[butterflymuzzy]

Why aren't regulations such as innovations (overset mesh and especially hexahedral and polyhedral mesh Methods with inflation like tetrahedral mesh) made to ansys fluente in ansys cfx?

I think that I refrain from making the development investment of ansys in fluent, in cfx. ansys treats cfx as a stepchild.

I hope we will see these innovations in cfx in the future. because cfx is really nicer than fluent in terms of interface.

[Opaque]

Ansys CFX is fundamentally a polyhedral mesh code by definition. Not sure exactly what you mean? are you saying to work with polyhedral elements? There is no exact analytical theory for interpolation functions for arbitrary polyhedral elements.

On the tetrahedral mesh with inflation, Ansys CFX has used those meshes since its initial release back in early 2000's. What kind of feature do you think you are missing?

[Gert-Jan]

CFX is a node based solver. This makes a CFX-solution with a tet-prism mesh (more or less) comparable to a polymesh solution in fluent. So, you could say that what they implement in Fluent has been in CFX for years.
More over, recently they implemented the coupled solver from CFX in fluent as well, and made it the default method.

Bottom line: all good things from CFX are put into fluent. The better GUI and reliability from CFX have not been tranferred to Fluent yet. So basically it remains a crappy C-Python-based non-intuitvely environment which crashes a lot. (I use both packages a lot)

ANSYS does this because fluent has a larger market share. They don't want to push all their existing customers to CFX, which would have been a better idea, in my opinion.

So, all the TECS we as commercial ANSYS users pay go into the development of Fluent, which is a dead end in my view. It would be better it they build something new from scratch, like CD-adapco did 15 years ago with Star-CCM+

I wonder if we should accept the current situation any longer.

[ghorrocks]

The future of CFD with ANSYS is Discovery and Live. It will be many years before they have all the physics from CFX and Fluent, but they will catch up some time.

[Gert-Jan]

Even worse to pay TECS for that. We are digging our own grave.

[ghorrocks]

It is a sad fact of life that the growth of CFD is not with people like you and me. It is in the myriad of general engineers, students and non-CFD specialists. They would not know a coupled solver from an advection scheme, but they still want to model the latest gizmo they have designed.

[Gert-Jan]

Mass is cash. In other words, it is all about maximizing shareholder value.

[ghorrocks]

ANSYS is a listed company. So you are correct, they have to chase the best return for shareholders.

[Indra]

Originally Posted by ghorrocks

The future of CFD with ANSYS is Discovery and Live. It will be many years before they have all the physics from CFX and Fluent, but they will catch up some time.

Discovery is mostly meant for designers, not simulation specialists. So I think Fluent will remain the specialist product.

[Indra]

Originally Posted by butterflymuzzy

Why aren't regulations such as innovations (overset mesh and especially hexahedral and polyhedral mesh Methods with inflation like tetrahedral mesh) made to ansys fluente in ansys cfx?

I think that I refrain from making the development investment of ansys in fluent, in cfx. ansys treats cfx as a stepchild.

I hope we will see these innovations in cfx in the future. because cfx is really nicer than fluent in terms of interface.

Fluent is much more attractive with its GUI(especially dark mode ) and multi-physics capabilities and superior meshing. Not sure if you have seen the latest versions but they are so easy to use, even making turbomachinery simulation easier using turbo workflow, aero workspace making external aero easy and much more.
All in all, for new users, Fluent offers a better and more friendly GUI. If you have been using CFX, you might find it different and hence it would appear "nicer".

[ghorrocks]

Your points are all true, Indra.

However I do not see a long term future for CFX or Fluent as neither of them work well on GPU systems, and GPU systems are becoming dominant in high-performance computing platform. Live is ANSYS's only CFD software which really uses the power of GPU, but it was never intended to be software for CFD specialists. So ANSYS has a problem to develop a high-end CFD package which runs well on GPU systems. Some of its competitors are making progress here and ANSYS has the potential to be left behind if they do not get a move along.

But this will take many years to play out. CFX and Fluent have years left in them before this issue really starts to bite.

[Indra]

Originally Posted by ghorrocks

Your points are all true, Indra.

However I do not see a long term future for CFX or Fluent as neither of them work well on GPU systems, and GPU systems are becoming dominant in high-performance computing platform. Live is ANSYS's only CFD software which really uses the power of GPU, but it was never intended to be software for CFD specialists. So ANSYS has a problem to develop a high-end CFD package which runs well on GPU systems. Some of its competitors are making progress here and ANSYS has the potential to be left behind if they do not get a move along.

But this will take many years to play out. CFX and Fluent have years left in them before this issue really starts to bite.

Fluent now offers you GPU capabilities with some promising speed improvements( "new Multi-GPU Fluent solver (beta) showing four high-end GPUs provide the same performance as more than 1,000 CPUs"). Never used it personally though. Check out the latest version.
Given the size of ANSYS, I believe they have the money and resources to make all the latest updates and not be left behind.

[ghorrocks]

I had not seen that in the release notes. Thanks for pointing that out.

In previous releases the GPU capability was very restricted, it only worked on a few physics models (particle tracking from memory). It will be a game changer when the GPU works with the solver on all physics models (or at least most of them).

