Monday, December 17, 2012

Model-Based Engineering, Are We There Yet?

  Part I

As I‘ve mentioned before in several other posts, I’ve had the pleasure of visiting many design and manufacturing companies over the last 20 or so years, sometimes several per month. With some of these visits I even get the pleasure of conducting full product development process assessments. With these assessments I get a first-hand detailed view of how companies design products, from requirements to manufacturing. I also get a glimpse of what their vision might be for the future of their design processes and where they see high-value opportunities in process improvement.

In a previous blog post from 2009 titled “Model Based Definition(MBD) – What’s the Hold-Up?” I gave my early views on the topic and asked my readers a few questions. It is still the most read article on my blog. Interest in this topic remains very high.

There is one consistency I find with any discussion I have regarding “Model-Based” and that is that everyone seems to have a different idea of what it means.  Do we refer to it as Model-Based “Engineering” or “Design” or “Definition” or what? And what's the difference? And what does the term “Model” really refer to?  I have found, however, that if we translate all of this “Model-Based” discussion and debate down to the product development process, what most process experts/owners will agree to is that they need to somehow:

1.       Eliminate the disconnects
2.       Remove the redundancies
3.       And move to a single product definition master
If this Model-Based “whatever” can help them with those three “opportunities”, they want it.
And I think it can, but we have to be in sync on what the real goal is. If “Model-Based” somehow refers to one or more of the three opportunities listed above, we have to know where these conditions may exist, where the priorities are, and what the vision is.
It was over 30 years ago that the CAD industry presented a solution to disconnects and redundancies in the process. They called it “associativity” (which, by the way, is still not a word). Everything was supposed to be associative to the 3D model, i.e. the master model, (an old term). If the model changed, all downstream deliverables would update. Create data once and leverage it to the max. Was this not enough, or did it just not work? Based on my experiences, even the most sophisticated of companies still have many, and I mean many, disconnects, redundancies AND master documents in their processes. When I ask a company what they consider to be their “engineering” master, (and I’m not just referring to the detailed mechanical engineering master) they may have trouble answering that question. And of course, their “engineering” master and “manufacturing” master are at least two different things.
For many people “MBD” refers to the merger of key characteristics of the manufacturing drawing with a rich 3D model, thus eliminating or greatly reducing the need for the detailed manufacturing drawing. But, what’s the big deal with fully annotated and detailed 2D manufacturing drawings? What's the value of merging this data into the 3D model? Let’s see how this stacks up against the three opportunities.
Most drawings today can be considered “Model-Based” in that they are extracted from a 3D model and are in most cases associated to the same. So in this case we should have no disconnect as the drawings are already "Model-Based".
It is possible that some of the work done to fully annotate the drawing is redundant to the work that was done when creating the 3D model? But does PMI and 3D annotation eliminate this redundancy or just move the activity to a different "document"? It is possible that we may be able to eliminate some standard dimensions/tolerances by better leveraging the 3D model. So there will be some value here.
The Single Master Document
Perhaps the most significant value comes via the merger of one of the engineering masters (the 3D model) with one of the manufacturing masters (the 2D drawing). There is always high value in eliminating even one document in the process, and this is a big one. The big question with this is; how consumable is this rich document (the 3D model) to downstream functions. Functions such as; Tech Pubs, Mfg., CAM, Tooling, Process Sheets, and the tricky one – QA. We are getting better at this, but still much progress to be made.
Is that all there is to “Model-Based”? Merging the manufacturing drawing into the rich 3D model? On the contrary, in my humble opinion this particular merger is perhaps one of the higher cost, lower value fruits that can be realized through a “Model-Based” initiative. Perhaps it’s a reasonable, although complex, place to start, but don’t let it stop there. There is so many other high value opportunities with “Model-Based”.
In Part II of this post I want to share some other areas in the product development process where “Model-Based” can bring huge value, and in many cases at very low cost. If you have been able to eliminate or greatly reduce the number of 2D manufacturing documents being produced you have made great progress, no doubt. But I can almost guarantee you that even after all that work you most likely still:
1.       have not eliminated all the disconnects
2.       have not eliminated redundancies
3.       and do not have one single product definition master

Friday, July 13, 2012

History-Based Modeling and Direct Editing

Most all parametric history-based CAD tools on the market now have some level of direct geometry editing capabilities. Without direct editing, users of history-based CAD tools can only edit preexisting information captured in the structure tree during model creation. Information like sketches, 2D constraints on the sketch, 3D feature parameters and feature order. To make these edits, it is important for users to have some level of knowledge about the structure tree and the process and methods used in the creation of the structure and resulting model. In the case of imported models, this information is typically not available and as such the imported model may not be editable, with the exception of adding new features to it. In the earlier days we tried to recognize features in imported models to develop an editable feature tree, but that typically only worked on relatively simple models. In other cases we tried feature tree translators. There are several companies that provide these types of translators, but a feature tree translation can be problematic depending on geometry characteristics, it will be version specific, and can be somewhat expensive.

