(Unregistered) 02/24/06 07:26 AM

Today's Culture in Tolerance Analysis Blocks Real Benefits

 
Today's Culture in Tolerance Analysis Blocks Real Benefits Use this link to read the full article

icconian (Unregistered) 02/24/06 07:26 AM


Here's another example of halfway into the article, I could guess who wrote it. Hot Dawg  it's Norm stirring the pot again and making us think!
I was a drafter for 10+ years before being introduced to GD&T in 1990. That was more than 10 years of working alongside engineers who must also have not known or cared about GD&T, let alone stackup analysis. (When I think of the countless hours of handlettering paragraphs of notes for runout, TIR, etc, and how GD&T symbols would have been so much simpler, I still roll my eyes & shake my head). It's a long ingrained culture that is resistant to change, but the hope is that people comfortable with 3D design will also champion 3D everything else, & they eventually will _become_ the culture.

Sturdy (Stranger
)
02/28/06 06:01 PM


Drawing attention to the generally poor ability among engineers to use GD&T effectively (if at all) is always timely. Tolerancing errors sometimes become glaring visible even to management, yet the root cause of prototype or production problems is rarely addressed. What is also a huge cost goes unnoticed. Unnecessarily tight tolerances added substantial cost to nearly every design I reviewed; I generally found the offending engineer unable to understand the analysis and unwilling to risk loosening the belt and suspenders.
Interestingly, it was during the transition from 2D to 3D, but prior to effective CAM tool availability, that I noted a dramatic increase in the GD&T knowledge of quality engineers. This was often enforced by the necessity of programming CMMs. In fact, I frequently worked with mere inspectors who had a better working knowledge of GD&T than the resident Mech and Manufacturing Engineers.
o
Since few engineers now work in an environment that houses manufacturing, and more design is done in 3D, GD&T knowledge is more critical than ever before. The design might well be manufactured half a world away by a team that would have difficulty providing feedback even if they spoke the same language. Facility among engineers with GD&T is part of an effective DFM strategy. To have quality or manufacturing engineers, or "checkers" clean up a design after the fact is not as efficient as having the design engineer get it right. And knowledge coupled with use of tools that evaluate stackup is imperative. Ignoring this guarantees that even advanced mechanical engineering will follow manufacturing offshore.
Operations Mgmt (emeritus)
(errr...good deed doer)

Norm Crawford (Unregistered) 03/01/06 10:28 AM


Thanks for support of my "stirring the pot". But really it is not so much to "stir the pot" as my intent is to simply recognize that tolerance analysis contiues to be over looked. What I hope to accomplish is that some of the higher level management and executives that have the obvious objectives to reduce cogs, get to market faster with the best quality and want to reduce physical prototypes to reduce R&D cost and time; go right into production tooling, etc. begin to realize that colorful spreadsheets do nothing to "produce" product. They make colorful design reviews of information and can be very effective in collecting information. But if the information collected is not valid then its garbage in and garbage out. For example, if a CTQ (an ever popular term for Six Sigma) is directly cascaded (another popular term) to a specific measurement (transfer function) between to features of a part or two parts of an assembly and the toerance analysis is invalid or doesn't exist; "What value is the answer on the CTQ scorecard?" What value is there to the design review? The bottom line is margin/profit. Ultimately when product is produced at a cost and sold for what the market will bear, the truth comes out and the scramble is on in a corrective action mode instead of continous improvement.

Pinsan Cheng (Unregistered) 07/10/06 05:23 PM

Tolerance analysis is not a solid engineering/science yet

 
After I read your message, I decided to let you know my experience with tolerance stackup. You are right on talking about the engineering culture and practice. However, I sadly say that tolerance analysis is not a solid engineering/science yet. Take a look at visVSA manual or training course books, there are all the detailed steps how to build a model and get results. THERE ARE NO THEORETICAL BACKGROUND mentioned. If you want to compare VSA with FEA, you can check all major FEA software manuals and you will notice the theoretical manuals there. I don't understand why visVSA does not provide its theory. You can image if I just follow you without understanding what the software is doing behind, I will not and can not go far. In my practice with visVSA, one of difficulties is the judgement of the results. How do I know I got issues in my models. I prefer thinking the issues with my models, instead thinking the bugs in the software. What will happen if we do have a bug ! in the software? How do I know? Yes, technical support there. How much time is needed. In industry, how many chances we have to report to management about the "trooubled" results?

