Model size 2

[edgray]

I would like to get a feeling for the relative "size" of the parts we build here. I am hoping people will be willing to share what they consider to be an average size part and what they consider to be a large part.

I am talking about feature count.

I try to be maximally efficient with my features. One of the first things I usually have to do after inheriting a part from a collegue is to remove vast swaths of features and replace them with a few more thoroughly thought out features.

Thanks in advance.

 

[looslib]

One of the items on my desk is a list of Commandments for modeling in another CAD package. Your comment on replacing lots of features with other features that lowers the feature count struck me. So, here are a couple of the commandments.

Always start with a thought out plan for constructing your model
Thou shall keep sketches simple and not create complex features with a single sketch
Thou shall avoid creating rounds and chamfers in sketcher, as they should be added as separate features, when possible
Thou shall add chamfers, drafts, and rounds at the end of the model and not create children of chamfers, drafts, and rounds
Thou shall save early and often


"Wildfires are dangerous, hard to control, and economically catastrophic."

Ben Loosli

 

[edgray]

Sounds good.

How many?

 

[jerry1423]

How many of what?

 

[edgray]

I am just trying to get an idea of the number of features in typical individual parts.

 

[jerry1423]

There is no way to come up with a number . . . it can be anywhere from 1 to 1000+

 

[edgray]

Wow I didn't know I was asking an odd question.

I guess it comes down to the "simplicity" of our product line. We make three different types of parts, but each one is diabolically complicated.

I always use copier companies as the direct opposite of our product. "Simple" individual parts but "large" assemblies. In our case we have simple assemblies, 3-20 components, and complicated individual parts 500-3500 features.

I just wanted to see how other companies fit in between these two "extremes"

I work in the auto industry so I also have some expeience with the complicated-complicated CAD situation.

 

[cowski]

I worked at a company that made household consumer products, lots of injection molded parts.
[ul]
[li]small: < ~200 features[/li]
[li]medium: ~200 - ~1000 features[/li]
[li]large: > ~1000 features[/li]
[/ul]

Usually a project would have 2 or 3 'large' parts, the majority would be medium parts, and a few assorted small parts. I have worked on a number of, what I would consider, really large parts that were 5000+ features.

My goal was never to minimize the feature tree, after all you could export a parasolid and claim you did an entire complicated part in only 1 feature; but rather to model in a logical flow that minimizes unnecessary features and dependencies. If a large part has been worked on by several different people, you can almost assuredly eliminate dozens of features (offsets on top of offsets when simply changing the extrude distance would have given the desired result, pruning 'dead branches' in the model tree, etc).

 

[edgray]

Thanks cowski, that is exactly the kind of responce I was looking for.

My main reason for trying to reduce the feature count is that our parts re in continous revision, and as you know it is impossible to eliminate feature failure, so the less feature I have to fix the faster I can complete the revision. However I think I have gone overboard with my sketches at times. I am going to have to review this.

The thing I am most proud of in my parts is the ability to make early (inthe feature tree)chages in the parts with minimal disruption down stream. Remember I make very few parts with less than 1000 features. I spend most of my time fixing failed features. My tone is kind of snotty sorry.

 

[cowski]

I was often tasked with the large parts, so I know exactly what you are going through. The key is to minimize dependencies, my first boss drilled that into us and is reflected in looslib's list as:

Thou shall add chamfers, drafts, and rounds at the end of the model and not create children of chamfers, drafts, and rounds

I heard about "horizontal modeling" a number of years later and as best I can tell, that became the term for minimizing dependencies.

http://www.designnews.com/document.asp?doc_id=226495

 

[edgray]

That is interesting.

My rule is to maximize the dependencies! But, it must be done in a thoughtful way. I am not talking about fillets and chamfers. for me fillets (we only use chamfers rarely) are to be placed with the Par feature they belogng with. there for when I am working on a mount boss say around feature 850 out of 1800, I want the fillets right there so I don't have to go to the end of the part to check if the fillts have been added or not. Fillets can also have a signifcant positive effect on down stream features in certain situations, so I don't like to generalize too much.

