Direct Stiffness Method 3

[abusementpark]

This is a question for those who have learned the direct stiffness method and how the structural analysis programs work. Did you gain any real practical knowledge from it? I am curious because I am taking a graduate level analysis class on this stuff in the fall and I have a hard time seeing how much it will truly help me.

I have no ambitions to write my own FEM code. It seems like it is just trivial knowledge otherwise. Yes, I understand it is nice to know how the program assembles the stiffness matrices and solves for the unknowns, but I don't really understand how this will make me a better structural modeler. I sure there is some value to it, but I just don't see any offhand. I'd appreciate feedback of any kind. Thanks.

 

[271828]

I assume that the grad level class will include some advanced topics like buckling, small displacement 2nd order analysis, cables, etc. There's great value in understanding how the programs do these things. People mis-use these feature very frequently when they don't understand the underlying assumptions and methods.

 

[civeng80]

This is an interesting question.

I did the direct stiffness method at school and the Professor spent a lot of time on it. To be absolutely honest I found it wasn't really worth the time we spent on it. Understanding of the meaning of bending moment, shear force, and second order analysis, end moments, etc is probably more important than the method. Having good structural sense such as approximate solutions to problems is more important so that a check can be done on the output files.
You dont have to know how to set up a stiffness matrix to be able to use a sophisticated structural analysis program.

 

[OzEng80]

I did a fair bit of FEM theory and associated calcs at uni and would rate the exercise as not worth the time and effort required. You can gain as much understanding by reading some general FEM documentation and interrogating your software manual (and support staff) for answers and advice for the essential items listed by 271828. I would rate knowing how your specific software program works much higher than a detailed understanding of the history of FEM and the methods employed by each program.

 

[271828]

Is there a reputable university anywhere that would let somebody get a MS without taking a course in the DSM (Matrix Structural Analysis)?! I'm familiar with several universities and I am completely confident that none of them would allow someone to squeak by without it.

You guys are typing about being a "technician," or "button pusher," not a MS-level structural engineer. That's a road that shouldn't be gone down.

Sorry--that's just the way I see it.

 

[rb1957]

completely agree with 271828 (your locker combination?)

 

[Skicat]

I wouldn't be so fast to dismiss the benefit of a Matrix / FEM class. I took one in grad school and it was one of my best and favorite classes. I use SAP2000 and Etabs every day and I feel much more comfortable using those programs now that I know what processes they use internally. I feel that I have an advantage over my co-workers who haven't taken an FEA class when using the programs. The math isn't any fun, but to gain a different and in my opinion useful perspective on the behavior of elements, frames and structures is very beneficial.

 

[271828]

rb, it's an inside joke, and a fairly poor one at that. My little girl is adorable and her name starts with "E" so I started calling her "Little E" which of course became e. EngTips' stupid username limitations wouldn't let me have a decimal!

 

[Lion06]

I had a class on matrix structural analysis at the undergrad level - and my degree is an ET degree....... go figure!

 

[Dinosaur]

I think a class in this is essential before someone should be turned loose with any stiffness program. I hope you will learn the sign convention used by these programs, the difference between an element that includes shear or does not include shear deformations, the fact that it takes special coding to accommodate buckling and plastic analysis if you need it and other "under the hood" checks that should be understood.

I believe an engineer could go his entire career without working out a single integral. But I'd never want to work with one that did.

 

[prost]

Why would any university force you to take a DSM (Matrix Structural Analysis) class for your MS (I am referring to mechanical or aero engineering here. I don't know what Civil engineering MS students take)? Many universities are quite flexible with their class options--you might have to take 2 'structural mechanics' classes (in addition to math and fluids and thermal), but the list of structural mechanics is often quite long, with classes in finite elements, fracture mechanics, elasticity, plasticity, and continuum mechanics. Now it may be a GREAT idea to take DSM or MSA classes as you have suggested, (given the ubiquity of these methods in current engineering practice), but why force that class down someone's throat (more or less)? People take many different career paths upon graduation, and many may not see the need to understand DSM or MSA theory so thoroughly.

 

[OzEng80]

I don’t think anyone has claimed that undertaking any course or form of education of any kind is a complete waste of time.