Jeremy Wright (Unregistered) 07/12/06 01:21 PM


You may want to look at CETOL by Sigmetrix. I am currently using this software @ Caterpillar and in the training there is definite talk of the theory behind it. It uses the vector loop approach as opposed to a Monte Carlo simulation approach to solve. There is some definate advantages to this method. There is work being done at universities on this topic as well (Prof. Ken Chase @ BYU for instance) So I'm sure that this field will come up to speed in time. Remember too that FEA is a mature tool (NASTRAN came out in 1968) so I'm sure the look of Tolerance Analysis will change and grow in the coming years.

Tim McLellan (Unregistered) 09/12/06 03:39 AM


Having looked at both VisVSA and CETOL one of the things that implementors may want to consider is how your tool stores it's data. Does it have it's own data base? If so, how is the data associated with the production geometry which goes through many changes and then released (solid models). You may find that you will need to figure out how to manage yet another data base then figure out how to tie it to your Solid Model Revisions to reflect real world results.
Norm's points are good ones. Simply put, it's about cost avoidanace. Advanced packaging, Solid Modeling, Tolerance Analysis, CAE, Plant Layout, etc are all cost avoidance tools and is something which is difficult for most to understand.
Tim

DavidO (Unregistered) 09/16/06 06:55 PM


There are a couple of comments I'd like to add.
First, on the issue of no theoretical VisVSA background material  keep in mind the following. VSA was a company purchased by Engineering Animation Inc. prior to EAI being purchased by UGS. VSA was primarily a consulting company that wrote software for the use of its consultants. VSA was never serious about the sale of software per se. As such, not only was there never any motive to document its software, anything that imparted knowledge to the users took consulting business away, so it was in VSA's best interests for this to remain a "mystery". It is Monte Carlo simulation based, which is about all anyone knows about it. Its subsequent owners had an appreciation for GD&T about equal to the average engineer, so its no surprise that information about it is limited.
Second, on the issue of the vector loop vs. Monte Carlo simulation. Vector loop methods are in general very fast, and the do not require actual shop floor results data to be effective. The tradeoff for this is that vector loop performance deteriorates rapidly as the size of the problem increases, and it does not take into account variability that is introduced as a result of variations due to manufactuing methods and machines. In all cases, IMHO a MonteCarlo based simulation approach yields superior results, and can do so on very large problems (i.e a complicated assembly with corresponding tooling), but to be most effective requires results from shop floor operations to model the variation upon which Monte Carlo simulation depends.

JR Haag (Unregistered) 10/03/06 08:32 AM


I can help you with understanding CETOL data strategy and discuss the new CETOL v8.0 architectural enhancements that will make data management and CETOL model reuse a much easier process with the CAD system.
CETOL v8.0 will be released soon, and has extended our Exact Constraint Technology, our solvers and analysis solutions that will allow options for synchronization with the CAD model, storing outside the CAD model, and reusing the CETOL model as a template to be connected to different geometry of similiar design. CETOL maintains a unique identifier with the surface ID on the geometry, which a persistent relationship that allow for the updates you are referring to. You can contact Sigmetrix directly through our website at http://www.sigmetrix.com

Norm (Unregistered) 10/05/06 03:38 PM


Now if this isn't piggy backing off my creative writing to make sales commercials. What is? LOL
VisVSA can be used with ANY cad system. Doesn't even need a CAD system. Reuse and updating templates on different designs has been around since the day of conception.
No wonder others are trying to look more like VisVSA. LOL

TedA (Stranger
)
10/09/06 01:42 PM


You mentioned in your post that CETOL is based on vectorloop technology. This is no longer a valid comment, even though the original research with Sigmetrix and Dr. Chase involved vector loop models, and Dr. Chase continues to use vector loop models for some of his publications and explanations on the ADCATS website. Early releases of TI/TOL (the precursor to CETOL) were based on a vectorloop model. That was from 1994 through 1998. There are many inherent limitations to the vectorloop approach. The current CETOL software uses a much more sophisticated modeling approach that has two components: a kinematic assembly modeler with advanced joints based on precise contacting surfaces and a part modeler based on feature constraint technology. This modeling approach supports the precise modeling of complex 3D parts and assemblies for tolerance analysis.