I try to make sure that the engineering relationships are preserved. My sketches are regularly dimensioned to or constrained to other sketches. I try not to dimension or constrain to edges, that is asking for trouble. Since my parts are thin walled plastic parts I make extensive use of the shell feature, and this leads to creating a lot of reference curves in the sketches that indicate where the other side will be. like I said earlier I can go a bit crazy in my sketches.

 

[cadrebel]

Just looking for input....
after the introduction of syncronous modeling in NX6, why do designers still keep all these features?
I have heard that they keep them for easy changes but when I read this thread about features being built on top of features and then being cleaned up, it seems that there would be a more effient way.
I have found that removing parameters makes the model very clean and easy for the next designer to navigate through the model.
I have also found that removing parameters makes the model much smaller. When I first started using NX I kept my parameters, I made changes to a block that were extensive, then I tried to move the block. The first two times I shut down the machine because I thought it locked up on me. The third time I just let it go and I timed it. It took ten minutes to move the block! When I removed the parameters it was almost instant.
I have been told by many that I cannot work this way, that it is against the commandments that that I should be hung from the highest tree, but I do not know what I am missing here.
I love synchronous modeling.... it is why I bought NX

 

[edgray]

Cadrebel, I'll take it under advisment.

We definately have issues with unintended consequences. This is an argument for parameter removal.

However, my problem with this is that we have extensive interpart relations that help to keep modeling errors from causing assembly disasters. For example: Rotating clearances that are relative to the size of the rotating part, are coded into the expressions. We do have troubles with our novices getting tangled up in the expressions but we have a low incidence of, part too big to fit in hole.

Any relationship that the engineers deem to be critical to the performance of the part is coded into the expressions, along with containment for the results of the variations in the inputs.

But back to my original question what is the range of the feature count in your parts before you "dumb lump" it?

 

[jerry1423]

Look into "History Free Mode" in NX, you may like that over the feature thing.
I agree with what you said about syncronous modeling, but in my opion it really depends on the model c omplexity and the application on what is the best way to go.

 

[looslib]

How many times do you have to go back and change a feature by few thousandths of an inch to get it just right? If you still have the parameters, you change one value and the model updates. As with any CAD system. the real advantage is not in creating a new part design, but in editing that part design later for whatever reason. Synchronous modeling allows this, but I can see where too many synchronous changes would render the original parameters useless.

Has Siemen's, or anyone, published a white paper comparing parametric editing versus synchronous?


"Wildfires are dangerous, hard to control, and economically catastrophic."

Ben Loosli

 

[cadrebel]

ED - I do not have a number for you "before" because I constantly remove parameters as I go. I typically do not even use sketches unless I have some really funky shape. When I get a body like I want, I destroy the evidence. (any jobs in accounting out there???)
Jerry - I agree, it really depends on what you are making. I make molds this way and I am pretty happy. There are a few pits in the trail but for the most part it works well for me. I don't use the history free mode because I like a little history for a short while but it also is a good option for file size.

 

[edgray]

If it works it works!

 

[bruwel]

Rebel, once you destroy the evidence, how does the next poor schmuck figure out what you have done to revise it? I ask because I am usually the next schmuck that gets to move a hole and add a couple small enhancements to a wonderfully compact model.

 

[cadrebel]

bruwel - tell me more about why you need to know what he has done. not trying to be smart here, it is a question that comes up a lot for modeling this way. if what was done is done, and you come along and have another change, the synchronous modeling allows you to make that change regardless of what was done in the past.
we keep design change forms in our document control system that explicitly will tell us what was done in the past and also why.
I think that going through the whole history tree to find the line that was changed in the beginning is cumbersome and you end up just doing what you want anyway because as you dig into the tree you find that some other "schmuck" didn't want to look either and they just added another level to the current one which is why guys are going in and cleaning up the whole tree eventually.
what do you think?

 

[edgray]

Cadrebel, I get your point. That is why I suggested that I would take it under advisment.

The reason that I want to know the hows and whys is that making a change to some features that are early in the tree will have effects further down the line. The way we model it makes it easier if can discover the relationships. I do not hink that it would be possible to make the kind of changes that we do solely with direct modeling, but I am going to keep it in mind. Expecially when I get to see the advances in NX7.5. I am still on NX5 for a few more weeks.

With the parameterizatio that I have incorporated into our parts we can some pretty elaborate what if scenarios while preserving our engineering guide lines.