Wednesday, November 18, 2009

Predictability With Direct Editing - Part II

In my previous post titled Predictability with Direct Editing I showed a simple example of what it means to get expected and predictable results with “direct” or “explicit” editing. It was a simple example in that the geometry was mostly analytic – planes, cylinders and cones. With this example I am adding much more complexity in that most all surfaces being represented are complex b-spline surfaces. This example will greatly push the limits of direct editing.
As I have mentioned before when pulling on faces using direct editing, the system needs to extend or shrink the adjacent faces, but should also properly take into account rounds, fillets and chamfers. This gets much more complex if the adjacent faces happen to be complex freeform b-spline surfaces. How well will the system extend or trim the surfaces? Can you get predictable and consistent results?
For this article I also created a simple video to show the example. I start with a simple cube and then using direct editing capabilities to deform several of the planar faces. If you are trying this on other CAD systems you may have to use other methods to get the complex surfaces, however it shouldn’t matter how you create the surfaces. Once you have them just add the rounds and chamfers to get the basic shape.
Hopefully the topology of your part matches the one in the picture. Pay close attention to how your CAD system creates the chamfers. Are they accurate? It’s a bit surprising how some modern CAD tools treat chamfers.

Now let’s try some direct editing.

Try this example on your favorite CAD tool. Can you get expected and predictable results? How well do the surfaces extend? Is chamfer and blend sizes maintained? Are tangencies maintained and accurate? Are the surfaces extended in a logical and predictable way? Are the chamfers and blends order dependent? Live rules and parametrics can help control change, but if you can’t get through the basics of change they add no value.
Even though I considerably increased complexity with this second example, besides adding blends and chamfers I still did not change the topology of the model. So perhaps sometime in the near future I can bring it all together, including complex topology changes, with a realistic part.

Monday, November 2, 2009

Inventor Fusion Preview II – The Fusion

I too was given the opportunity to see Inventor Fusion Preview II. I was a bit shocked when Autodesk invited me. I told them again that I work for PTC, but they still wanted to show it to me. I guess attention is attention. So here’s how it works (as near as I can tell):

The demonstration started with a history-based model that was created in Inventor. In this case the model was fairly simple and included only a few features. This part is saved as a native Inventor file.

In the example given, another user would load the Inventor file into Fusion to make some modifications to the file using the direct editing provided in Fusion. When the Fusion user is satisfied with their edits they can save the modified part as a new Fusion file.

Back in Inventor the designer may be notified of some suggested modifications to the design of the part. The designer would then load both the original Inventor model and the modified Fusion model. When this is done, Inventor will scan the topology and geometry of both parts looking for differences between the two models. After the scan, edits (modifications, deletions, additions) will be highlighted by different colors. Inventor will provide a list of the modifications that are being “suggested” by the modified Fusion part. These modifications can then be accepted or rejected.

If accepted, Inventor will look for the effected features in the history tree (sketches or parameters) and will drive edits to these features updating them based on the recognized changes from the Fusion part. Accepted edits that cannot be accommodated through modifications of elements in the history tree will be captured as new features at the bottom of the tree. The end result is a modified history tree based on direct edits from Fusion.

Basically the Inventor Fusion technology is identifying differences between two B-Rep solids, recognizing which history-based features need to be edited, added or removed, and then applies the accepted modifications to the history tree.

There are two basic technologies at work here. The first is the scan. Technology for scanning two B-Rep solid models and geometrically comparing then is nothing new; it has been around for many years. CoCreate for example has been using this technology for many years to provide version compare and change management for its users. It does have its limits however. Robust direct editing can result in a model that is completely unrecognizable from the original by even the best compare tool, especially if the topology of the model changes. I can conceive of many examples where this scan will result in inaccurate results. If the scan delivers inaccurate results, the second technology is meaningless.

