This, of course, should be fine.  If there is a dependency between two timeline items, you should be blocked from reordering a dependent item before its predecessor.  I suspect that is the root cause here.
• however, I cannot figure out what this dependency is.  It is not a projection or any dependency that I can find
• if you can shed any light on that, it might help.
• we will continue to investigate this, from the design and the screencast - it really helps that you provided both of those.

More info as we find it.

[kristianY3VSG]

I´m not sure, @jeff_strater , what you exactly mean by "shed light on that", but I´ll try:

• "Adamsachse Radius" was inserted later, when "Grundriss allgemein" already existed.
• Dimensioning in "AR" works fine (screencast at 0:40).
• Then I project some necessary elements to "GA" (1:00).
• Returning to "AR" after this, dimensioning leads to circular dependency (1:20)
• After deleting the projected elements from "GA", dimensioning in "AR" works fine again.

Maybe all this is influenced by the fact, that the blue elements in "GA" were copy-pasted before from another sketch "Vorderkasten Boden" originally situated right from the actual history marker position (see V6 before and V7 after copy-paste), later deleted, as there was no need to have these profiles in separate sketches (V8 - last version before inserting "AR"). 

Originally there was an insert-derived sketch in this design. As a weird projection error occured, I deleted that sketch, as the insert-derive function is suspected to be problematic. 

I hope, this can help you.

Kristian

[kristianY3VSG]

Could this be a clue? I work on a notebook, so I normally don´t quit Fusion at all and have several designs opened at the same time. Can some unexpected interferences in the data space of Fusion between simultaneously opened designs using insert-derive be the root of the discussed problems?

[jeff_strater]

Thanks for your patience, @kristianY3VSG .  We are making some progress, but still have no clear answer.  A couple of points worth mentioning in response to your post:

• I don't think that having several designs open at the same time can be the cause - most people work that way, and I have not seen any reports of problems from that workflow
• You may be right that the Derive/Project workflow could be a problem, but it's not clear to me yet how it relates to this design.
• you may be on to something here:  "Maybe all this is influenced by the fact, that the blue elements in "GA" were copy-pasted before from another sketch "Vorderkasten Boden" originally situated right from the actual history marker position (see V6 before and V7 after copy-paste), later deleted, as there was no need to have these profiles in separate sketches (V8 - last version before inserting "AR"). "  Copy/Paste could be a source of problems

Here is what we found out today:

• the problem is indeed only related to the two sketches ("GA" and "AR").  You can delete the entire rest of the design and the problem still shows up.  So, we are very interested in the history of those two sketches
• you say '"Adamsachse Radius" was inserted later, when "Grundriss allgemein" already existed.'  That is good information, and also agrees with what I see in the design.  "AR" has a much later timestamp than does "GA".
• There is already a dependency, however from GA -> AR.  From our debugging, it looks as if when AR was created, GA was selected as the sketch plane.  This is perfectly valid.
• But, and this is the critical part:  At some point, AR was reordered before GA in the timeline.  This should not be possible.  If you have any recollection about how this happened, we would definitely like to know.
• Because AR is now before GA, Fusion lets you project from AR into GA.  This is what causes the cycle, it doesn't have anything to do with the dimensioning.  If you just edit GA, project from AR, and exit the sketch, then do Compute All, the cycle will show up.  The dimensioning just triggers a compute which returns the cycle error.

Thanks for sharing your designs - debugging would be impossible without those.

[kristianY3VSG]

@jeff_strater  I´m very sorry, but I don´t remember, how I created AR 😞 As I created it for elements supposed to be projected to GA (at that time named "Vorderkasten Boden" in V8, forgot to mention this, sorry), it´s not likely, that I created it  right from GA (only by mistake, may have happened). And creating it left from GA shouldn´t let me select the GA-plane (and I´m not used to select sketch planes for new sketches). I played around with V8 now, creating AR in vyrious ways, but couldn´t find any way to replicate that error, even creating AR right from GA, projecting something from GA, then deleting the projection link and moving AR to the left of GA. 

Ad insert-derive: Originally I worked with "Hinterrahmen MP" containing two insert-derived sketches from Masterplan. The reason was, that a mysterious project warning occured, and after having problems with "Sidebar" ("Rahmenwange") I decided  to substitute the inserted sketches by fixpoints or dimensions, untill the Sidebar problems will be solved, thus creating Hinterrahmen fixpoints. My idea is, that even the new file may still be contaminated by some data mess originated at that insert-derive. May the Master-Slave laboratory experiment could give some clue?

Thanks for your effort!