Learning Forum

[Glenn]

I have been trying to run my first calculation model but am having some difficulties.  I am extremely new to CFD/Fluent but have tried to read up as much information on the interweb as possible.  Much of it is confusing however but I am trying my best....

My model is of an internal combustion engine inlet tract terminating in a new valve shape.  I believe my mesh is reasonable but I stand to be corrected. 

Air enters the airhorn (massflow_inlet) at 45.4m/s which implies around 0.2048kg/s of air (68.4mm diameter and 1.225 kg/m3 density).  This value has been entered in the boundary conditions.  I am wanting to analyse the flow through the entire model (under a fixed inlet flow of 45.4m/s) and to see how the flow (speed, pressure, uniformity, turbulance) looks at the outlet.

I have been researching whether this should be a pressure or density based calculation and settled on density but still confused.  I assume it should be a 'time steady' calc but again not 100%.

I am getting an immediate issue when running the calculation that Fluent reports a "floating point exception" and stops the model.  I can also see in the console window that it shows reverse flow at either or both the inlet and outlet which confuses me as I am not expecting backflow anywhere.  I am nor expecting this forum to teach me how to use this tool better but any pointers (in basic/beginner terms lol) would be appreciated.

I have run the model previously with other settings (of which I did not take sufficient notes and have therefore forgotten) that either did complete the iterations but did not converge or against eventually gave a "floating point error".

Guidance appreciated.....

PS - is that the correct file to upload??

[peteroznewman]

Glenn, you don't say what version of ANSYS you are using. I opened your file with 2019 R2.

I am also learning Fluent, so I am willing to be corrected by the experts.  I believe the correct selection is the Pressure-based solver, not the Density-based solver which is better for air flows that have extreme pressure gradients that cause the density of air to change, such as supersonic flow.

A Steady State solution is always a good place to start a study.

CFD needs good quality elements. Skewness is an important quality metric. You have 10 Pyr5 elements that have a skewness of 0.991 and that value can ruin the whole simulation result.

You ideally want elements with skewness < 0.75 so slice and simplify the geometry until you get there.

It can happen that during the iterations to a converged solution, some cells on the pressure outlet might have reversed flow, but as the solution improves and the residual errors are driven down, the reversed flow will probably resolve itself.

[Glenn]

Thanks a heap Peter.  I did eventually realise that the Ortho quality and skewness were extremely poor.  I therefore started to look at the CAD file and find any areas that had artifacts or poorly fitting parts.  I have identified a couple of issues and am trying to rectify before reloading into the ANSYS software.  I assume the picture above with the black areas are some of the ones you found to be very poor quality/skewness.  They are next on my CAD list to examine.

I will also try the Pressure based solution.

I am using 2019 R2 Student.  I am currently close to the 512K element limit so cannot make my mesh any finer but hopefully fixing the above will give me a good quality mesh.

[peteroznewman]

It's not about making the mesh finer, it's about simplifying the geometry so larger, less skewed elements can be made.

When you request mesh metrics, you click Control to change the limits on the X and Y axes, then you can click and ctrl-click the bars on the graph to cause the elements to highlight.  I highlighted the 53 worst Pyr5 elements in the mesh.

[Glenn]

I have spent considerable time cleaning up my CAD model.  There are still areas that I am not happy with but my CAD skills are only slightly better than my Fluent skills so I have had to run with it.

 

I believe these are a much better set of stats (?);

 

I still have alot of cells with poor orthogonal quality and skewness;

 

The calculation ran this time with no floating point errors.  I still have a long way to go in order to research and understand all the various model inputs such that I can have any amount of faith in the calculation output......(currently put 0.2kg/s air in at the inlet but getting consistent negative flow at the pressure-outlet!!!!!  Not sure how energy is being conserved in that calc!)

[DrAmine]

Your metrics are poor. Check if only small number of cells are affected. Sometimes even a small number of poor cells can affect the whole run.

Check if poor mesh numerics can help there. More details in User's Guide!

Translation to polyhedral cells might enhance the performance. An auto node move in Fluent Meshing might enhance the quality.

[Glenn]

I have spent the day running re-meshing under a huge number of different setup options.  Some have plain not worked but I am starting to get a feel for what will and will not work.  However, looking at the Mesh Metrics bar graph I can never get rid of the handful of low quality elements <0.13.  They number something in the order of 1K out of a total of 670K (need to get this under 512K).   This is the metrics;

 

 

I use 2019 R2 student and cannot find an auto node move function.  I found a 'Node Move' icon which created a Mesh Edit item in the tree but cannot work out what to do with it.  Suggestions?

 

I have not loaded the updated mesh into Fluent yet in order to look at the Ortho and Skew Qualities but assume they will be rubbish.

 

How important is it to remove all the low quality rubbish cells that may ruin a calculation vs leaving a few in?

[DrAmine]

Is casde dependent. If you are lucky it might work but I saw cases where some few cells are affecting the whole field.

Now just start in Fluent and see if it is running. Also post then settings etcd..