Absolute vs. Relative design? 2

[ParabolicTet]

In my experience there are two types of analysts:

1) The Absolutist: Will model the physics as accurately as possible with measured loads, system interaction and detailed features.

2) The Relativist: Analyzes a new design and compares to analysis of an established baseline with a proven field history. Typically done when fatigue is predominate failure mode and loads are not known.

Are there any other types of analysts? Is the future of analysis heading more towards the absolute type? Or is the relativist analysis inherently a better way to tackle a problem? I think a lot of analyst gravitate towards the absolutist analysis because they are conditioned to thinking every problem is unique with it's own solution.

 

[GregLocock]

The relativist is the guy who says - if the old one broke I'll make the new one twice as thick. Call me a pessimist but I don't need FEA for that.

How can the relativist optimise his design?

On the other hand I do believe correlated models are essential.



Cheers

Greg Locock

 

[TheTick]

The first casualty of automation is judgment.

There is an appropriate balance between "absolutism" amd "relativism". There should always be a sufficient level of detail to capture the details of interest. It is up to the analyzing engineer to determine how much detail is enough and how much is too much.

Due to illness, the part of The Tick will be played by... The Tick.

http://www.EsoxRepublic.com

 

[corus]

As in life there are two kinds of people, the optimists and the pessimists. The difference between an optimist and a pessimist is that the optimist doesn't have all the facts. All analysts are optimists.

corus

 

[IRstuff]

That's not been my experience.

A pessimist says "more data, please," the optimist says "I don't need no stinking data."

And I've seen more pessimistic analyst than optimistic analysts.

TTFN

 

[GregLocock]

I have just spent /4 months/ correlating a non linear model. Its dynamic behaviour now agrees to better than 93% for all important measures at all times during the correlation events, and has an R^2 >.9 for all those measures over the whole event.

Do I believe it?

I'd use it for trend analysis, but no, not really BELIEVE.





Cheers

Greg Locock

 

[Drej]

In my experience, as is my analysis outlook in general, all calculations are guilty until proven innocent.

-- drej --

 

[magicme]

i tried not to drop my dime on this topic, but the temptation has been too great....

whatever you may call the two types of analysts, i find that younger engineers tend to be what you call "absolutists" and i believe it's because they inherently need to flex their intellects and demonstrate they are great thinkers and can apply the "latest" solutions to the problem at hand. they also love to "optimize" designs (often implying that "we need to start with a clean sheet of paper".)

older engineers generally have grown out of that era and recognize that there is very little that is totally new under the sun, plus they have archives of "oh i saw that back in 19XX and this worked or that didn't work" ...... and they generally have become "relativists"..... they try not to reinvent the wheel or the lever.

so i guess by my thinking, when you stop trying to re-invent the wheel again ...... you're an old timer!

pretty philosophical for an engineer, huh?

daveleo

 

[AlohaBob]

Oh I get it. It'a rhetorical question. Right!

 

[EnglishMuffin]

In relation to GregLocock's second post, I have heard it said, and have personally found it to be true, that when an engineer analyzes something theoretically, others tend to have much more faith in the results than he does himself. But when someone has made actual measurements, especially if the results are unexpected, it’s the other way round - only the guy who actually made the measurements believes them. This phenomenon has undoubtedly been a contributory cause to a number of catastrophes.

 

[ParabolicTet]

I agree that the main problem with relative analysis is that you may be comparing to an over-designed baseline. In this case your new design won't be optimized.

With absolute analysis, there are a lot more variables for inputs. With each variable there is also a great deal of uncertainty. So your final answer has a big variance and becomes almost meaningless sometimes.

I think the best analysis should always begin with relative/ comparative analysis and only proceed to absolute analysis if the designer has time/ resources to furthur optimize the structure.

 

[gurmeet2003]

If the product being analyzed is one of a kind, one has to go for absolute analysis. However if there is field experience available with similar products already in use, then it would be unwise not to use the field experience.

I have done this quite often in my work. I tend to use an existing product as a benchmark. This product may have been designed without the benefit of FEA in past. I use the benchmark to figure out existing factors of safety. Then as a start I use the same factor of safety for the new product with FEA. I think just the process of putting the product through FEA results in better product, even if same factor of safety is used. For example one can easily figure out where the material is not doing much good. Also one can figure out the optimized shape for a part.

Of course with more experience with FEA of product one can refine the design more and more.

Gurmeet

 

[LeonEng]

whether become a absolutist or relativist would depend on the field that we're working on....
If you are dealing with rather simple application(for example design metal part to work below yield point), or in an academic institution where you may have the luxury of time & budget to characterise the related material & build mathematical models, then likely you will be an absolutist.
On the other hand, for those who work in the industry and deal with much complicated problems,where the material properties & physics of failure are not well characterised yet...and the management always want the result on YESTERDAY, then probably you'll be a relativitist...

 

[elogesh]

Hai,
I absolutist most of the circumstances.

Just a dynamic case may explain, why absolute values are important. Usually there is a thinking that stiffer is better.But in occasion, you may be stiffeneing the structure in relative sense and hence may
be moving closer to the resonating frequency of the structure.Correlation of the modal model makes the absolute values

But few occasions, I was relativist. One such occasion was, there was a rigid new part to be designed for fatigue loading.And also there was an existing part used for different family product and it has been in the field for many years and passed the accelerated life testing. The loading is completely reversed loading(That is R=0) and the part is made of ferrous material with fatigue strength=0.4*tensile strength This existing part is developed before FE based software came into existence in our organisation.It was recommended to carryout analysis on this existing part and compare the results of the new part with the existing part.As a surprise the stress values for the existing part were higher than the allowable limit.Then all aspects of FE analysis was checked.Later on our metallurgist, mentioned that the measured tensile strength is always used to be at the upper bound.But the tensile strength referred for FE is lower bound,being a conservative.In this situation apple to apple comparison will hold good.But in the report I clearly mentioned that I accounted upper bound of tensile strength for being cost competitive.If one batch of sample is supplied with lower bound of tensile strength,then it will become hell.After reviewing my cautious report,they asked to redseign the same new component with same material content but effectively redistributing it and asked me to use the lower bound of tensile strength.May be I might have overdesigned by accounting lower bound in wide range of tensile strength value.

As mentioned by gurmeet the factors for size, surface finish,etc can be determined better by relativity.

As leoneng perfectly puts it,the industry always wants the output quicker, even before the input is
collected for the analysis.

Regards,
Logesh.E