Finite Elements

[haynewp]

How many of you are using finite element analysis on a regular basis? I have never really trusted finite elements and have always used the code equations whenever possible and tried to be conservative when the codes don't quite fit the application. But I think some engineers try to use finite elements as much as they can.

I have heard some other engineers say that FEA is a method for the future of the profession. I don't know...I am just looking for any opinions on FEA in general.

 

[mrMikee]

haynewp,

I currently use a program (based on the stiffness matrix method) to do the stuctural analysis part of my design and then use Mathematica to do the code checking, not that I am recommending this approach here. I feel comfortable using computer programs for analysis but haven't been able to warm up to the code checking parts yet. You could also use general purpose FEA programs to do the same as long as you use line elements in your model.

When someone mentions FEA I think about models made from plates or solid elements. Some of these programs can solve for buckling but I have no experience with how well they work and accuracy that can be expected. The codes call for ASD or LRFD however, so I wonder what use the more complex solutions would be.

Just my 2 cents.

Regards,
-Mike

 

[haynewp]

I am also thinking of plate models and understand some people are using finite elements to routinely design things like structural slabs. I am not sure how this compares with the ACI prescibed methods for slab design??

I would think about using FEA in a situation where you have an element that is out of the ordinary (like a curved steel plate under a non-uniform load) where FEA can give you a tool to look into the behavior. I am just unsure about using it in more routine applications and how you can compare the results to what the codes stipulate, particularly in buckling situations.

 

[miecz]

I've used finite elements to derive forces, shears and bending moments in steel plates, concrete slabs and SBR tanks. I do the code checking with Mathcad. I found tank analysis by FEA to be cumbersome, and I'm back to using PCA's tables.

 

[Dinosaur]

I have used FE for some problems, but only when the complexity of the problem exceeded my ability to perform the calculations by hand. Fortunately, I can, given enough time, do FE by hand, but there are few folks I work with that enjoy this level of understanding of the FE method.

If I were solving the same sort of problem often, I would think that would be a good application for FE. But most often I see folks use FE because they don't understand the problem well enough to do it by hand.

 

[corus]

I would have thought that those who understand the problem realise the complexity and hence use FE. Of course there are those who don't know how to do a problem by hand and thus resort to using FE and producing a pretty picture. Usually those are the ones who don't understand the results.
If a design code method is applicable then it's wiser to use that. In a lot of cases, however, the code doesn't apply or a situation or failure occurs that is outside the scope of the code. The use of hand caluclations can make many more assumptions than you would with finite elements and hence the errors can be greater. I don't know why haynewp says that finite elements is the way of the future as I've been using them for 25 years. Unfortunatly the design codes haven't kept up with techinology advances and still use 'Noddy' calculations with huge safety (or don't know) factors, just in case.

corus

 

[mrMikee]

A few years back I worked with an engineer that used FEA to design concrete plant structures. He modelled beams, angles, tubes, etc. with plate elements and solved for a stress of 20 ksi. When I asked about column buckling, lateral-torsional buckling, and tension-only members, he said he had enough extra beef so that there would be no buckling. In my opinion he was a very good stress analyst but his approach (again my opinion) was not appropriate in general for building code type structures. Buckling is important as are most of the other ASD and LRFD criteria, I suppose. Being design codes they will sometimes be a little conservative.

I first used FEA about twenty years ago and quickly became impressed with the sophistication of the analysis and the complexity I could build into the model. I used line elements only, no plates or solids, and did the code checking with my own program outside of the FEA. I used three dimensional models not really because I had to, but simply to do the bookkeeping involved with orthogonal loads. Not being a highly skilled mathematician I am happy to use software for this part of the design process.

I am working with different programs today but still use the same process. Structural analysis with commercial software and code checking with Mathematica. The next step for me would probably be integrated code checking, but beyond that who knows.

-Mike

 

[jike]

I use finite element for solutions to problems where sometimes the tables that I have don't quite fit. For instance, designing a concrete wall of an areawell. I model it as a plate with the top free, the sides fixed and the bottom pinned. I load it with an out of plane triangular earth load plus a uniform surcharge. I have also modeled this as four walls but have found that I obtain sufficient accuracy with just modeling the outside longwall.

This is just one example that can be modeled in less than 5 minutes with today's software.

