How to face the first finite element simulation of a phenomenon that you have never studied

Finite element simulations cover a wide range of phenomena and possibilities. As a CFD / FEM engineer it is highly likely to master certain types of simulations, but not to know many others in depth. Either due to the need to tackle a new type of project or to expand our own knowledge, sooner or later it will be essential for us to learn to develop new types of simulation. To make the process as light as possible, in this article we will leave some tips that will allow you to overcome this task.

Define perfectly what your problem is.

To simulate is an extremely broad task. A simulation can be done to study a specific variable at a certain point, to observe the field of different variables on a specific surface or volume, to analyse the temporal evolution of a variable, etc. Focus on what your objective is, perfectly defining which will be your inputs and which will be your variables to analyse. If it is a phenomenon far removed from the ones you usually study, you must begin with a basic problem in a steady state of which you analyse one or two variables in some specific region, for example, the output of the control volume. Although the final objective is more ambitious than this (and sometimes it is not even), once we validate this first simulation, we will have enough guarantees to develop more complex challenges.

Create a simulation that you can verify.

Surely, this point is the most important of all. When setting up the simulation, it is vitally important that you can compare the result with some real data or approximate analytical solution. It is useless to invest hours in simulating if once you have finished you do not have any data that allows you to guarantee that the result is correct.

A finished and correctly converged simulation only means that the case is well configured, but this does not mean that you have reproduced the problem that you wanted to study. If you have experimental data and the solution does not match, you can investigate where you have made an error, or even check that this type of simulation requires more complexity than the case you have prepared. Of course, you must make sure that the experimental data has been measured correctly or that the approximate solution is valid, since your simulation may be correct and you are looking for inexistent errors.

If you are not able to obtain any of these data, but you are interested in knowing how the case is configured for this new type of phenomenon, you can do the simulation in two different software and verify that the result agrees.

Look for specific information.

It may seem obvious, but the truth is sometimes, and due to our own knowledge and experience, we try to develop everything by ourselves. If you want to study something, someone has probably already studied it. Nowadays it is extremely easy to find information related to almost any problem on the net in scientific articles, video tutorials or personal blogs. But be careful, you could not see the wood for the trees. If you are truly clear about the phenomenon to simulate and your target variable, use these words as key search terms.

Evaluate which software is the most recommended to do it.

While doing the search on the networks, you should also investigate about which software offers the greatest advantages to solve the simulation. There is software that has faster and specific modules to solve specific problems. Obviously, this will depend on the availability that each one has, but it never hurts to know which the best software is to do something specific.

Evaluate the degree of complexity your simulation requires.

This can be difficult the first time you study a phenomenon, but it is important to keep it in mind. A common mistake is to think that the more detailed and complex we build the model, the better, but this is not necessarily the case. The degree of complexity required for a simulation uniquely depends on the objectives. To make it better understood, we will put a simple example.

Suppose that we want to solve a metallic structure formed by different types of profiles, of which we are only interested in knowing the maximum deformation of a certain profile, for basic engineering. We can build the 3D model of each of the beams, mesh it, solve it and evaluate the deformations and stress levels. We will almost certainly face two problems in this process: performing a correct meshing and the presence of singularities when evaluating the results. Problems that, although we can overcome them, will take us a significant number of hours (and headaches). For a basic engineering of this type, there are software that allow to build these structures in 2D and assign the profile shape to each beam (standard or graphically defined by us). The 2D mesh is done almost instantaneously and the singularities disappear, obtaining a good result in a noticeably short time.

Final advice.

Finally, it only remains to add that when carrying out a simulation we should not be mere users of the program in question. In other words, when we carry out a simulation, we must be completely sure that we understand the physics that occurs behind the problem. For example, if we are trying to solve a heat transfer problem and we want to determine the temperature on a surface, we should additionally verify that the simulated heat fluxes make sense, if the LMTD obtained at the entrance and exit of the control volume are possible, check that the fluid patterns follow the expected paths…If you solve the problem, but have not done all of the above, you will only have gone half way since you may have made mistakes without even realizing it from them.