i comparing result between simulation and experiment.But there are huge different between this result.Which kind of error should i look to justify the difference between this two result? especially in simulation
thank you for answering
It is assumed that the experimental results represent the real behavior of the object under test with specific measuring errors as hinted above by Milad. You have to be sure that this error is bound and lies within certain margin.
On the other side the simulation results represent the behavior of the same object based on its theoretical model Then there are some procedure to get the simulation results:
- development of a physical model for the object
- development of a mathematical model for the object leading to system of equations.
- solving the system of equations
-post processing the the results of the solution to get the intended performance parameters.
The two last steps can be verified and the results of the solution get confidence. While the discrepancy between the real performance and the simulation lies in the difference between the real object and its assumed model either physical or mathematical.
So mostly the discrepancy lies between the the real object and its physical and mathematical description specially if the other error sources are minimized.
wish you success
Hi Mr. Mamat,
Always in laboratory errors due to devices, human force, bad material, ... existence. So. if your numerical work is exact and based on theoretical conceptsc is in one direction (finite element method, software results, ...) it seems better modified experimental results and experiment again. Of course, if result of other researcher in this field exist, the best way is help from them (exp. or num.).
Kind Regard,
Milad
It is assumed that the experimental results represent the real behavior of the object under test with specific measuring errors as hinted above by Milad. You have to be sure that this error is bound and lies within certain margin.
On the other side the simulation results represent the behavior of the same object based on its theoretical model Then there are some procedure to get the simulation results:
- development of a physical model for the object
- development of a mathematical model for the object leading to system of equations.
- solving the system of equations
-post processing the the results of the solution to get the intended performance parameters.
The two last steps can be verified and the results of the solution get confidence. While the discrepancy between the real performance and the simulation lies in the difference between the real object and its assumed model either physical or mathematical.
So mostly the discrepancy lies between the the real object and its physical and mathematical description specially if the other error sources are minimized.
wish you success
The simulation may not always produce accurate results, it works on logical manipulation. Whereas analytical results are accurate as these are obtained by proven mathematical manipulation.So, you should compare the analytical result with simulation and experimental result; if you find a HUGE difference in the result in one of them compared to others, you should modify it.
Assume that your experiment is performed without any errors. Now:
1) what you measure is related to the frequency responce of the device. Are you sure that it is suitable to describe all the fundamental frequancies of the flow problem?
2) After you measure, I suppose you will do some statistical analysis. Is the number of samples sufficient?
3) are you considering integral or pointwise quantities?
Now, concerning the numerical simulation, you have to consider what formulation you would use between RANS/URANS, LES, DNS. Each one resolves a totally different variable (statistically averaged, filtered, pointwise) so that to match the experimental data you may need to arrange opportunetely the numerical results.
Finally, in particula RANS/URANS are problematic simulation that depends strongly on the turbulence modelling that affects all the scales of motion. Only using the DNS you get a database from which you can extract all the statistics you want.
simulations basically involved in bringing mathematical or numerical model of your system into real world with the aid of computer soft ware, while experimental analysis involve real life measurement of materials in applicable to apparatus of your system, so how close your simulation result is to that of the experiment determine the validity of your soft ware model simulated result, how ever errors due to measurement, faulty equipment, external force from analyst are also likely to be encounter during the experimental analysis, so simulation result are also significant vice vasa
If the experiment is done correctly, there are three more major error leading to the discrepancy: Modeling error, discretization error and iterrative error!
From my experience, two things are going hand in hand to 'create' the discrepancies:
And during modelling, what I call "side effects of higher orders" tend to be neglected: "this is a marginal omission". Sometimes I was happy when at least well-known "first order side effects" had been included in the model. Errors do not add up - they multiply!
(This doesn't seem to be active knowledge for many 'modellers'.)
And even the best experiment has some residual error(s). And experiment errors multiply as well.
So the model predictions tend to be "better than ideal", the experiment results "worse than real(ly possible)". Thus the differences...
Let first address the semantics of your question.
Once you have a problem, you can address it with SIMULATIONS.
They can be either THEORETICAL, or EXPERIMENTAL. Yes, because also the experimental approach represents a simulation. In fact, generally, you cannot experimentally reproduce exactly what you are studying.
As far as the theoretical simulation is concerned, because of the complexity of the equations you have to deal with, you usually solve the problem in a numerical way. Therefore, you can speak of a NUMERICAL SIMULATION.
The other one is an EXPERIMENTAL SIMULATION:
So, according to the path you choose, you have to build a THEORETICAL MODEL, or an EXPERIMENTAL one which must include ALL the peculiar aspects of your problem. In any case, they should be the SAME for the two models.
I believe that the problem you are facing is that the two models do not coincide so they lead to different results. Please, accurately check them.
Good luck for your work.
.
