Method of Manufactured Solutions - Turbomachinery
 

Hi

I am trying to get to grips with the Method of Manufactured Solutions; I want to do a Code Verification that is appropriate for a turbomachinery simulation (centrifugal pump). The solver I am using is ANSYS CFX.

I would appreciate it if somebody could nudge me in the right direction with an example of a completed (turbomachinery relevant) MMS study or something in the literature (perhaps referring to a good paper/article that deals with the specific application) that can get me on the right track?

Thank you very much in advance.

Dear TurboFlo,

what is the point of your study. A MMS only veryfies the implementation of the CFD code. ANSYS has done this trillion times.

Did you implement additional stuff and want to verify that?

Regards

i really doubt this part for the reasons that most of the time it requires special boundary conditions that by default are not available in CFD codes.



Also never came across any such paper so far. Do you know any such study i would love to know.

I am really sure that this is the case for most stuff implemented in commercial tools.

Boundary conditions are part of the implementation. That is the reason why i asked

Nobody knows without more information.

Regards

I was developer on starccm and when I worked there I did do benchmarking and to do so I was forced to code the BC because solver directly does not allow them.


I developed Wildkatze solver http://www.digital-sol.co.jp/wildkatze/index_en.html and yes during development I used manufactured solution to make sure that all the terms are properly accounted for. I can still not offer that to user because once you offer it as a solver you have to provide lot more things to treat BC that it becomes very difficult to allow fixed velocity and pressure at BCs.



This is indeed true.

Hi all

No, I did not change anything neither am I implementing something alien to the code that requires verifying. I am attempting to frame an order of accuracy for the CFX code.

A direct quote from ASME V&V 20-2009: "Prior to estimating uncertainty of a numerical solution due to numerical error (unum), it is necessary to verify the code itself [i.e., to determine that the code is free of mistakes (code verification)]"... "The objective of verification is to establish numerical accuracy, independent of the physical (modeling) accuracy that is the subject of validation."...and a bit later in the text: "code verification is distinct from solution verification and must precede it, even though both procedures utilize grid convergence studies. In general, code verification assesses code correctness and specifically involves error evaluation for a known solution. By contrast, solution verification involves error estimation, since the exact solution to the specific problem is unknown."

Essentially, a formal order of accuracy (p) is to be established that will serve in later efforts to compare the observed- and formal/stated order of accuracy during verification.

I have no doubt that the ANSYS team have performed exhaustive studies, but the assumption that it has been done satisfactorily would have to be proven in written works. It would be great to see a verification baseline case from CFX, but I have not been able to track anything down (maybe I do not know where to look).

Various texts do refer to the MMS as a good method to use, and I tend to agree. I would just like to see a worked example of a CFX case?

When I started my LES journey with Fluent I had several troubles with the bi-periodic channel flow. Only by verification with the MMS on the Taylor solution I was able to identify a bug in the green gauss gradient for bi-periodic settings that is not mentioned in the manual. Using the MMS properly, in turn, involved few subtleties, like properly modifying the reference pressure location and initializing the first two time steps (instead of the first one only).

So, I guess that it matters or not according to your own competence. If you are at least concerned I think you should proceed to do your own verification. Unfortunately I don't know of any specific such solution. But the nice thing is that the MMS can be used with whatever solution, just come up with something meaningful for your field, like a vortex flow or sort of

@sbaffini; Thanks for the reply.

Yes, I definitely will attempt the verification; I have found some great reference works from the proceedings and articles published by the AIAA (https://arc.aiaa.org/), in particular from authors like Roache. One paper https://arc.aiaa.org/doi/10.2514/6.2005-682 also directly performs some code verifications using commercial CFD codes (TASCflow, Fluent and CFX) for various 2D cases - CFX does not have 2D capabilities and one should construct a one element deep flow field to complete a (almost) 2D case. Another paper https://arc.aiaa.org/doi/10.2514/6.2010-127 details a 3D case with unstructured mesh application. Obviously the reference works by Roache (http://www.hermosa-pub.com/hermosa/h...oks.html#VVCSE) are indispensable to the understanding of this method.

Some databases (http://manufactured-solutions.github.io/MASA/index.html) also makes it a bit easier to implement the MMS if you know what you want to achieve.

I will continue to look for other baseline cases (like the backward facing step and some additional 3D applications), and perform the required work.