I believe there is a vast difference between being a button pusher and being a button programmer. I agree that is essential to know the in’s and out’s of your program and it’s concepts.

I may be speaking from my own experiences with an impractical FEM course but I believe that time spent grinding through heavy calcs and theory is not going create a better engineer than equivalent time spent researching and understanding the practical application of the theory developed by others.

With the vastness of knowledge, training and limited time available to most of our profession, a focused and practical approach has to be taken to training.

I admit it would be nice to know everything though.

 

[Denial]

271828

Given your little inside joke concerning your nom-de-EngTips, you should read Richard Feynman's delightful autobiographical book of anecdotes "Surely You're Joking, Mr. Feynman". In particular the anecdote titled "Safecracker Meets Safecracker".

 

[NS4U]

I have an example

I have an MS, I took both matrix structural analysis and FEA (DSM & Structural Dynamics were the only required courses for the structural MS, the rest are electives). I work for a company that has two divisions, standard structural design (what I'm in) and high end analysis (mainly done on FE software).

I am currently working on this project which requires both design/detailing and high end analysis. Part of my job as the designer is to do a hand analysis and design, which is then checked in FE and more refined results are given.

The other part of my job on this project is to check the FE results. This would have been IMPOSSIBLE for me to check if I did not know things about ill-conditioning, stiffness matrices, integration points, element types, mesh sizes, local FE effects and numerous other topics, all of which I learned in my grad classes. Now, we are all human... and I did find quite a few discrepancies and was able to intelligently discuss with the engineer who was doing the FEA and get them resolved. If I didn't have this experience they would have had to bring on another guy to check the FE stuff, or I may not have gotten on this project all together.

Not to mention you should never use a program in which you don't know enough about how it's arriving at the results, it's very poor practice in my opinion.

 

[271828]

OzEng80, say a guy reads the manuals and does his research on the job, but has never had a class. Say this guy wanted to use rigid links in his model. He's he going to know that the program is going to give strange answers if he makes the rigid link stiffness too high? Should just be able to make A, I, and E arbitrarily large and nothing should go wrong, right? (Just one example, right off the top of my head.)

 

[OzEng80]

I am pretty comfortable saying that everyone in this thread knows more about FEM modeling than me.
There are inevitably going to be situations during the modeling process which require hand verification, experience and a ‘gut feel’ for confirmation. I personally pepper the programmer with questions every time I have even a glimmer of an inconsistency, anomaly or unexpected result. I have found that this as a resource, has been far more beneficial to my development than the tedious, demanding and mostly forgotten FEM course I undertook. It is obvious that different courses will offer different levels of practical material – speaking from my experience, I gained little – and certainly not practical tips on ‘how’ to generate a model or what to what watch out for.

271828 have your skills been honed by your graduate level FEM class or later software and design courses, experience and gut feel? Again, all education has merit but with time being such a precious commodity I believe any training regime should be directed to the most beneficial and specific skills required.

 

[abusementpark]

Thanks a lot for all the responses guys. It is interesting to get the mixed opinions. Usually at the beginning of the semester, professors allows students to weigh in on the syllabus. Are there any specific things that I should push for in the class to make it practical for someone who is a practicing engineer?

 

[271828]

OzEng80, I believe it has, but it's hard to be objective. I think I spot a lot of issues that I would be a lot less likely to spot otherwise.

As full disclosure, I'm not in design any longer. I had a DSM class and a FEA class during my MS program and I think they helped. I've since taken additional coursework, including 2 other DSM classes. A LOT of this, for me, falls into the "Dang, I wish I knew that when I was using these programs!!" category. Things like virtual work optimization, buckling, response history, different types of nonlinearities, etc.

I admit that this subject is debatable. IMO only, at least one class (DSM or FEA) should be taken by anybody with a MS. OTOH, a curriculum only has a certain number of classes and there might be something else that is more valuable.

Whether programming should be a part of the class is also debatable. I *think* that could be dropped without hurting the class. That being typed, in my last two DSM classes, we had to program everything, all the way up to cables, buckling, etc., and I do believe that increased my depth by quite a bit.