Verifying results

[phillip mobley]

Hello everyone,

Working with my very simple simulation of a charged parallel plate capacitor has shed much light on how to setup a simulation for Deal.II

However, there has been one thing on my mind. For more complex simulations, how would you verify that your simulation is correct?

For me, I have a simulation of a charged parallel plate capacitor from a different program (COMSOL) to verify this. But what if I am simulating something that no one has simulated before. How can I ensure that my results are correct or accurate?

[Bruno Turcksin]

Phillip,

On Thursday, August 13, 2015 at 5:03:07 PM UTC-5, phillip mobley wrote:

However, there has been one thing on my mind. For more complex simulations, how would you verify that your simulation is correct?

For me, I have a simulation of a charged parallel plate capacitor from a different program (COMSOL) to verify this. But what if I am simulating something that no one has simulated before. How can I ensure that my results are correct or accurate?

a very common method is to use a manufactured solution (you can google it). Basically, you choose a solution for your problem and then plug that solution in your equation. This will give you a source term. Then, you can use that source term in your code and check that the solution that your code gives you is the one that you expected. This method can (should) be coupled with comparison with other codes and experiments. All of this is necessary to get some confidence in your code but that doesn't mean that your code is bug free, maybe if you try a new geometry you will find a new bug. There is basically no way to be sure at 100% that your code is bug free (we still find bugs in deal.II) but you should be able to get pretty confident in your results ;-)

Best,

Bruno

[phillip mobley]

Hello Bruno,

Thank you for your response. This will help me much later when I get to more complex simulations