CAD Forum

Sometimes with all the chatter about the various top-down modeling methods and very little about plain old bottom-up modeling I wonder who else does no or very little top-down modeling.

Trying to be specific I'll add a little list off the top of my head; I don't know that every one of the lines below are really top-down vs bottom-up type things, but they tend to get drawn in sometimes.
- Piece parts in part files, weldments and assemblies in assembly files.
- Multibody modeling only used for ref geometry or tool bodies; one part number per file.
- No Master Models
- No Skeleton Sketch Parts
- No parts referencing other parts through a parent assembly.
edit: - Primary mating method is part geometry to part geometry (on rare occasion ref geometry if face ids likely to be change such as imported customer/vendor model); Mating origin to origin or coordinate systems is not used.

Who else models in this way in day-to-day practice?

Thank you.

Much of my work doesn't involve external references of any kind, but it does occasionally, such as letting the shape of concrete drive the rebar shapes and sizes. When I do that I usually edit the rebar Parts inside the Assembly instead of having a master Part.

I do often use multi-body Parts. When I do they stay as multi-body Parts. I never save the bodies out as separate files.

My philosophy on this is pretty simple. There is one major question that divides the two possibilities: will this part be used in multiple contexts or in only one context?

If a part is used in multiple contexts (fasteners and standard bearings are good examples) it MUST be done bottom up. Otherwise changes in one context will affect all others. Other examples are company-wide standard parts that are manufactured to spec instead of purchased off-the-shelf. In our case, we have common shared parts such as motor brackets, towing components, and wear bushings.

On the other hand, if a part is used in only one context it makes sense to use top-down. Sheet metal parts that are common to only one model or product are good examples. As long as it's not used in another assembly, top-down works great.

I do not use top-down/skeleton sketch method. I use bottom-up as you describe it. I make mostly machined parts and use a lot of "standard" parts such as something you might download from McMaster-Carr. Our assemblies represent how the parts actually go together so this also means (gasp) that we mate faces to faces and holes to holes rather than primary planes to primary planes. I know to some this is heresy. John Stoltzfus would rail on me for my methods, but the simple fact is this works better for us.

Top-down has its place, just as bottom-up has its place. The key is to use what works best for your situation.

Most of my work is bottom up.

Only on the very large assemblies do I work top down. And then, ONLY if I have a really good feel for what the final product will look like before I start.

95% of my modeling is bottom up. Even when I use a sketch to transfer information across sub-assemblies/parts, the vast majority of the work is of the bottom up design.

My life may be simpler than most because I don't use weldments or multibody parts.

Here is an example of one of my machines. You can see several sketches that are part of my SSP. The only thing that is driven off of the sketch are the belts. The rest of the information is used to allow me to work in the context of a smaller sub-assembly, while still knowing where features are located in the other assemblies.

DennisD wrote: ↑Wed Dec 08, 2021 11:32 am I do not use top-down/skeleton sketch method. I use bottom-up as you describe it. I make mostly machined parts and use a lot of "standard" parts such as something you might download from McMaster-Carr. Our assemblies represent how the parts actually go together so this also means (gasp) that we mate faces to faces and holes to holes rather than primary planes to primary planes. I know to some this is heresy. John Stoltzfus would rail on me for my methods, but the simple fact is this works better for us.

Top-down has its place, just as bottom-up has its place. The key is to use what works best for your situation.

Ah yes, I forgot about assembly mating/relationship method gets tangled into the top-down or bottom-up method. I'll update my OP to have
- Primary mating method is part geometry to part geometry (on rare occasion ref geometry if face ids likely to be change such as imported customer/vendor model); Mating origin to origin or coordinate systems is not used.

SPerman wrote: ↑Wed Dec 08, 2021 11:41 am Here is an example of one of my machines. You can see several sketches that are part of my SSP. The only thing that is driven off of the sketch are the belts. The rest of the information is used to allow me to work in the context of a smaller sub-assembly, while still knowing where features are located in the other assemblies.

image.png

So the sketches provide the profile for the belt loft path? Makes sense.

