一段有关Fluent中LES的评价
Commercial CFD software is a relatively new phenomenon, one that really only got underway in the late 1980's. Even now, the typical CFD company is relatively small and closely tied to academia. For a nice list of commercial CFD companies, click here.
------------------------------------------------------------------------------ --
The following, originally posted on CFD Online by "Andy," illustrates the stengths and weaknesses typical of commercial software. Although this concerns FLUENT specifically, similar things could be said of any commercial CFD package. "FLUENT is a general purpose code and can address a wide range of problems in fluid mechanics. Although this generality can be what is needed it is also the key to FLUENT's main failing: it cannot predict particular classes of flows as well as CFD codes written solely for that class of flow. That is, if you are only interested in hypersonic flows then a CFD code written only to solve hypersonic flows can be substantially more accurate and more efficient. However, the costs of working effectively with specialized codes are different to those of working with general purpose codes. It can be higher or lower - the key is usually access to educated people who know what they are doing. Perhaps using a bigger computer, waiting longer for the answers and accepting a higher risk that the answers are misleading is perfectly acceptable. It depends on what you are using the code to achieve. "A personal example: I recently considered proposing FLUENT for a set of LES simulation (it has an LES button and could be driven by relatively inexperienced people). After a phone call with FLUENT it became clear that: "The code is very inefficient requiring around 30 times more computing time per time step than a purpose written code. This figure is very approximate but was based on a real prediction running within FLUENT at the time - I could not believe the computing resource it required. "The low accuracy of the convection terms is a poor choice for LES. i.e. requiring much too fine a grid to match a scheme which uses a more accurate differencing scheme. "It was not possible to generate curvilinear grids of a high enough quality. Related to accuracy of convection terms (or perhaps not given the low accuracy differencing!). "The computer memory requirement was not investigated but it looked to be around 10 times more. "I was not convinced that FLUENT were on top of the problems of near wall treatment for this type of flow since the parameters to control it were not in evidence. "To be fair, I think this should be recognised as something of an extreme example caused by FLUENT adopting low accuracy differencing, unstructured grids and general purpose implicit solvers for a class of flows which are best served by high accuracy differencing, structured grids and very simple efficient solvers. "To conclude, I believe codes like FLUENT can help address a significant range of engineering problems. However, they are best used by people who understand their strengths and weaknesses and they are best not used in isolation." |