Checking calculations 19

[P205]

I'm writing this thread in regards to what was said in this old thread here:

http://www.eng-tips.com/viewthread.cfm?qid=321928

It leaves me feeling uneasy to say the least. I also feel compelled to ask the following question:

- In the interest of meeting the deadlines AND feeling comfortable, why not be generously conservative in most of your designs?

- Many admit to finding errors often in their own work and in others. It stands to reason that not all errors get discovered. This would lead me to want to be conservative in my design.

- Another thought, given the old triangle (cost-quality-time), if deadlines are very rushed (low time) and the structure simply needs to be satisfactory (low quality/less refined), then being conservative (higher cost) appears to be the way to solve this puzzle.

Thoughts?

---

Personally, I have 8 years of design experience, and spend a lot of time reading books and teaching myself as much as I can. I usually buy a few books a year to refine my skills and knowledge. Honestly, I don't see too many peers doing that. I'm just looking to stimulate some discussion.

 

[Teguci]

We engineers are truly an arrogant lot!
Yes, mistakes happen and no, we can't (pre-)fix them with a safety factor (although, after a sleepless night, those extras came in handy - also note: sleepless because I checked and found a mistake). My strategy, when up against a schedule, is to simplify and envelope the design. As an abbreviated example, if I have a distributed load, find the total load and apply it as a point load at center span in order to size a beam. An enveloping strategy like this is not a mistake, often meets the client's requirements, easy to check and good in a quick pinch.

Because we all have to sell our services, its important to manage the clients perceptions. The obvious downside to enveloping like this is the potential for a poor reputation as an over-designer. However, most of the projects I deal with today are happy with an aggressive schedule over material costs.

 

[Tomfh]

Yes, mistakes happen and no, we can't (pre-)fix them with a safety factor

I'm still not sure why not.

Whenever a safety factor catches someone's underdesign (as opposed to catching substandard material) has it not fixed it?

And the bigger the safety factor, the bigger the issue that can be caught.

 

[PEinc]

"And the bigger the safety factor, the bigger the issue that can be caught." The bigger the safety factor, the bigger the project's cost. Clients who hire design engineers expect the engineer to know how to properly and economically design a structure according to code and project requirements. Pumping up a safety factor, without a really good and understood reason, in order to cover expected mistakes doesn't seem right to me. If a designer is expecting to make mistakes, maybe more training or schooling is needed. If you expect the contractor to build it wrong, you have the wrong contractor. If you expect the inspector to do a poor job, you have the wrong inspector. Design it right; build it right; inspect it right. I know it's not always that simple, but try?

http://www.PeirceEngineering.com

 

[P205]

If you expect the contractor to build it wrong, you have the wrong contractor

Tell me, how many projects have you been hired for your engineering services where you've also been tasked with personally choosing the contractor of your liking. How many developers/businesses/institutions call you up and ask you which contractor they should hire?

If you expect the inspector to do a poor job, you have the wrong inspector.

You've used the term "inspector" several times in your replies. In Canada (Ontario), our professional organization strictly avoids using that term because it has legal implications that are far beyond what is expected as the engineer of record. Using the terminology used by the professional organization I belong to (PEO), which is also the terminology used by contract lawyers who teach seminars to engineers on providing professional services, we as the engineer of record perform periodic reviews of the construction work to ensure that it is in general conformity to the plans and specifications.

My point being that we don't get to see everything the contractor does. We have to make the drawings clear and easy to understand and build. Having 30 different column types that are designed just right, and most economically, is a construction nightmare, likely to lead to mistakes by the contractor. Lawsuits happen all the time, and even if it's 100% designed right, built wrong, you end up having to share some of the costs/liabilities.

--

A lot of people in this thread have shared examples of situations where they use some judgement in being on the conservative side. I appreciate everyone's input and I have gained some valuable take-aways. Thanks everyone.

Edit: spelling

 

[SwinnyGG]

I'm shocked by the arrogance in this thread.

If safety factors aren't explicitly designed to cover possible mistakes (by either the engineer or contractor or possibly both) then I would love to hear why people think we use them for everything.

 

[Retrograde]

If safety factors aren't explicitly designed to cover possible mistakes (by either the engineer or contractor or possibly both) then I would love to hear why people think we use them for everything.

Doug Jenkins (a regular here on Eng-Tips) wrote a paper on limit state design principles. To quote:

"Statistical analyses used to calibrate code load and resistance factors do not include significant sources of risk, such as failures in the design and construction process, or unforseen events after completion."

 

[Tomfh]

Statistical analyses used to calibrate code load and resistance factors do not include significant sources of risk, such as failures in the design and construction process

That's a worry, because if it's true it means we are completely ignoring the risk of design and construction deficiencies, preferring instead to rely on a fantasy of making zero errors.

 

[racookpe1978]

Worked on the interference analysis of a USN man-rated mini-submarine a few years ago.
Budgets were tight, the boat was overweight, the spaces and clearances were "negative" = Everything hit everything when you tried to replace it or remove for maintenance or even to install the internal comuters, radios, wires, fans, crypto gear, HVAC and atmosphere controls, etc.
So, in the middle of everything else, we had to put in four man-rated submarine pressure hull hatches.