 

[pappyirl]

I use FEA and computer software for every slab I design. The pressure of today’s design office and the increasing complexity of structural arrangements make hand calculations mostly redundant and I find that generally the code is not directly applicable. For nice rectangular grids with equal spans and uniform loadings the codes are fantastic, and these assumptions of "equal spans and uniform loadings" are regularly stated throughout. This does not tend to happen in the jobs I do; irregular grids, opes everywhere, irregular loadings, steps, recesses, transfer columns off grid, triangular and circular floor plates etc. The numerous assumptions made to fit the design of the building around the code is putting the cart before the horse and generally results in an over-designed structure. While I realise the code can't cover every scenario, it should be more applicable to modern day FEA.

Sure there are some people do not understand FEA and believe the software will do the work for them. Part of the problem lies in the understanding of FEA. I spent 4 years solving matrices and learning the many different methods of structural analysis. Good fun at the time, but of limited use in any of the modern day design offices I have worked in. Building character and rigour was probably the intended end product. Part of the problem is that it requires effort and time understand structural analysis. These same people would have struggled with stiffness or moment redistribution methods in the days when these were cutting edge techniques. The profession, as any other, will always have lightweights.

I currently design both PT and RC slabs. I have done PT calculations by hand and enjoyed doing it. There is no doubt you get a better feel for what you are designing, although this is probably because you have sent an hour looking at it. PT design is not a quick process. Add to the equation the client who changes his mind like the weather and FEA is the only option. In my opinion FEA is not the future, it is the now. It is time the codes and universities we work to and attend recognise this fact.

 

[prost]

I realize most people tuning in to this forum work with buildings, nevertheless I think you can say this about aerospace vehicles: it would not be possible to design today's advanced air/spacecraft (in which weight seems to be the driving constraint) without finite elements. Regardless of how well Bruhn or Flabel or Petersons or Roarks formulae work, it just would not be possible to 'push the envelope' with many of the air/spacecraft designed today by relying on established, tried and true, pre-FE methods.

As far as where FE fits into our engineering toolbox, obviously it's here to stay until something better comes along, but that good engineering judgement and common sense demand that FE not be the only tool in that toolbox, and that hand stress calculations, handbooks such as Bruhn, etc. will forever remain essential tools in that toolbox.

 

[DRC1]

Just a couple of points, FEA is like most things fairly simple when broken down. The advantage of the computer is that it can perform 100,000 of these simple computations and come back seconds later asking for more. I like for routine work where if you did it by hand, you would use moment distribution or some thing similar. The biggest concern is being sure I modeled the supports and connections correctly. I think the danger with it is is that a lot of people think it will make them a structural engineer. Even in this forum, you see posts from people asking how to apply FE to problems they clearly could solve by hand in the time it took to post the question, leading one to wonder how well they understand the concept of what they are undertaking. But overall I find it a great tool.

 

[Dinosaur]

I want to elaborate a little further about the difference between moment-distribution of earlier times and FE used today. When using moment-distribution, you can make mistakes. As a matter of fact on even simple problems, until you get the sign convention straight, you can make errors. But on these simple problems, there is usually a reference that will help confirm your answer so you should be able to stay out of trouble. Even if you are not skilled at M-D when you start, you will become pretty good at it or you will just not go anywhere. Going through the M-D exercise helps give engineers a feeling for what is going on in the structure.

FE on the other hand, does not readily indicate mistakes once you have a connected model. It also doesn't help an engineer understand where the loads are going nor does it show him what is driving the problem. You can make mistakes in FE and learn very little from fixing the problem, which is unlike hand solutions.

Observations related to the high pressure of the office environment and irregular geometries are well taken. I'm not sure how an engineer gets real experience today. It is my belief that folks come out of engineering school with understanding or they don't and little analysis will be learned after graduation. Codes can be learned on the job, and plan preparation, specification writing, contracts, ... can all be learned in the office. But substantial new analysis is probably not going to happen.

I've seen a lot of work submitted to me for checking or review to scare the heck out of me. I wonder how someone can get paid to do a job with no capability? I don't think they are so cavilier to not care about the potential consequences to others and themselves, so I have to believe they don't know any better. But, that goes for hand calcs and computer calcs.

 

[zcp]

The power of fea for structural work is the ability to test 5 or 6 basic designs in the time it would take to calc 1. I often run several simplified models quickly to put me in the ballpark of sizes and locations..then check / calc to make sure it works. As was said earlier, some of today's designs can not be done by hand without making it into a full thesis. What IS required is patience and training by senior engineers for the ones getting out of school today.

ZCP

http://www.phoenix-engineer.com