But the other sketches/elements are not driving the things they represent (space claim)? Are those sketch elements being driven by the parts they represent? If not, do you update both when one moves or changes or is it just a rough guide for reference and not to be taken as exact?

I don't understand the need to rigidly define "This" or "That" modeling methods. I chose and pick what works for me. In some cases, that means mating origins to origins. In other cases, it means mating construction geometry to construction geometry (typically planes or axis) and in some cases it means face to face and hole to hole.

Learn the pros and cons of different methods and use what works best in a given situation. Stop worrying about if it follows "X" procedure or violates some arbitrary rule. Those only limit your potential to get the job done efficiently.

SPerman wrote: ↑Wed Dec 08, 2021 11:53 am I don't understand the need to rigidly define "This" or "That" modeling methods. I chose and pick what works for me. In some cases, that means mating origins to origins. In other cases, it means mating construction geometry to construction geometry (typically planes or axis) and in some cases it means face to face and hole to hole.

Learn the pros and cons of different methods and use what works best in a given situation. Stop worrying about if it follows "X" procedure or violates some arbitrary rule. Those only limit your potential to get the job done efficiently.

It's not so much dogma as it is best practice. I'm not trying to shoehorn anyone into my philosophy, CAD or otherwise.

If anyone really thinks best practices are all nailed down, why is there still such a debate whenever modeling methodology comes up? Opinions vary and I'm fine with that, it's just that I'm usually not shy about expressing my own.

SPerman wrote: ↑Wed Dec 08, 2021 11:53 am I don't understand the need to rigidly define "This" or "That" modeling methods. I chose and pick what works for me. In some cases, that means mating origins to origins. In other cases, it means mating construction geometry to construction geometry (typically planes or axis) and in some cases it means face to face and hole to hole.

Learn the pros and cons of different methods and use what works best in a given situation. Stop worrying about if it follows "X" procedure or violates some arbitrary rule. Those only limit your potential to get the job done efficiently.

Edit: like Mike said, I'm not trying to say top-down is bad or that list I put in the first post are best practices. I'm certainly not saying anyone else should CAD the way we do where I work. Hopefully this doesn't become another debate thread.

Well, that's kinda what we have and it's difficult to get new help up to speed and they get very frustrated because it takes a long time to learn all the "what worked best for whoever modeled it five or ten years ago". When everyone has their own way to do things that really could be unified it's hard to maintain the dataset. People make mistakes when working on existing files or new products that use existing products.

Often times the quickest or "best" way to create the model at first is bad as it costs many hours of fixing down the life of the model and offspring models. Sometimes its a while before the problem manifests its symptoms, by then the file can be used in several assemblies and copied to a new part number or two. When there are no guidelines people will follow whatever method, they last saw for something similar.

On the other hand, it is as you mention about rules limit potential and if the rules are full of exceptions, then what's the point of having rules? We deal with that question daily. So its interesting and helpful to hear how others do it.

mike miller wrote: ↑Wed Dec 08, 2021 12:02 pm It's not so much dogma as it is best practice. I'm not trying to shoehorn anyone into my philosophy, CAD or otherwise.

If anyone really thinks best practices are all nailed down, why is there still such a debate whenever modeling methodology comes up? Opinions vary and I'm fine with that, it's just that I'm usually not shy about expressing my own.

If we were all doing the same job, in the same industry, there would still probably be debate over the best method. With the wide range of uses and industries there is no single "best practice" to cover all CAD.

When I'm modeling concrete some of it is often top down. When I'm modeling a guardrail it's all bottom up.

For me, it depends greatly on what I'm doing. Most of my work is master model / multibody to start, fanning out to single body mate-at-the-origin assemblies, and usually small assemblies with relatively complex parts.

I use bottom up process only for the simplest parts and assemblies. If I'm putting off-the-shelf parts together into an assembly, or simple stock shapes, or simple machined parts.