Turns out the only "acceptable" hatch was from the deep-diving submarines because they were certified for underwater explostions because they had been tested post-WWII on diesel submarines, so they were apporved for the newer nuclear submarines so they were approved for our little shallow -water submarine. Get the logic?

Turns out the fabrication dwgs were from 1932. Had never been updated from the "made a wood form and cast it in the foundry" era.
Were the four hatches too big, too thick, too heavy in some places? Ansolutely.
Where were they over-built, and where was the steel "just enough" and where was it "never failed yet" ... but was still undersized? Don't know. Can't tell.
But we could not get a "new lightweight hatch" designed and approved, even one that would have been designed in 3D with CAD/CAM certs and a finite element pressure check.

Typical of that project. Yes, it failed. Too expensive, too heavy, not enough power to do the intended job.

You've GOT to do your best design, and use a second-check to verify YOUR assumptions and YOUR approximations. Else the project fails because it becomes too expensive to be built, to be run, to operate as needed. Or, like the nuclear power porgrams, they get managed to death without adding the ultimately needed levels of safety and reliability.

Because you CANNOT truly know what your design will actually face.

 

[271828]

jgKRI wrote:

I'm shocked by the arrogance in this thread.

If safety factors aren't explicitly designed to cover possible mistakes (by either the engineer or contractor or possibly both) then I would love to hear why people think we use them for everything.

I am not familiar with other standards, but in the AISC Specification, the resistance factors do not account for human errors in design.

From Page 81 of Galambos 1981 AISC Engineering Journal article, "Load and Resistance Factor Design," which is the basis of modern LRFD in steel: "The reliability is to be interpreted as 'notional,' i.e., it is a comparative concept. It should not be confused with actual structural failures, which are the result of errors and omissions. Only the national statistical variation of the parameters is included, and, as in other traditional specifications, human errors must be guarded against by other control measures."

Part 2 of the AISC Manual spells out what sources of variability are included. Design errors are not in there.

Here's how LRFD phi factors are determined:
1. Test as many specimens as possible to get measured failure loads. Dig more out of the literature if possible.
2. Select a predictive model. Predict the failure load for each specimen.
3. Compute the ratio of measured-to-predicted load for each specimen.
4. Use reliability calculations, such as in one of the SSRC Guide appendix, to determine the phi factor.

Nowhere in that process is a factor for human error, other than in the development of the predictive model. Errors in the predictive model are deficiencies in the best current human understanding of the failure mode. These errors are quantified in the process listed above. They are not arbitrary. These factors are not intended to help with potential errors -- of totally unknown magnitude -- that designers might make. How would one include those types of errors in a reliability calculation?

I am sorry, but using an arbitrary bump-up is an awful practice. Errors most certainly will occur. They must be dealt with some other way.

 

[SwinnyGG]

In my mind the terms 'load factor' or 'resistance factor' or phi from LFRD are not interchangeable with the term 'safety factor'.

That's a worry, because if it's true it means we are completely ignoring the risk of design and construction deficiencies, preferring instead to rely on a fantasy of making zero errors.

I couldn't agree more.

 

[Guastavino]

"Statistical analyses used to calibrate code load and resistance factors do not include significant sources of risk, such as failures in the design and construction process, or unforseen events after completion."

Yes, but the structure doesn't know the difference between being overloaded because of a code limit load and being undersized for a more realistic maximum load. So in the words of a wise sage "I'd rather be lucky than good."

I am sorry, but using an arbitrary bump-up is an awful practice. Errors most certainly will occur. They must be dealt with some other way.

What specific way do you deal with them?

 

[JAE]

The derivation of the phi factor is aligned with what 271828 lists above -
I might add that this includes, within the reliability methods, all the individual parameters that contribute to the capacity of the structural member.

For a concrete beam, those parameters would include concrete strength, rebar yield, depth to rebar, width of beam, and the equations used to determine strength.

Each of these parameters has a "natural" variability in itself and the phi factor is developed based on their variabilities plus calibration with an expected probability of failure.
The "natural" variability of the parameters, including the predictive equations, creates the phi and none of this involves egregious errors.

However, when combined with load factors, you have a margin of safety now built in that does in fact cover some errors as long as the errors are not excessive.

In some cases, we engineers find ourselves telling owners and contractors "the structure doesn't work under code" while they are scratching their heads looking at a structure that has stood for several decades. Been there many times.

So I would suggest that small errors are many times unintentionally covered by the load and phi factors while larger errors are not intended to, and can not be, covered by the level of built in safety.


