Dassault CATIA versus UG NX5 in automotive design and manufacturing
On May 3, 2010, Automotive News reported on a move at Chrysler to replace CATIA with Siemens’ NX (originally created by EDS) to work better with Fiat, which uses NX. Engineer Bob Sheaves was part of the evaluation group when Chrysler chose CATIA in 1988; Chrysler spent almost $2 billion over five years to convert everything, from product to process, to CATIA.
Chrysler has used CATIA since 1989, starting with the Jeep/Truck Engineering group. Before this, Chrysler had used manual drafting and an internally written program, Chrysler CADCAM, supported by the internal UNIX and mainframe computer groups.
Chrysler’s CADCAM was a great advance on manual drafting; design and engineering efforts were more repeatable and more accurate, with less “guesswork” on the projection of skewed views (views that are not parallel to the major drawing views, side, plan/top, front, arranged according to various standards which vary by country and company.) [Chrysler claimed, with the Plymouth Scamp launch, that CAD had increased productivity by 3:1 over hand drawings.]
Manual drafting is, simply, making a two dimensional representation of a three dimensional object. Technically, it’s an “approximate representation,” because one cannot apply properties such as mass or moment of inertia to paper. The more complex the object, the harder it is to physically project true views to ensure the representation is accurate. Manual drafting is also slow, tedious, time consuming, and prone to errors. This key thought is important to understanding CAD systems.
Any CAD (computer aided design), CAE (computer aided engineering), or CAM (computer aided manufacturing) system — or, if you prefer, Product Lifecycle Management (PLM) — system is used for three reasons:
- It increases accuracy, cuts rework, and reduces the technical risk.
- It lowers costs by minimizing lost effort and repeated effort.
- It provides an automatic engineering-program history to lower costs on future programs of the same type (in other words, PLM provides for continuity to not “reinvent the wheel”).
Defining the need
- How do we manufacture the car?
- How do we make sure all the parts fit together correctly?
- How do we ensure the parts will survive in the real world?
- How do we position all these parts to assemble them consistently and correctly, according to the customer order?
How do we design the car?
- How do we ensure all the parts to make a car are completed and fit together?
- How do we ensure that we haven’t missed designing any parts?
- How do we make sure the paper matches what we want the customer to receive?
- How do we verify the design works?
- How do we make sure that we do not have any OOPS’ and how do we verify this WITHOUT building “umppity-hundred” mule cars to test every single change and component?
- How do we account for variances?
- How do we account for the build tolerances in the vehicle components without breaking the bank on costs by having too tight a quality check that provides no benefit to the customer?
- How do we cut overhead cost?
CAD/CAM/CAE/PLM discussion
Dassault Systemes has grown through the Chrysler and Boeing relationship (as well as other applications) into, until a couple of years ago, the largest supplier of CAD/CAE/CAM and PLM software in the world. This has lead to the “fat, dumb, and happy syndrome” within Dassault, leading them to believe they could do no wrong; and they showed this arrogance heavily towards their customers.
CATIA v5, in particular, has had a checkered history, to be polite. Unstable and hampered working with v4, the early versions (up to version 5, release 15) proved to be extremely hard to implement, administer, and stabilize.
Siemens stepped into this gap between the customer and Dassault, by purchasing UG from EDS (Ross Perot’s company, acquired by and sold off by General Motors) and developing the NX family of software. NX was directly intended to supplant CATIA at the top of the heap.
NX was, out of the box, easier to install, administer, and much more stable, but had no interconnection with any other system, as the data files (in this case, principally Dassault) are always in a proprietary format which is unreadable by any other system. STEP was used to transfer data between systems, by converting to intermediate formats (e.g. International Graphics Exchange Standard or IGES) for interchange, just as people, long ago, might transfer a WordPerfect file to MacWrite through RTF, losing any complex formatting along the way.
In general, file interchange works well with systems except CATIA, because of the extreme precision on the CATIA 3D data, compared to other CAD/CAE/CAM systems. There are a number of “significant digits” in any number; after a certain point, they are too small to make a significant difference, to a degree (though the more iterative processes are involved — that is, the more number processing, the more loops, etc., and the greater the need for precision — the more significant digits are required.) CATIA, because of the 19-digit calculation precision in its core kernel (NX provides 11), is far more…stubborn…in creating Class A surfaces than anything else. The surface is either right or it is wrong, there is no middle ground for acceptability or tolerance.