I don't really do the same thing twice, or even similar stuff more than once.

bnemec wrote: ↑Wed Dec 08, 2021 11:51 am So the sketches provide the profile for the belt loft path? Makes sense.

But the other sketches/elements are not driving the things they represent (space claim)? Are those sketch elements being driven by the parts they represent? If not, do you update both when one moves or changes or is it just a rough guide for reference and not to be taken as exact?

In this case it was a bit of an evolution. The envelope of the machine was predetermined, so I began with the front sketch to determine the position of the pulleys such that they fit available belt sizes and could be properly tensioned. That drove some major component placement on the front and rear plates, but once all of that was determined the relationships were broken so that they were no longer driving the position. At the end of the design, I try not to have the sketch drive anything. Parts get reused, and local changes may need to be made and I don't want any downstream unintended side effects. To me SSP, is a part of the journey, but not a destination.

I should add that I have the benefit of being a one man band. When I try something out and it doesn't work, no one is affected but me. I can see the need for more defined processes in a bigger work environment. (I would definitely be cussing someone helping me out that was doing some of the stupid things I've tried. Let's not talk about the "fix all" phase I went through.)

SPerman wrote: ↑Wed Dec 08, 2021 1:05 pm I should add that I have the benefit of being a one man band. When I try something out and it doesn't work, no one is affected but me. I can see the need for more defined processes in a bigger work environment. (I would definitely be cussing someone helping me out that was doing some of the stupid things I've tried. Let's not talk about the "fix all" phase I went through.)

I agree. It's always awkward when I open an old part and get upset at the guy who modelled it.................then I realize he looks just like me.

Thanks everyone. I could have added "and why" to the question, but I was trying to avoid the typical "What's better" debates. I was interested in the "Why you do what you do" and not "Why you think others should try something different"

I'm always looking for improvement opportunities; ask the "but what if..." question. Sometimes just trying something out can come with a significant risk when I don't know what I don't know. We have recently tried a several new methods in the new CAD system that we didn't know much about. Some seem to be working or we haven't stumbled into the potential unknown problems yet. But we are also working to clean up the mess from a couple of the other things we tried.

I found the reasons that were shared for using one method over the other for certain cases to be in line with what we had found through the years.

To quote Lou Bega about using many various methods to get the job done suitably to any situation,

A little bit of Monica in my life
A little bit of Erica by my side
A little bit of Rita is all I need
A little bit of Tina is what I see
A little bit of Sandra in the sun
A little bit of Mary all night long
A little bit of Jessica, here I am
A little bit of you makes me your man

I am glad that I don't to the same thing all the time, nor in the same way. Curveballs take a different swing to knock it out of the park.

I am using bottom up , same procedure with mates in assembly.
For weldments i like to use multibody.
Sometimes i use multibody and save bodies external (this method always worked for me without any issues)

bnemec wrote: ↑Wed Dec 08, 2021 11:13 am Sometimes with all the chatter about the various top-down modeling methods and very little about plain old bottom-up modeling I wonder who else does no or very little top-down modeling.

Trying to be specific I'll add a little list off the top of my head; I don't know that every one of the lines below are really top-down vs bottom-up type things, but they tend to get drawn in sometimes.
- Piece parts in part files, weldments and assemblies in assembly files.
- Multibody modeling only used for ref geometry or tool bodies; one part number per file.
- No Master Models
- No Skeleton Sketch Parts
- No parts referencing other parts through a parent assembly.
edit: - Primary mating method is part geometry to part geometry (on rare occasion ref geometry if face ids likely to be change such as imported customer/vendor model); Mating origin to origin or coordinate systems is not used.

Who else models in this way in day-to-day practice?

Thank you.

We operate exactly like this. Top-Down would only be used for really early conceptual designs, typically before part numbers are assigned.

"Bottom-up" doesn't get a lot of discussion because there really isn't much to discuss.

You build, you assemble, maybe assemble some more. No context. No continuity. Easy-cheezy.