The reason I am a bit sceptical of the ability of CFX and Fluent to adapt to GPU platforms is that the basic numerics of the CFX and Fluent solver requires solution of large matrixes of linear equations, that is the equations are implicit. GPUs tend to not perform well in these setups. GPUs work best in explicit equation models which are algebraic equations and highly suited for massively parallel systems. But CFX and Fluent are implicit and to make them explicit is a re-write from the ground up.

But if ANSYS proves my scepticism wrong then I will be very pleased.

[butterflymuzzy]

I am a Naval architecture engineer and i have been using implicit ansys cfx for 14 years in ship resistance and propulsion analysis. but recently, I have found that star ccm applications in cfd analyzes in the maritime sector have started to be used more frequently in the literature. The star ccm cfd interface is used in ship resistance with a high froude number, ship maneuver calculations with the overset mesh method, and propeller analysis. I found that star ccm has the ability to analyze with polyhedral and hexahedral mesh with inflation in a smaller number of meshes. Habits are what keep me going with CFX. I hope the applications in star ccm will be brought to CFX as well. Let fluent develop as much as it wants. I always like CFX more as an interface.

[Indra]

Originally Posted by butterflymuzzy

I am a Naval architecture engineer and i have been using implicit ansys cfx for 14 years in ship resistance and propulsion analysis. but recently, I have found that star ccm applications in cfd analyzes in the maritime sector have started to be used more frequently in the literature. The star ccm cfd interface is used in ship resistance with a high froude number, ship maneuver calculations with the overset mesh method, and propeller analysis. I found that star ccm has the ability to analyze with polyhedral and hexahedral mesh with inflation in a smaller number of meshes. Habits are what keep me going with CFX. I hope the applications in star ccm will be brought to CFX as well. Let fluent develop as much as it wants. I always like CFX more as an interface.

I understand that at the end it's all down to familiarity and habit. Fluent does the mix of polyhedra and hex mesh called moasaic meshing which saves quite a bit of computational time along with only poly mesh as well.

CFX mesher turbogrid now does a tet-hexcore meshing as well.

[Saeef]

Originally Posted by Opaque

Ansys CFX is fundamentally a polyhedral mesh code by definition. Not sure exactly what you mean? are you saying to work with polyhedral elements? There is no exact analytical theory for interpolation functions for arbitrary polyhedral elements.

On the tetrahedral mesh with inflation, Ansys CFX has used those meshes since its initial release back in early 2000's. What kind of feature do you think you are missing?

Hi, Opaque

CFX is my favorite CFD software. The biggest advantage is that it is easy to operate! CEL and adding variable or source term are much more convenient than Fluent or Numeca. But it can be improved in the following aspects, at least.
1. GPU parallel computing is not supported.
2. Pressure-based solver has insufficient robustness in terms of supercritical CO2 compressor/turbine/heat exchanger/ejector simulations. Most people will use NUMECA / fine turbo for supercritical CO2 compressor/turbine flows. It's worth stressing that CFX has a large proportion of rotating machinery users！
3. Subroutines can only support the old FORTRAN language. NUMECA Fine/Open supports a much more convenient Python script.
4. Lack of adjoint solver. Fluent and STAR-CCM support it. The efficiency of adjoint optimization method is much higher than that of GA.
5. Users can modify much fewer parameters than Fluent. For example, Fluent can modify most of the coefficients in Spalart-Allmaras turbulence model, while CFX can only modify two parameters.
6. There are defects in the inlet boundary conditions of large eddy simulation. CFX cannot automatically set the boundary condition of large eddy simulation that meets the energy spectrum. In this regard, CFX is worth learning from Fluent.

I hope developers can improve them.

[ghorrocks]

While your points are all good and valid (except for the fortran/python point IMHO ), there is little chance ANSYS will do anything about them. ANSYS is hardly spending any development effort on CFX - the last few releases had very little new stuff in CFX, while Fluent and Discovery got lots.

[butterflymuzzy]

why is ansys treated like a stepson to cfx ? I I have been waiting the “adaptive grid refinement” and faster converge of “rigid body-free surface” technologies in the ship resistance and self propulsion problems since 2016. The free surface can easily converge but on the other side I m waiting too much tine the sinkage or trim of the vessel. And I can’t do the free surface zig zag maneuvering or 3 dof with self propulsion propeller. In the naval architecture cfd literature papers star ccm’s popularity rising. Cfx please get up *♂️

[ghorrocks]

I think my previous post applies to your comment as well. you raise good points, but there is no chance CFX will address them. ANSYS is focussing its development on other software.

[SaschaGH]

Originally Posted by ghorrocks

I think my previous post applies to your comment as well. you raise good points, but there is no chance CFX will address them. ANSYS is focussing its development on other software.

Yeah absolutely. I just talked to one of the software engineers who told me that they are only going to push fluent in the future and that there is finally some real action going on at ansys to do so.

I feel like they keep saying this for the last 10 years, but let's see.


I was just assigned to plan a new in-house cluster, something around 512 cores or similar. As I have been working with CFX for the last 10 years and now seeing that cfx is likely to be discontinued & fluent seems to work well on GPUs, what would you experts go for? Mass CPU or GPU equivalent?