Check out Eng-Tips Forum's Policies here:
faq731-376

 

[PEinc]

I agree with stiman86's comments about inspection. However, it is common and usually code-required that companies be hired to do on-site inspections or special inspections. These special inspectors are supposed to be hired by the owner or owner's rep. It is the special inspector's specific job to make sure construction is done correctly, keep records of their inspections, and to notify the proper authorities or entities if there are construction problems. When I said "you picked the wrong" engineer, contractor, or inspector; I was referring to whoever does that hiring, not necessarily and not usually the design engineer. By the way, I have many, many times been asked to recommend contractors to my clients. I always recommend more than one contractor so that my client has a choice from several contractors that I know are experienced and capable of performing the specific work required.

EDIT: P.S., in New York City, the design engineer of record has to sign a form that he or she approves the chosen special inspector.

http://www.PeirceEngineering.com

 

[271828]

njlutzwe wrote:

What specific way do you deal with them?

Depending on the situation, one or more of:

1. Knowing from experience approximately what I should get in most cases. Bird's eye view of the sizes on the plan.
2. Check my primary calculation method with standard tables, textbook examples, etc.
3. If a result looks weird, break down the problem into a simplified version which should give a similar result.
4. Peer reviews.

We all make mistakes. Our work is too complicated to pretend otherwise. I just disagree with the particular method described in the OP. I am sorry for being too forceful in this thread, though. Maybe I need a load factor.

 

[Guastavino]

@271828

I agree with your methods of checking on things. I often do quick dirty calculations and am surprised by how close they are to a complex RISA-3D analysis complete with semi-rigid diaphragm load distribution.

Regardless of the exact interpretation of the OP, i think there is more common ground than uncommon...

Another thing to consider is the fee/schedules of typical design shops. The adage “you get what you pay for” is so true. And, i think that if clients were willing to pay more and give more reasonable design schedules, engineers could save them time and money by spending more time designing with a sharp pencil. For example, connections with the UDL. That is typical, but we could fine tune that, especially for short beams, but we often don’t.

Now, with that said, I think that if the fee was $10,000 or $100,000, with established practices, you would probably get about the same product. In other words, if clients were willing to pay more, I even wonder if we as engineers would give them a better product, or if we would revert to the current design methodology we follow.

Ultimately, what i’m Getting at is that every engineer could sharpen the pencil more than we do. And the level that we do is based on judgment, experience, and ability to sleep at night (Both structurally and ethically). So it’s something to hold in tension. Folks like 271828 make us better engineers by encouraging us to sharpen the pencil. Meanwhile, Stiman and company make us more efficient. I think there is a time and place for both, and the tension is healthy.

Just my two cents...

 

[Tomfh]

In some cases, we engineers find ourselves telling owners and contractors "the structure doesn't work under code" while they are scratching their heads looking at a structure that has stood for several decades. Been there many times.

Here's one I sometimes wonder about. Let's say we design a steel beam to support a 200mm thick concrete deck. We factor the dead load by 1.2, and reduce the steel beam's capacity by the relevant phi factor. Assume the strength ratio comes out to be 0.98, the beam is working hard, but has not exceeded its capacity. The contractor pours the slab, and actually pours it 220mm thick. Our original load factor of 1.2 covers this 10% variance in dead load - the load factor has done its job.

But say a new engineer comes back and measures and checks the structure from scratch - the beam doesn't work anymore - because they're now factoring the actual slab thickness, which is larger than the slab thickness in the original design.

 

[IJR]

Sorry for joining late.
I vote not to overdesign, instead:

1)Search hard to see if there exist already standardized stuff: a)Some place somewhere somebody solved a similar problem, executed it and catalogued their work, b)German engineers steel use standard tables the way engineers did in the fifties and sixtees where competent designs/methods were sought for and tabulated. Modern software killed this great tool.

2)Sharing designs/methods : I asked a friend for design of a huge spiral staircase which he knows worked, I modified it to make it more economical and remove errors, built it and passed my design back to him. Yes sharing works. Modern I-am-the-best approach kills sharing.

What do you think?

ijr

 

[csd72]

Most serious failures do not occur because the member is not quite strong enough, most occur because of something fundamentally wrong with the structural scheme or with the way it has been installed.

Conservatism will not negate this only checking of the design will negate this.

Even with preliminary schemes we do what we call a 'sanity check' i.e. we get one of our experienced colleagues to run their eyes over the job and see if it all looks right to them. This can take as little as 5 minutes for small jobs.

I have checked work from graduates to those top of their field and almost ready to retire. I have picked up errors in all levels although the later end errors are generally a lot more subtle.

 

[kipfoot]

Here's an excerpt from a book that shows the concept of loads and structural capacity graphically.

The book is Load and Resistance Factor Design of Steel Structures, Geschwindner & Bjorhovde, 1994

 

[ldeem]

IMHO you have to consider the life cycle of the structure. I work in power and mining industry and almost all structures in the field have extensive section loss from corrosion just due to the conditions. I have also seen several very questionable repairs. To compensate I favor fairly conservative design just knowing how things will be treated during their expected life.

As a second point I have done several projects where I need to go back and check an existing structure for additional loading. It gets very expensive for the client when they need to add 10% more weight but the original structure was designed to close to IR=1 that now they need to additional steel work.