The second and most significant technological contribution that is being made with Inventor Fusion is in taking the results of the scan and identifying what feature, sketch or parameter from the history tree must change to provide the same geometrical results of that of the part modified with Fusion. It is basically attempting to map a “geometry” modification back to an element in a history tree. In some cases, such as the blend and chamfers from the above example, it can remove a history based feature and/or add and organize new features as needed. (I’m not sure if it can change the order of features yet.) If it can’t make sense of the Fusion modification, a new element will be added to the tree representing the faces involved in the direct edit. When the history tree is regenerated, the resulting model should geometrically match the Fusion model. It’s an interesting technology and I hope to have a chance to dig into it a bit more. I am still skeptical about how well it can create and order new features, in a way that will support the design intent. Time will tell.

I am a firm believer that technology is only an enabler to process. If it doesn’t enable, or bring value to a process, it is useless. So the question going through my mind is; what processes, or use models, can be enabled with the Inventor Fusion technology?

Above I roughly outlined a process. It could go something like this: Perhaps a designer is working on a plastic part. The design has progressed to the point where a review with the tooling supplier is required. The 3D model as an Inventor file is sent to the supplier. On review the supplier has some suggestions to make for this part to be better suited for manufacturing. Maybe for some reason the tooling engineer prefers using Fusion rather than Inventor. Using Fusion, direct edits are made to the model to communicate the suggested modifications. The modified 3D model, a Fusion file, is then sent back to the designer for review. The designer desires to keep the native Inventor data up-to-date and so loads both the original Inventor model and the modified Fusion model into Inventor. The suggested modifications will be highlighted for the designer to see and review. The designer can then either accept or reject the suggested modifications as needed. Accepted modifications will drive updates, deletions and additions to the history tree keeping the native history tree up to date.

I am trying to come up with other use models. Help me out.

When and why is it important to have direct editing when your CAD system of choice is history-based? Some of the more common reasons include the following:

1) I need to make some last minute modifications to the model, but the structure of the history tree does not support these required modifications. I will either need to rebuild the part or use direct editing. Will Inventor Fusion technology help with this? Most likely not. If you can’t make the edit in the tree manually, it is very unlikely that the system will be able to make it for you automatically based on a modified B-Rep model. I asked Autodesk during the preview what would happen if the Fusion direct edit would invalidate a feature in the Inventor history tree. Also, what would happen if the history tree structure in Inventor would simply not support the suggested Fusion edit? Here is an example they showed:

They demonstrated that in those cases the modified faces would simply be captured as new features (sculpt) in the history tree – very similar to adding a direct edit to a history-based model in other history-based systems. So, if the edits found in the scan/compare cannot be translated to a feature in the Inventor history tree, new elements are simply added to the tree as new features; as is usually the case when doing direct edits in any other history-based environment. In this case the result would be very close to the same with or without Fusion.

2) I want to use direct modeling for concept design because of the flexibility, but then want to use my standard history-based tool for detail design. Will Inventor Fusion technology help with this scenario? I am not sure. I would like to check into this more. What will Inventor do when a part that was originally created in Fusion is loaded into Inventor, with no counterpart from inventor to compare it to? Will it only yield a single feature solid as has always been the case, or will it attempt to do what FeatureWorks does in SolidWorks?

3) Preparation for FEA is often a good use for direct editing with history-based data. Perhaps we want to simplify the model by removing some features. Many times this removal process can be very painful in the history-based system that was used to create the part in the first place. In many cases the references and relationships are just too complex to make the necessary edits. As such many companies are choosing direct editing to support this process. Will the Inventor Fusion technology support this process? First of all, there is probably no reason to push the direct edits back to the native (master) Inventor model. We only need this data for FEA. Secondly, I do not see that the direct editing within Fusion is robust enough (yet) to support the sometimes very complex process of simplification.

Perhaps you can think of process or “use cases” where direct editing of a history-based model is of benefit. And then consider when and how the Inventor Fusion technology can bring value. When is there value in relating a purely geometric modification back to an element in the history tree? Autodesk has certainly introduced some new and interesting technology with this second preview. Now, how will we use it?