How much discrepancies are normal?


Hello I’m still kind of new to Simscale so I’m apologizing for this stupid question but I was wondering to what degree discrepancies between simulation and real experiment are normal?

I working on a heat conductivity project and simulated it in the Thermal Simulation at work. But when I conducted the experiment with this LZT Meter I got different results. I don’t think they differ too badly but still. How much is considered normal in your experience?

Thank you in advance.


Hi @acalavare!

First of all how big is the difference in percent? The admissible discrepancy is highly dependent on the application and simply saying 5% is “still okay” would not make much sense - context matters.

Can you share the project and give us a rough idea of what you are investigating (thermal conductivity? heat distribution at specific points?). Next thing is not to trivialize a complex problem too much. Remember that a simulation only tries to model reality but we can (almost) never do it for 100% which is sometimes what we want.

I would suggest that we discuss this in more details. I also invite our PowerUsers and experts @cjquijano, @BenLewis & @ggiraldo.




There are no stupid questions!

I agree with Jousef, the amount of tolerable discrepancies between a numerical model and a laboratory experiment could depend on many factors.

On the other hand, and please correct me if I’m wrong, but the equipment you indicate is used to measure some material properties and their temperature dependence. How did you simulate this using SimScale? What kind of results are you comparing?

Take into account that the accuracy of the simulation results highly depends in the input parameters, such as material properties, but those are inputs, not outputs. Even more, in a typical workflow one would take the output of the equipment experiment as inputs to the numerical simulation.


There are sources of error is simulation and in physical measurements. How closely the two match depends upon how well the simulation models real life behavior and on how accurately the instrumentation measures physical quantities. If the two sources of information differ, by more than an amount you are comfortable with, then you need to spend more resources tracking down and minimizing sources of error.

In my experience, you should be able to get the two sources of information to agree within acceptable limits. But it is not unusual to get a large discrepancy on a first pass.