TedA (Stranger
)
10/09/06 01:53 PM


You have presented many interesting points with respect to vectorloop modeling and Monte Carlo simulation, although it is not really valid to make a comparison of these two technologies. Vectorloop modeling is a constraint system definition, not a solver technology. The public information available on this subject is a little confusing and limited. Most of this information doesn't precisely discuss defining the assembly functions and part feature relationships through different modeling methods vs. how the functions are initially solved, and then how the solved solution is finally analyzed/reported. There are generally two tolerance solver technologies used in current commercial applications: Monte Carlo Simulation and derivativebased analysis, both of which are supported in CETOL. A vectorloop model can be solved with either of these solution approaches.
In your post, you mentioned the fact that Monte Carlo simulation supports the ability to account for variations that occur due to manufacturing methods and machines. Actually, that capability is supported by both of the solution approaches (Monte Carlo and derivativebased analysis). In fact, that is much more conveniently accomplished in the derivativebased analysis approach because the manufacturing variation can be incorporated into the analysis as a postprocessing function. Once the derivatives (sensitivities) have been calculated, the results can be recalculated quickly (almost instantaneously for most models) when manufacturing variation data (or any other input data, for that matter) is entered. With a Monte Carlo approach the entire simulation of the original model with the new manufacturing variation information must be rerun, as it would with any change in input data. This is one of the inherent problems with Monte Carlo simulation.
There are conditions where one solver may be more beneficial than another. Both are capable of generating results for large, complex problems However, with Monte Carlo solvers it is difficult, if not impossible, to ascertain the validity of the model that is being solved or, more importantly, the accuracy of the results. A derivativebased analysis provides the derivatives (true sensitivities) and contributions that show the precise mathematical relationships between the measurements and dimensions. Thus, the sensitivities and contributions can be used to verify the validity of the model and accuracy of the results. This is the most significant differentiation in the overall benefits of one solver method vs. the other: the ability to use the output to clearly understand the model, eliminate the error associated with modeling assumptions, and provide higher confidence in the results' ability to predict and correlate with the real manufactured product.
Many of commercial applications that use the Monte Carlo solution approach provide tools for estimating the sensitivities and contributions. However, these calculations are incompatible with many of the models created in those applications because of the discontinuities and nonlinearities inherent in the modeling approach used. Thus the sensitivities and contributions that are reported in those applications can be highly misleading.

Tom R. (Unregistered) 10/30/06 07:18 AM


We use both VisVSA and CETOL were I work (QMC).
We have expert users for both tolerance analysis packages.
I worked for VSA for 8 years and used the VSA software for over ten years.
I have used the CETOL software tool for the last 6 years to conduct tolerance analysis studies of large complex systems. If you have any questions about the application of either package please contact me at www.qmc.com.

Dave S. (Unregistered) 11/07/06 08:09 PM


The way VSA worked was easy for anyone using it to understand. This was because VSA generated an intermediate plain language file that was then compiled and processed.
Want to see how it worked? Read the source code that was fed to its compiler.
This situation gave a huge advantage in that any program that could be gotten to talk could be included into the VSA analysis  Have a cylinder combustion chamber temperature program that depends on the compression ratio? Take the outputs of the varied diameter, stroke, wristtocrown variations, et al, and use those to generate temp outputs and sensitivity contributors to same.
The feature VSA did not have was the ability to handle variation due to default angular plus/minus tolerances. It did best with features controlled by feature control frames and their geometric characteristic symbols. OTOH nobody I know handles default angular plus minus tolerances in their analysis.
Vector loop or Monte Carlo, shop floor info is only necessary if yields are to be predicted, otherwise it doesn't matter. At least the models I built never depended on shop floor data. We just checked to see if, when within tolerance, the parts would function.

Norm (Unregistered) 11/28/06 10:31 AM


I reccomend NOT using +/ angular tolerances in favor of GD&T angular conrol that provides a total width zone versus the pie shaped zone.
The absolute need for a +/ angular tolerance has never happened for me. So, if there is a case in your applications, I would like to learn about it.

Srini (Unregistered) 11/29/06 11:45 AM

Looking for opportunity in Tolerance Analysis
[re: DavidO] 
 
Hi
I am looking for a VSA GD&T Tolerance stack up position.
Can any one help me? I have 10year exp in product development and 3years in using VisVSA software for different assemblies like Chassis, Suspension,Airbags, Electronic packagings like PCB packaging.
srini_chetlapalli@yahoo.com