This shows how CATIA v6 allows the user to integrate FD&T (Functional Dimensioning & Tolerancing) into the 3D models. This information is carried along with the assembly, upward in the manufacturing process to include the relationships between the higher level assemblies. ... GD&T includes symbols that tell the machinist and quality control people what the tolerances of the part is, to ensure functionality, interchangability of parts, and establish the limits of acceptable variance. Carried through downstream manufacturing modules of CATIA, such as KBE (Knowledge Based Engineering), this provides the automatic parameter programming for DELMIA manufacturing using DNC, CNC, and NC processing, all without human intervention, including no APT or G-code programming manually of the machine tools.
This “stubbornness” can be a great thing (providing a stable and repeatable 3D model that is at the utmost accuracy) or a really, really bad thing when you try to work across various software platforms.
Because CATIA has more functionality built into and integrated with it, CATIA costs more and is harder to learn, but it is more robust. NX is far easier to learn, has many third party applications, is less expensive, but is sloppy in ultimate accuracy and repeatability.
Clarification: Dassault CATIA, Siemens NX, and accuracy
The Toledo plant pictured above is as real, mathmatically, as the final physical plant in Ohio. If you were to view it in a "tank" (a specialized room for viewing hologram data) you would see and walk through the plany exactly is if you were in the physical plant (and remember this is from a few years ago). It "exists" as full scale data, so all the line, mechanisms, human motion studies are figured out long before you pay to build anything.
The accuracy issue has been a primary point of contention between all software providers for years....ever since CATIA was first written by Dassault in the late 1970s. Mathmatically, with CATIA, a model is always designed to nominal, just like on the drafting board. The issue arrives in the definition of "accuracy." Let's take two examples....
In the first, we are going to build a house of 2x4s. In the real world, a 2x4 can vary considerably in the end measurements, because of non-uniform drying, warp, etc. The variance can be mathematically described as GD&T (Geometric Dimension & Tolerances), but the 3D model must be constructed to represent the nominal. In this case, physical clearances between each 2x4 are loose. Accuracy of 19 places is not required.
In the second, we look at a diesel fuel injector with tolerances of practical machining at 1/10000 inch. Since you will machine the part directly from the 3D math data, accuracy takes on more importance. The model of tolerance stackups between close fitting parts (like the injector) must be as accurate as possible so your NC machine’s model for predicting wear follows the correct path; and interchangable parts (you really aren't going to do this for just one part) all work as designed when machined.
The choice is leads to a question: how accurate do we need to be to not add excessive cost to the end customer? If we change from CATIA to NX, how much data will be lost? Remember, CATIA v5 is Chrysler Engineering, in all its facets...
(The graphic representation method is user selectable in CATIA v5 and v6. Depending on what you are doing, you can set the coarseness of the solid model representation on the fly. The user can set the graphics from a 0.030 mm sag up to 10 mm sag value. The finer the resolution, the more graphics memory is required to attain a usable design speed. The sag is graphics only, it doesn't change the actual model, only the display of the math behind the screen. )
Addendum: how they choice was made in 1989
There was never any chance for remaining in the same place. Iaccoca said "fix it" and that was what was going to happen.
The top 12 or so design engineers were assigned training in all the other programs, to carry out the evaluation with TCC (the Technical Computer Center — which was responsible to Engineering, not the IT group). An evaluation period was set and test problems were assigned. Basically, this was a fake pilot program that was carried out in each package. Evaluation criteria were:
- Speed to learn (how fast a person could be trained to be productive, how much OJT was required to reinforce the training)
- Speed to perform (the time to accomplish the same design task between all systems)
- How much analysis could be done concurrently with the design (the capability to be used in the "no rework" process direction)
- How much interoperability existed between all the different software and the translated accuracy differences, as well as interfacing with the various supplier systems and processes
- How can we change the design and engineering process to minimize time to complete and take advantage of the built in software modules for varios procedures (such as vehicle dynamics simulation, FEA, etc.)
These measurements led to a complete rethinking of how Chrysler designed and built cars.....things like SCORE, integrated manufacturing, etc that were the hallmark of Chrysler in the early 1990s. Now, today...TCC doesn't exist any longer, thanks to Daimler.
With the SCORE program, guys like me set up the suppliers to match Chrysler hardware and software, trained the supplier employees, and assisted with OJT and even design help to shorten the time of the learning curve. If you were a T1 supplier, this was not negotiable, until Eaton came in and forced the suppliers to eat everything and do it on their own. This was the beginning of the end.......and another story entirely.
Read more:
- Using CATIA to create the 1997 Dakota
- The Digital Factory project - designing production on the computer
- Using computers in the Neon engineering process
- SCORE: integrating suppliers into engineering to save billions
- CATIA presentation on how PLML works
- Siemens NX site • Dassault CATIA site