Finite Element Analysis

1Q. I've heard so much about Finite Element Analysis (FEA) and I want to know if it would be worth me sending someone in my group to learn it so she can teach it to others?

1A. It depends. You haven't told me what your group does. In general, it will take a good engineer a year or two to come up to speed and be able to produce meaningful results. A course or two just won't do it. FEA isn't CAD and someone excellent in CAD also has to be very knowledgeable in engineering. Stress analysis, heat transfer, boundary conditions require a knowledge of the engineering principles. The engineer will have to start out with simple examples. If there is someone in your organization whom she can mentor with, this seems to be the quickest way. Doing examples at home on your own is also necessary. Keeping refreshed with the method is a commitment of time and training as well as the expense of maintaining the software up to date and is something you should be able to support.

We tend to send large FEA jobs to contractors who do this on a daily basis and can build models more efficiently than someone who only does a few models a year. Yes it's less costly to build a simple model in-house and run many cases, but sometimes you just don't have the time.

There is a big plus for having someone in your group learn the Finite Element Method and that is to understand what the consultants are doing. Usually no one knows your equipment better than you and if you can explain it simply to a consultant it can greatly reduce the cost. The simplified axis-symmetric model can easily cost one tenth that of a full 3D model, and can be easier to understand.

I was once asked how to make a CAD technician knowledgeable in heat transfer, stress analysis, vibration analysis, and strength of materials. I told the manager to send him to get his Mechanical Engineering Degree.

2Q. How do you know if the answers you get from a complex FEA are correct?

2A. Well I'm never 100% sure and this is where engineering experience is valuable. As one of my experienced colleagues use to say, "If it doesn't pass the funny look test, it's probably not right". I usually run a simplified FEA, or simple closed form solution to check the sensibility of the results. I bounce it off of others and watch their faces. I compare the results with things I have witnessed in the field or in technical papers others have done. I always try to get data after the FEA results have been used, meaning has the design worked after say two years. I store that as a successful run. I always add a considerable margin to the design, since the heat transfer coefficients, boundary conditions, material properties, modeling technique FEA simplifications, etc. are always estimates too.

3Q. We want to design a new product and have read so much about FEA. Is this the way to go?

3A. FEA doesn't design products, Engineers do. FEA is only a tool. You can design a product without FEA but you can't design a product without good engineering and experience. FEA saves money by allowing designs to be tried on the computer before any manufacturing is done. It's a great tool especially when you are trying to sell to a customer, management or refine a design. However anything totally new is usually driven by someone experienced in designing successful products and is usually one person with the concept and focus. The rest of the team helps and finalizes the design. Very rarely has it been the Specialist in FEA who comes up with a new product.

4Q. What is an axis-symmetric model and what good is it?

4A. If you draw something in two dimension such as a rectangle and it can be rotated about an axis to form say a tube then it is axis-symmetric. This allows a finite element modeler to draw the mesh on the rectangle and "trick" the FEA into thinking it's the full tube. Actually the computer code puts restrains on the model to act like a section of a 3D model. The benefits are that if this can be done for complex models there is a great saving in time and expense. For transient coupled heat transfer stress analysis it is especially useful. Many different boundary conditions can be tested with little work. In addition sometimes the models are easy to visualize where 3D models can be more difficult.

5Q. Do you use finite element analysis for modeling?

5A. Most of the FEA I do is like the analytical analysis I do. I reduce the equipment into a simple system and then analyze it. This has been out of necessity since when working to solve plant problems you don’t have time to dwell too long solving a problem. You need the best answer to address the problem so the equipment can be put back into production. When I can’t solve a problem with calculations, I use FEA as in a transient temperature problem and the stress, displacement, loads and moments they produce. Not easy to solve analytically but relatively simple with FEA. Axis-symmetric,  ¼ or ½  symmetry, plane stress and plane strain are some simplifying techniques to use in FEA. I do very little complex 3D FEA, because I’m not fast enough and don’t have the time. I go to an outside shop when I can’t simplify it to 2D.

6Q. What is CFD?

6A. It stands for computational fluid dynamics and it is also FEA but with fluids, meaning liquids and gases. I've used simple models to solve fluid flow pressure drop problems such as shown in the Tech Talk section. When problems get complicated I always utilize experts in this area than try to do it myself. I simply don't do or know enough of it to stay proficient.