[nezmat]

Is negative results omission considered falsification?
For example, you obtained positive and negative results when performing the same CFD experiment with old and new software versions, respectively. Is it unethical to present/publish the positive results only?

[wyldckat]

Greetings nezmat,

Mmm... it depends.

From one perspective, here is an example:

1. A colleague runs a simulation with version 2016A of a CFD software in 2017, and reported that it gave good results and is consistent with experimental data.
2. You then try to reproduce the same results with version 2018B in 2018, but the results are not the same and it doesn't make any sense nor it is related to the experimental data.
3. Therefore, the two main unethical problems would be if:
   1. The colleague that ran with 2016A lied about the results and/or instead simulated with another CFD software or another older version.
   2. Or if you re-run the simulation with 2018B, see the bad results, get 2016A and run and get the good results... then report that you ran with 2018B and state "yeah, yeah, 2018B reproduced the same results..."

When it comes to software, changes to the code may improve things in most cases, but the implementation could either render some types of simulation no longer possible or perhaps the implementation requires a different set of boundary conditions in order for it to work properly.


For example, with OpenFOAM, back in version 2.3 was introduced an improved buoyancy handling methodology that made old multiphase cases no longer work properly when they were configured to work with OpenFOAM 2.2. But once people readjusted the boundary conditions to use the new convention, the results were either the same, improved or possibly made worse due to stability issues that were not as well handled in specific situations (sometimes mesh related or discretization schemes needing adjustments). Essentially, the "problem" was associated to how the pressure equations were being handled in each version and how buoyancy fitted into the whole implementation.


But then comes the other perspective when it comes to these situations: If the negative results are meant to be reported on an article/paper to be published, will any peers accept it? Or will they disregard it and simply say "you're wrong"?
Because CFD software should be validated before it's made public and/or sold, therefore if it properly passes the validation standards, then either the manufacturer lied or you've made a mistake in setting up the case...

Best regards,
Bruno