Direct editing is being introduced into history-based CAD systems now to simply provide another option in history-based modeling for the editing of 3D models. With good direct editing we can be somewhat less dependent on a well-structured feature tree. However it does seem that some parametric history-based purists consider direct editing within a history-based CAD tool as somewhat of a risk. Perhaps they are concerned that it provides a way to “corrupt” the design intent. I find this a little strange in that if in fact the design intent (feature structure and definition) is already pure and perfect, there would certainly be no need for direct editing. Unfortunately this is rarely the case.

In reality direct editing in a history-based tool is just another parametric modeling feature. And like any parametric modeling feature, there needs to be clear best-practices and modeling standards associated with it. If used correctly, direct editing can be one of the most powerful parametric modeling features you may have in your CAD toolbox.

I want to try to highlight some of the potential power of direct editing within a history-based model in the video below. In this case I will be using Creo Parametric. PTC added direct editing to Creo Parametric with what they call “Flex Modeling”. Mature direct editing requires tools for geometry selection, methods to define the transformation (edit), and robust predictable results.

So having another method; direct editing, for creating an intelligent and ordered parametric feature can be very powerful when used correctly. It should be nothing to fear, but rather leverage when and where it makes sense – as with any parametric modeling feature.

PTC has done a nice job integrating the necessary functionality with a robust kernel to make their direct editing (Flex Modeling) very capable, predictable and robust. The same can be said of Siemens and their direct editing (Synchronous Technology) in both NX and Solid Edge. But as you consider direct editing, pay close attention to these three things:
  1. There must be good geometry selection methods including feature recognition, geometry rules and the ability to add and remove from the selection
  2. There must be intuitive methods for defining the transformation including dynamic handles and dimensions. AND, you should be able to make multiple transformations in one edit.
  3. You must get stable, predictable and robust results. In some cases there may be multiple results to a given transformation. The system should provide the possibilities and allow the user to select the appropriate one.
Not all history-based CAD tools do direct editing very well. Direct geometry editing requires functionality from the geometry kernel that typical history-based modeling systems have not had to do. History-based CAD systems are basically Boolean engines that are programmed through the history/feature tree, with the parametric feature being one of the primitives in the Boolean. With direct editing, besides Booleans, the system needs to know how to do local operations, how and when to close gaps, add and remove faces, and extend and trim geometry - all based on actual geometry manipulation. If it doesn't do these things well it will certainly show.


Monday, July 9, 2012

Parametric Direct Modeling

I continue to enjoy seeing what is possible with history-free direct modeling when coupled with a synchronous parametric solver; i.e. Parametric Direct Modeling – or whatever you want to call it. This is the truest form of uniting the control of parametric modeling with the flexibility of direct, or explicit, modeling. If, on the other hand, the modeling process is the basis of your design intent, as is the case with history-based modeling, then you are forced to plan ahead before modeling. This firm coupling of the modeling process with the definition of design intent can greatly inhibit flexibility.

When adding direct editing technology to a traditional history-based parametric modeling tool, you still end up with a structured and ordered model. The direct edits are just another form of a “parametric modeling feature”. They, of course can be very powerful and useful in the right context and I'll discuss this in a future post.

By combining parametric control with direct modeling it is possible to have history-free geometry and assemblies that behave according to the designer’s intent. You can develop relationships and intent independent from the modeling process. There is no need to plan ahead before you start modeling, and certainly no need to ever recreate a model just because the model construction process and methods no longer support the designer’s intent. Parameters can be added to any geometry at any time. There is no dependence on where, how or in what order the geometry features were created, or how they were orginally constrained.

Here are several examples. Some you have seen before on this blog and some you have not. The video is organized into these three groups.

