robust design vs optimized design

[magicme]

hello again

i would like to post some comments on "robust design" versus "optimization".

20 years ago, i learned a hard lesson.
i was tasked to reduce the weight of a truss structure (that supported a turbine) as much as was possible, using FEA.
this went on and on and got fairly intense. people reviewed my stress plots and wanted to reduce the amount of metal anywhere in the structure that we had stress margin. i caved into the pressure and ran endless cases and cut out all "un-necessary" metal.
the intent was to reduce weight, which was a major customer criteria.
i tried to get the dead weight removed from the "payload" (the turbine), but lost that battle.

we ended up with a structure that looked like toothpicks holding up an elephant. but the FEA said it would not be overstressed, given the cutomer's specified load conditions.

well....you can guess it.... after all the drawings went out and the steel was in the shop and they were welding up plates.... the customer came back with an "update" set of loads..... of course this perfectly optimized structure could not survive a change in loads, and of course i was the guy to blame for producing a design that was not "robust". we lost the contract to a design that looked like a solid block of steel (weight be da##mned).

i have ever since shredded any thought of optimizing any design i have worked on, and add my own personal "comfort factor" on top of all the other factors, so i can sleep at night. this has worked out very well over the years, because INVARIABLY, the week before a critical design review, something external will change that will test the robustness of the design and no-one will care how "optimized" it is if it can't survive the changes.

okay.... i got that off my mind.

regards,

magicme

------------------------------------
"not all that glitters is gold"

 

[corus]

If the major customer criteria was reduced weight and they went for a block of steel, and they also changed the loads, then I think you're working for the wrong people. If you were in a company that had previous experience with the design then they should have advised the customer beforehand. On top of that you might have known better from previous experience what typical ultimate loads would be applicable.
The first lesson in any design, I think, is does it look right? If you don't think so then the customer won't either.
You live and learn I guess.

corus

 

[cbrn]

Hi,
I agree with Corus. "Edge"-optimization and polyvalence (another way to call "robust design") are completely antithetical, and if the customer is not able to understand this then it's better to loose the contract. On the other hand, it's the task of the Mechanical Calculations' Team to warn everybody around about that: management is worried about material weight / cost? "OK, no problem, we can optimize, but be careful, Sirs, because you project managers MUST be sure that the design conditions are definitive"!!! Which brings me to another important point: if the Salesmen have put exotic weight values in the contract proposal, just in order to win the contract, then it will be THEIR fault if the contract is actually over-expensive, or if it can't be done at all...

This said, I wouldn't disregard optimization because in today's world it's a primary necessity... Or we may design machines just like in the far passed decades, when the instruments and knowledge were both more limited (*).

Regards

(*): ... but people knew very well how to use them, which is not always the case today...

 

[EdDanzer]

A couple of things to ask customers before starting down the optimization road:

1. Will deflection cause any problems?

2. Will you guarantee the loads, direction of the loads, and frequency of the peek loads in writing?

3. How much per pound is the optimization premium worth? It almost always cost more per pound to produce a highly optimized design over a simple robust design. In fact the highly optimized design may cost more than the simple robust product to produce.

 

[dik]

The economy was probably not wittling away anything that didn't look like a truss, but in the system selected, ie. using HSS chords and angle compression struts and diagonals, or whatever... I generally send a letter to the client confirming loading conditions stipulating code minimums and asking if he requires special loading... I too was chastened at an early age...

Dik

 

[EvanGr]

I work in a company designing and building oil refinaries. One type of our major structures are pipe-racks. When I first started to design the steel and concrete elements (based on the loads of our piping dept) the stress ratios that we were instructed to use was 80-95% (were are talking about detail design not prelliminary) since it was on a lump sum contract.BUT for ALL the structures we designed and after the procurement of the materials, we were receiving an "updated" list of loads that would exceed the previous issue at 25-75%. And of course since the available stress margin was very small and the materials already bought there was little we could do appart from crossing our fingers. Once even, the structure was already installed ( a pipe rack of 100m long and 25m high) with both in-situ and precast concrete columns and beams and of course its mat foundation.
What I learned from this first project? Always to make allowance for "unforseen loads increase of the last minute" I use larger dimension of concrete elements with extra reinforcement and the stress ratio of the steel is max60%.

The client will never guarantee the loads. But in order to justify our low stress design we demonstrate the standards for deflection. It usually works.

Vangelis