HerrTick wrote: ↑Wed Dec 15, 2021 6:18 pm "Bottom-up" doesn't get a lot of discussion because there really isn't much to discuss.

You build, you assemble, maybe assemble some more. No context. No continuity. Easy-cheezy.

Agreed, mostly. I could have asked in a bunch of different ways, but each thing I tried became too wordy, so I tried to just keep the question short and open ended.

There's a bit more to it that what you stated though. I was reading into the extras that people offered about why they stick with bottom up when they do so. Where simple bottom up becomes not so easy is when the entire data set is best described as the "Standard Parts Library." Maybe 2% of our hardware part numbers are OTS, the rest are special and have been or may be revised. I'm trying to get a feel for how other places with similar situations use the software.

Our production assemblies are exclusively bottom up and I can't imagine it any other way for us. We highly encourage part reuse so that bottom up is pretty much a necessity.
Top down can be used as a design tool, but we can't allow it into the production assemblies due to globally allowing and encouraging part reuse regardless of context.
We make lots of "similar", but custom machines, so there is a lot of value to be gained by sharing parts and not locking them down to a single context.
Also, we like to reduce bottle necks by encouraging "partial" releases. Managing full top down assembly trees in the context of frequent partial releases without full approval and review is a recipe for mistakes.
You can get some top down "like" tools by building reference geometry parts and assemblies. We have a universal "planes drawing" for machine lines, These are used as mating surfaces in assemblies and a tool for references and/or making temporary "in context" references to when designing. The temporary references are to be removed and fully defined with as much design intent as is practical in a single part context prior to production release.

nordstjernen740 wrote: ↑Wed Feb 09, 2022 11:35 am Our production assemblies are exclusively bottom up and I can't imagine it any other way for us. We highly encourage part reuse so that bottom up is pretty much a necessity.
Top down can be used as a design tool, but we can't allow it into the production assemblies due to globally allowing and encouraging part reuse regardless of context.
We make lots of "similar", but custom machines, so there is a lot of value to be gained by sharing parts and not locking them down to a single context.
Also, we like to reduce bottle necks by encouraging "partial" releases. Managing full top down assembly trees in the context of frequent partial releases without full approval and review is a recipe for mistakes.
You can get some top down "like" tools by building reference geometry parts and assemblies. We have a universal "planes drawing" for machine lines, These are used as mating surfaces in assemblies and a tool for references and/or making temporary "in context" references to when designing. The temporary references are to be removed and fully defined with as much design intent as is practical in a single part context prior to production release.

@nordstjernen740 are you using PDM Pro? If so how do you handle partial releases and versions?

We utilize pdm pro and manage. The "calculated" engineering bom in Manage is the master from which releases are cut.
Solidworks portions of the Manage bom are always "live" to the latest version of the solidworks assembly (with optional visibility to "as-built" or previous versions)

We postprocess excel exports from the bom in Manage to "help" engineers select their full or partial releases which ultimately get sent to sap and prod ctrl.
Some of this relies on the engineer to manage their release strategy with notes and common sense, but we have built tools like "release from bom compare" to help them out and for the most part engineers are able to do so.
We are also working on closing the loop fully by storing release context data in the bom database in some way, but we aren't quite there yet, but have a couple of feasible roadmaps to do so. Release data is project context specific data for each release line item in a bom so it is difficult to store in a standard "live" bom (portions of which may be may be shared in other projects) without the use of so-called "items" or context specific bom fields. Manage has "ref specific fields", but they have some limitations that we are working on to get us there.

Our system is highly dependent on enforcing a "backfit rule". If the part doesn't backfit it gets a new number and/or you cut releases for the machines in the field to make the new version work. no exceptions. Any revisions requires that the engineers gives information in the part card on what to do with unissued inventory to bring it to the current rev.
Because of this, the "where used tree" is important for our engineers.

A similar system could definitely be implemented with just pdm or anything that gives you live access to the computed solidworks bom, but Manage just has some better bom management and visibility tools especially when it comes to non-solidworks data.