When conducting a search on Google, have you ever wondered how quickly you get to what you are looking for? Most often, it is the first two to three attempts. At other times, it might take a good while before you can even get close to what you need. But it is quite likely that you do not get it right the first time unless you are absolutely sure about what you are looking for.
What you put in the search field is what you really get out of the corresponding hits.
The same holds true for any simulation. The success of a good result from FEA is hugely dependent on what you feed into the computer. If the user faithfully translates all physical operating conditions onto the virtual environment, the chances of getting good results in FEA is extremely high.
All simulation results, irrespective of the knowledge and experience of the user, are only as good as the following assumptions:
So how can one ensure that they are setting up the problem correctly? In other words, how can one get accurate results the first time?
A little known trade secret in this industry is that the only difference between an expert and a non-expert is that the experts remember to ask all of the above questions.
The truth about FEA is that it is seldom the software as much as the user that can cause it to go wrong. The algorithms used today for FEA or CFD have been evolving over decades and have been honed and fine-tuned to suit numerous problems over a wide range of industries. Mathematical establishments such as NAFEMS conduct rigorous examinations of the validity of the codes using a plethora of benchmarks. While the software is now at a stage where the outcome is barely affected after the setup, it is the actual setup of the problem itself that poses the biggest challenges.
So what is accuracy in FEA anyway?
In reality, claims of accuracy or correctness in FEA are usually blown out of proportion. The value of most FEA done in product design is from the insight that the data provides, and not the hard numbers. In fact, hard numbers only lends itself as a distraction from the bigger picture. The difference in stress or deflection patterns between multiple designs can be much more useful than the actual value of the stress or deflection.
It is important to keep in mind that FEA results, at best, are representative of a controlled environment. It can provide extremely valuable insight on likely deformation, critical strength locations, causes of problems, and possible improvements.
What are some tips for success in design validation?
A trained user of Simulation tools would tend to focus more on comparative validations rather than true numbers. However, while doing so, it is critical that some key factors are kept in mind:
Keeping assumptions consistent across iterations, and models
Making sure that the inputs are reasonable and represent field conditions as closely as possible
Performing sanity checks, especially on the first few iterations (can the model really move in this direction? Are the magnitudes realistic?)
It is also important for the user to explore many iterations to gather as much insight as possible on new and existing designs.
Keep in mind that it costs nothing to experiment on the screen. But physical testing is still the final decision maker for design validity.