Normally you need to do a series of "assumptions" (simplifications) to develop your model. Some of them can be erroneous or oversimplify the problem, leading to results that appart - sometimes within a certain range of working conditions - from experimental results.
Otherwise you can also have problems in making your arrangement for experimental determinations (assumed working conditions are not accomplished).
Even if you make wrong assumptions for both the theoretical and the experimental models (which in the latter one means how you build up the experiment), as long as the assumptions are the SAME for the two, the results should coincide.
In fact, the solution should not depend on how you solve (theoretically or experimentally) a fixed, given model.
I still believe that the problem Mr. Mamat is facing lies in the fact that the two models do not coincide, so results are different. Most probably, the experimental model does not cope with the theoretical one.
All the best.
Dear Rais Mamat,
could you post an example of discrepancy in your results? And, please, provide some details about both the experimental devices and the numerical formulation.
That would really help the discussion
There are many explanations but they depend on the particular case.
Examples include uncertainty in the coefficients/parameters assumed, mistaken assumptions made about missing information about the experimental settings, poor resolution (spatial and temporal) of the numerical simulation, and choosing inapproriate mathematical model or submodel in the simulation that is not the best one for the particular problem being simulated.
Dear Mr. Mamat,
as prof. Denaro suggested, you should provide some details on the problem you are facing, as well as about both the experimental devices and the numerical formulation you are using.
What happens more often is comparing two different cases, or normalizing results differently. That happens to me most.
Probably, he is no longer interested in his question.
We are loosing our time.
Regards to all
One can do a simulation of a mathematical problem on the basis of the existing theory. Performance of a simulation of a mathematical problem is carried out to obtain expected results of some suggested phenomena which are already predicted by the experiments. The theoretical concept is based on the experiments. Because in science, the validation/confirmation of the theory is only based on the real experiments (with correctly evaluated experimental errors).
Attached lecture on model verification and validation may be useful in regards to this question.
actually i have similar issue with my research , moment curvature curve differ from experimental tests for 3 models (see attached) - i tried for more than 3 months changing most of parameters in abaqus but i failed
any suggestion
There are many possibilities to explain the mismatch of the experimental results with the numerical results, some due to experimental errors in the measurements and others, due to the numerical model realized, so that you:
- Validation of experimental measurements;
- The boundary conditions for the experimental tests corresponding to the model realized;
- The correct choice of coefficients in the numerical model with what is suitable for experimental tests.
Good luck
Always there are differences between simulation and experimental results. The experimental results are based on real time system, provides much accurate results compared to simulation results. Obtaining the correct results from a experimental set up is a challenge to any researcher. The difference is there due to the errors occurred from external disturbances, instrument etc. But I think the experimental results are more precise/accurate compared to any results obtained from any model utilizing any software.
Examine all assumptions made in your simulation model and see if you can remove those one by one.
Remember that both experimental and numerical approaches are based on a modelling process. As a result they are affected by the hypotheses in the model as well as by further resolution errors.
By absurd, a good matching between numerical and experimental results could be obtained even if they are both largely a poor approximation of the reality.
There would be a time in near future, when the error bars for both experimental and numerical simulations should be mandatory to display. As noted above, both these approaches are models and both are prone to errors. Errors are classified as (a) Epistemic: those due to ignorance of the physical problem and (b) Aleatoric: Those are not deterministic and often described statistically/ stochastically. The modeling errors in experiments and numerical simulations are epistemic and there is always a hope that these errors can be reduced.
Let me explain by examples. In experiments, sometimes utmost care is exercised to replicate theories. As noted in instability theory, Tollmien-Schlichting waves were theoretically postulated and it took about twenty years to verify it in lab under extremely careful designed condition. Outside the labs, these waves have NEVER been demonstrated clearly in realistic situations. Yet, the research community swears by these waves to be the cause for transition to turbulence for many flows. An example of science becoming a faith for a theory! This can be hopefully improved and is being done for this particular example.
Sometimes, in experiment not only the design of it causes error, this is due to limitations of measuring devices. How often, unsteady results are compromised by limited resolutions and bandwidth of used device! You cannot measure turbulence quantity by manometers!
In numerical computations, the subject of error dynamics is systematically developed. First, one should qualify the sources of errors. Next, such errors should be quantified, as it has been done using canonical examples. The popular von Neumann error analysis is demystified in recent times for (i) convection equation, (ii) convection-diffusion equation; (iii) diffusion equation and (iv) convection-diffusion-reaction equations. These may appear not related to solving governing Navier-Stokes equation of fluid mechanics. But, we have reached a stage, where the epistemic part of the error dynamics is being unravelled and one can find critical numerical parameters for solving quasilinear PDEs. Some beginning is essential for aleatoric part of numerical simulations, and some beginning has been made in relating effects of round-off error with long-time integrations via the phenomenon of focusing. This is perhaps the way, future is headed for high performance, high accuracy computing.
- Badges
- Science topic