Ever since I started working with simulation, there has been this dispute between the test engineers and the simulation engineers.

The test people claim that the simulation is not accurate enough, some simulation advocates claim that there will be no need for testing in the future and everything can be validated by simulation.

Both argue about an apparently limited part and are afraid that the other side will take this part away from them.

Let us take a closer look at both sides, while we are already on the important topic of accuracy.

If I want to determine an error, what am I referring to?

Simulation

The real geometry of the components usually deviates from the CAD models, but this already causes errors in the geometry of the CAE computer model.

Although the physics are increasingly better represented in the CFD software systems, it is an approximate solution. In addition, there is the influence of the boundary conditions, which are also often idealized.

Experiment

But the test also has its pitfalls:

The gaps of a compressor alone sometimes deviate considerably from the CAD model, but have a serious influence on efficiency and characteristic curves. The measurements are taken at specific points, and even the determination of the volume flow is an averaged value that assumes that the flow conditions are constant across the pipe cross-section. Measurement inaccuracies and measurement errors are also commonplace.

Discrepancies

Often we see with large discrepancies that the model we investigate in the calculation and the test model for the test have large deviations in the geometry alone because other components were used, e.g. another diffuser was used in the test.

But it is not only in the area of fluid that discrepancies between test and simulation occur.

In the case of finite element calculations, the material behaviour is already an assumption that was determined at some point on the basis of (faulty?) tests. The material characteristics scatter with the batch used and are only minimum values.

Order of magnitude of the errors

My many years of experience have shown me that the errors both in the experiment and in the simulation are of about the same order of magnitude. 3-5% is a good value in both cases.

Both test and calculation have their strengths and weaknesses.

For both we also assume a scenario which is also only a conservative(?) assumption for the use of the components in real.

The Problem – Communication

The problem is that test engineers and calculation engineers speak different languages.

Therefore, as simulation engineers, we prefer to look at the test setup in real, if possible.

Often, as in a concrete case, the test specimen of an electronic system, which we have assumed in our simulation model to be an aluminium die-cast part (data from the CAD model), turns out to be a milled part made of pure aluminium. Unfortunately, the thermal conduction coefficients here differ by a factor of 2, so it is no wonder that there is a difference of up to 40% in peak temperature between simulation and test.

Conclusion:

Back to the original question: Both are right.

Experiment and simulation complement each other. Both are flawed, both have their strengths and weaknesses.

The experiment often serves as a proof that there is no getting around it.

The simulation is faster and cheaper for variants and optimizations, offers a better insight into physics … and needs material data and fluid data that come from the experiment.

If we use testing and simulation in the right place at the right time and make sure that both sides exchange information without bias, we can develop new, safe components faster and more efficiently.

Yours Stefan Merkle

PS: Have you become curious? Find out in a non-binding, free and personal conversation how we can also support you in the development of your products.

Get in touch with us.