I'm currently working on a dynamic simulation of an already existing product. The simulation is supposed to help in future development of this product by enabling the engineers to get a better understanding of the influence different parameters have.

However the mathematical and physical models manifesting the simulation are only simplifications of the real system because of its sheer complexity. Therefore no quantitative results are created. Nevertheless this simulation is useful as it is faster and cheaper to change a parameter in a simulation rather than building a prototype and physically test it.

Now being at the point that the simulation runs as expected with acceptable performance I'm starting to implement the evaluation of the data.

Which gives me the problem: What am I even supposed to do with all of this data?

The Question now is: What are useful methods to analyze moderate sets of data of real world (mechanical) systems.

My discipline is hardware development and not data analysis so this topic is mostly new to me. (Apart from outdated mathematical lectures about statistics). Additionally I'm really junior in this industry so experience is rather limited.

I think this question (or more precisely the answers) could be a good resource for engineers in the same positions as me independent from the application.

A bit of background for my application:

  • 40 static inputs/starting parameters
  • 30 outputs presenting physical properties "plotted" over time (e.g. speed, height, current, etc.)
  • Monte Carlo method is already implemented

This question might me considered as to broad or not answerable, because of this I'm open for feedback or accept the fact that it might be closed.

  • 1
    $\begingroup$ Start with some graphical presentations, maybe present some of them in the question. Try to make the Q more explicit, tell us more details. And maybe most important: to get better data, look into statistical design of experiments, see stats.stackexchange.com/questions/162984/…, look through the tag experiment-design, get a copy of Statistics For Experimenters and read it! $\endgroup$ Jul 24 '19 at 12:04
  • 1
    $\begingroup$ @kjetilbhalvorsen I'm hesitant to share more explicit information about my application as I'm doing this as an employee (with more than one nda) for a company and not as a researcher. But the sources/references you provided are definitely helpful! $\endgroup$
    – GittingGud
    Jul 24 '19 at 12:10

I normally relly on Discrete-Event System Simulation by Jerry Banks to perform statistical analysis of simulation outputs. The topics covered in Section V include:

  • Modelling simulation inputs as probability distributions.
  • Validation of simulation output with respect to the real system.
  • Comparison of multiple system instances, with respect to a performance metric.
  • Optimization via simulation.

Among many others. It all depends on your analysis needs.


Not the answer you're looking for? Browse other questions tagged or ask your own question.