Where to Go with Parametrics?
Sample background sources:
- Matt Lombard: Is parametrics history?
- Jeff Rowe: The Modeling Pendulum Swings
- SpaceClaim: SpaceClaim 2007+ Challenge
- DesignNews: CAD Consolidation Continues with PTC/Co Create Merger
- WikiPedia: WikiPedia: Computer-Aided Design and WikiPedia: SolidModeling
There seems to be a large discussion about the future of the CAD industry, particularly the use of Parametrics. Parametrics add smarts and history to CAD models, which can be good and bad. Right now the companies banking on parametrics are the big suppliers of CAD software - the "Standards" - so parametrics are standard. However many long term CAD veterans, long term products, and quick upstarts are trying to chip away at that standard.
For many problems, using parametrics is an appropriate tool. However people run into problems at the extreme ends of the modeling spectrum:
- on very complex shapes (see Matt's column above) parametrics either won't generate the necessary shapes, or the built in intelligence gets in the way, since everything has to be recalculated and rebuilt, and small changes can cause unintended consequences downstream
- on very complex parts (such as complex injection molding parts that may have several hundred features) managing the history and the relationships becomes a large part of the exercise
I haven't made up my mind yet, but I am also not running into the situation where I need. Overall I see the following advantages to either system.
Advantages to being Dependency-Free (eg. Boolean Model, Dumb Solid)
- Easier Import/Export, because only topology needs to be translated.
- Facilitates CAD-Neutral Interoperability, because no intelligence needs to be translated from protected to different protected system, no need for training and acquisition of translation software.
- Protective Intellectual Property, because no intelligence is stored with the Data.
- Fast evaluation of concepts, because it doesn't care how it got there.
- Easier learning of the software, because fewer intelligence means fewer commands.
- Intuitive, because you don't need to understand how what you have got to be that way.
- Easy collaboration with people not the same software as yours (e.g. FEA, CAM, CAD, purchasing, vendors).
- Faster changes, because you don't need to rebuild the entire history every time.
- Easier to modify late in the process, because intelligence don't pile up and starts interacting with each other.
Advantages to being Parametric
- Smart, because it records design intent and thereby facilitiates modification
- Informative, because history (and annotations) is stored along the way you can verify why features are the way they are.
- Protective of Intellectual Property, because it is harder to modify you are provided some control.
- Deliberative design, because you have to think about design intent and make deliberate decisions.
- Reusable, because parts, subassemblies, even sketches and features can be extracted, saved and reused downstream.
- Intelligence, because all the information recorded by your model can be used elsewhere (like in tools such as FilletXpert)
- Efficient, because design tables and configurations allow you to store multiple parts in a single file.
- Integrated and controlled, because the design can in the same model create configurations for FEA, or mechanism design.
- Time Saving, because information (say 3d Annotations and Tolerances) are entered only ones. Drawings can pull information from 3d models.
- Flexible, because bidirectional associativity allows you to make changes ones and have them migrate across.
- Leading Edge, because having an enclosed software system allows the vendor to implement changes without having to worry about "neutral" standards.
Disadvantages
Any of the advantages above will bring responses of similar disadvantages, like
- fine if the tool is easy to learn, but do you want people to learn the tools easily without much training? For FEA results, it often used to be Garbage In, Garbage Out. The same applies for robust CAD models.
- fine if the vendor can easily make changes to the part the designer sends him, or is it?
- fine if the designer wants to build the FEA simplified configuration, but shouldn't the analyst do that? How would he get on the tool.
- fine if you have all that design intent stored with the model, but who cares? If I look at it 5 years from now, I may not want to understand 500 features before I can make a simple change.
- fine if the vendor can push leading edge tools, but will they be maintained for the long-term? Am I not better off with a vanilla flavored program and file formats that will be readable for the long-term (see ASCII).
I can see that there is a place for both. I have not run into a lot of problems, probably because of the type of modeling I do, and how I approach it. I do like having the smarts, because they help me communicate (to myself and others) what it is that I am doing. They also allow me to put a little bit of time up front for significant gain downstream. But you do need to be organized, and deliberate, and considerate to people downstream.
You can have a tool that allows you to chop away and quickly make many iterations without concern of what came before. Or you can have a tool where you have to think a little about how you got there and what you intend to do, but be smart about it.
I can see that once you reach the extremes (either specific complexity or overall complexity) that this intelligence gets in the way. Plus by using these highly specialized, instead of generic, tools, you're working in a closed software system with methodology that may become obsolete sooner rather that later.
So no wonder it's a concern to industry, and an opportunity for many.
Non-dependency modelers talk about parametrics on the fly, where intelligence pops up when you select any face to make intelligent changes when you need them. Meanwhile the parametrics vendors introduce things like SolidWorks Instant3D where changes are made to any feature on the fly (say by dragging faces) and then transmitted back to the original sketches to maintain the history.
I've found myself switching off parametrics several times. If you're looking at the design process, there is lots of activity up front until the design is complete. Is that quicker down cheap and low-impact, or deliberate and intelligent? Either approach may be reasonable. Does the intelligence need to be maintained and communicated? If you are looking design downstream, oftentimes you end up having to make very simple changes to sometimes complex parts (change a hole diameter, add a component). Understanding all the built-in intelligence downstream may cost more than the benefit derived from having it.
However, SolidWorks has developed into such a complex tool that it is relatively easy to regenerate any feature downstream from a dumb solid (see FeatureWorks, or any of the Multibody tools). In that case you could store the part, for archiving, as a dumb solid, and then work only on the feature you need, when you need to. For parts like that, saving a part as a Parasolid (or a STEP) part, may not be the worst proposal.
So where to go with parametrics? Be smart for now in your modeling, choose the appropriate tools, and see how the industry develops.