Geometry level design intent:
In these two examples we are simply controlling geometry with a few parameters. As this is direct modeling, it makes no difference where or how the geometry was created, or in what order the constraints were applied. The parameters are solved synchronously, not linearly.
  • Variable flange with pattern
  • Automotive wheel supporting a family of wheels
Assembly level design intent:
With these examples we are controlling assemblies. In some cases we’re using permanent constraints and in other cases the system is doing real-time solving based on the physical properties of the 3D models.
  • Iris Simulation
  • Complex gear simulation
Combined assembly and geometry design intent:
With these last two examples we are controlling both assembly and part relationships as well as geometry. In the case of the drill press we are using NO permanent constraints or relationships. The constraints and relationships are managed real-time based on part and geometry selection. In the V8 engine example we are using a large number of geometry and assembly constraints.
  • Drill press design change
  • V8 engine stroke change

I hope those examples make sense. There is still much room for improvement in the area of parametric control in direct modeling, but progress is happening. Some of the solving you saw happen in the embedded video was done using the DCM Solver technology from Siemens, and some was done with solver technology developed at PTC/CoCreate. I'm also excited to watch what the people over at LEDAS are doing with this technology. With the emergence of this type of technology CAD is certain to look different in the not too distant future.

Next up I want to show you some of the powerful things you can do by utilizing direct editing within a history-based tool - its not only about editing geometry.


Wednesday, June 20, 2012

Geometry Simplification Using Direct Modeling

There seems to be a growing need for part and assembly geometry simplification. There are many reasons for this. Many times designers need to simplify a part to prepare it for the meshing process and final analysis. In other cases manufactures are more often now being asked for the 3D models of the products they produce as they may be a component or system used in a larger system. In this case they not only have the need to reduce the actual file size of the data, but also want to protect valuable intellectual property (IP). In other cases perhaps you are on the receiving end of purchased components or systems and just need to have  simplified compact representations of the data to reduce overloading your assemblies with unnecessary detail and data. Whatever the case, the need is growing.

One of the best tools to use for this simplification process is direct modeling. There are two key reasons for this. 1) It allows for the direct interaction with the geometry and is not dependant on the original model creation process. 2) As there is no history tree involved it can greatly reduce the size of the data, even before simplification.

Once the simplification process is finished, the models and assemblies can easily be brought into any CAD tool or analysis tool as reference geometry and carried forward as necessary.

The video below is a demonstration of model and assembly level simplification. In this demonstration I start with a parametric model and import it into Creo Elements Direct Modeling (CoCreate). In its original history-based form it’s about 39mb worth of data. That’s the data we will start with.

I hope you can see that with direct modeling it can be quick and easy to simplify geometry, and in some cases automatic based on characteristics of the geometry.


Wednesday, May 16, 2012

Another Fun Project

I first want to apologize to my readers for not blogging more lately. There are two reasons for this. 1) PTC is keeping me way too busy lately and, 2) much of my personal time has been spent on finishing up the following project.

It all started about 10 years ago when my son and I agreed to purchase a fixer-upper car to work together on as a project. I let him pick the car. He eventually settled on a 2nd generation Camaro (1970 to 1973). This is the one we ended up with. It’s a 1973 Type LT, 350, turbo 400 automatic transmission. It was drivable when we bought it and my son used it to get to and from high-school until he ran out of gas money. This is the picture from the EBay ad where we found it.

Over the course of many years we started restoring various parts of the car. Here are a few pictures that represent some of the work:

Every piece of sheet metal on this car has been refinished, and we did it all ourselves with the exception of the sub-frame powder coat and the final external body and paint. We also built the engine, rear-end and suspension ourselves. Hard to tell where this car has been but there were signs of many years of use. We even found a bullet hole in the right front fender.
The car now has a 383 stroker small block engine, with a lot of goodies in it. It is nothing too radical but came out on the engine dyno right a 400hp and 436tq. (Since then, last weekend, we put a little more aggressive roller cam in it.) We also replaced the Turbo 400 with a Tremec TKO 5-speed, and put 3.73 gears in the back with a Detroit Locker Trutrac differential. After a few suspension trials we ended up with mostly Hotchkis springs, shocks and bars, with the exception of the Detroit Speed springs in back. My son also picked out the final paint colors and wheels. It is certainly lots of fun to drive, and we plan to show it this summer. Here are a few pictures of the finished product. (Although it is never finished. Brake upgrades are next. Don’t tell my wife).

1973 Camaro Type LT

Now I have to bring this back to CAD somehow. Do you realize that this car was designed with the help of no CAD system? They used drafting boards back then. That thought ran through my head several times as we were dissembling and reassembling this car, (many times). I even have a copy of the original assembly manual for this car. It’s full of hundreds of drawings, and in the title blocks are the actual signatures of the draftsman and the checker. Very cool. (And today we fuss about which CAD system we get to use.)

I hope to now get back to some serious blogging, although no promises.