Digital Prototypes? Sometimes, “Going Back to Basics” is Better
Appears in Print As: Digital prototyping? Sometimes, a “back to basics” approach produces a better automobile
|
Authored by Toyota Motor Corp., www.toyota.co.jp/en |
It’s easy to fall into the trap of relying too much on digital prototyping to test products. So says Brian R. Lyons, safety and quality communication Manager at Toyota Motor Sales U.S.A. Inc., Torrance, Calif. “For many years, Toyota has practiced a disciplined and metered method to manufacture vehicles. It lets the company build high-quality cars,” he says. “The approach included several prototypes for each model, obviously an expensive proposition. To cut costs, we increased our use of CAE and built fewer prototype vehicles.”
This helped reduce costs by letting engineers compare aerodynamic flows in different simulated engines rather than actually building physical engines. However, the company suddenly saw an increase in quality issues.
“We started seeing problems we never had before,” says Lyons. “For example, we extended the warranty on the 2001 through 2003 Prius because a ‘check-engine’ light kept going on. We had done extensive digital testing on the engine and its components, and everything worked fine in the virtual world for hundreds of thousands of miles. But after the cars had been in operation for several years, it became evident that certain fuels used in the U.S. leave carbon deposits on the throttle body. Digital testing didn’t match real-world roads and environmental conditions in the U.S.,” he says.
So the company decided to keep using CAE but also go “back to basics,” using what worked in the past. The approach is summed in the idea of “genchi genbutsu,” or “go and see,” says Lyons. “In other words, we ramped back up on physical prototyping. When the Highlander came out, we launched our Customer First initiative. As part of this, we increased the number of physical evaluations letting quality assurance, product planning, and even sales and marketing personnel test drive cars in real-world conditions. The Highlander was the most problem-free car we launched at the time. Now we’ve produced several models using this approach.”
The back-to-basics approach continues to lower warranty numbers, says Lyons. It also lets Toyota build cars that better target different markets. “We had to find out how consumers in different areas of the world actually drive the vehicles,” he says. “For example, unlike Japan, it’s common in the U.S. to drive on gravel and dirt roads. There is less traffic here than in Japan, and roads are wider with a lot of high-speed driving. Canadian and European drivers are different as well. In Europe, for example, you find extremely bumpy Belgium block roads made of slabs of rock and mortar.”
Want to use this article? Click here for options!
© 2012 Penton Media Inc.
© 2012 Penton Media Inc.
Delicious
Digg
StumbleUpon
Reddit
Magnoliacom
Newsvine
Furl
Facebook
Google
YahooAdvertisement
Supplier Search
Plastics Search
Advertisement
Advertisement
Advertisement
Advertisement
Advertisement
Comments
I wish if simulator are perfect.
I wish , if simulator are perfect, if it is so, it make my job much easier.
and save the a lot of time, but unfortunately, there is no such things as perfect simulator. I often makes antenna, RF circuit. I am not very good at using latest simulator, but one I used ( from 1999 ) , I still have to make prototype and change the circuit or dimension to adjust to work good.
but, I know many engineer, who are good at latest and the best simulator, they says they work much better, and require a little to no modification from simulation result.
Digital Prototypes
Trust but Verify
Digital prototypes versus actual prototypes
It is true that programs for Virtul prototyping are far from perfect, because tests are done separately and even air flow is done in 2D instead of 3D, all for lack of common description parameters that would enable computers to test everything in paralel, like it is happening in real life.
But reason given in article for going back to more physical prototipes testing is rather unconvincing, since as they say, it has taken years of driving that type of wehicle for such problems to show. Just as they did not had that kind of fuel in program, so surely they did not have it in their prototype testing laboratories.....
Problem lies in gap between people who know how to program computers that usually know nothing about engineering, and engineers that do not know possibilities of computers.
Solution is not in going back to physical prototype testing, but to make better programs that would be able to manifest synergy and even emergent characteristics of device tested.
I as programer with 30+ years of experience have viable ideas how to make parametrized description of parts and subsystems of any kind of device, which would enable that changes in parameters in one part automaticaly influence parameters of functionaly connected parts, and since that would be also accesible to program itself, it would be able to adjust each part as well as whole for optimal performance and production costs.
But most important would be ability of program to test it all in paralel, where results of one kind of testing would be reflected in results of other tests interdependently.
Unfortunately, I am in bad financiall situation and have to fight literaly to stay alive, so I cannot sit and make it.......
I discussed my ideas with several companies that produce such software, and each is to far gone wrong way so they would need to reprogram everything, and this cost too much so they would rather sell half unusable programs with few new >>features<< added each year, then to start it all over again.
If somebody would finance me, such program would fast corner all market for >>Digital Prototyping and Virtual testing<<, specially since it would have elements of Expert Knowledge databases, Fuzzy logic, Neural Networks, Evolutionary and AI programming.
I just dont know how to make contact with interested parties, and I hope that leaving my email address would not cause this post to be deleted...... So, contact me on mpollak at globalnet.hr
Sometimes Models Don't Cut It
We are producing a product where we've done extensive modeling. These have been valuable for several purposes, but have turned out to be nearly useless for predicting stresses.
The parts are geometrically complex, which drives FEA programs nuts in meshing, even with simplification. Predicted stresses are nearly double actual.
It turns out the most cost effective method was to build one and break it.
Bingo !
"It turns out the most cost effective method was to build one and break it."
Having had exactly the same experience, I couldn't agree more. Our certifying organization requires FEA / stress analysis results to be within 5% of strain gage measurements. That turned out to be virtually impossible.
In general, computer simulation is only as good as the theoretical model it's based on. And NO model is reality. Just close.
Leave a comment