Discovery Live : Strange results for a three-body car simulation

  • Last Post 14 January 2020
pierro42 posted this 28 May 2019


I've got strange results with a three body car simulation with ANSYS DISCOVERY LIVE.

I'm seeing a HUGE flow separation next to the rear windshield, which should not happen since there's a really low angle, ( Honda civic sedan 1996 )

I didn't have seen this with my old openfoam setup, neither with a R17 Fluent ( Not my license, i can't do it again )

I don't have pictures now, but i'll post them asap.


Do you know if there's a way to increase the cell number ? I've already maxed out the precision slider, and i can't visualise the cells size.

Or maybe it has to be a strange boundary condition

By the way, the car is 1:1 scale, wind speed is 40 m/s


I've attached a picture of what i'm seeing.  The flow separation should have been really small compared to what i'm seeing.


Any advice ?



Attached Files

Order By: Standard | Newest | Votes
pierro42 posted this 28 May 2019

Here's a picture of what i'm seeing with a 1:1 scale, 10m/s 20°air 1013hPa.
This is clearly wrong !


nmpatre posted this 29 May 2019

Hello pierro42 

Discovery Live results are largely affected by the GPU memory of the graphics card. What is the memory of your graphics card?

Can you try reducing the extents of the outer domain and check the results again?

Lastly, Discovery Live is really not intended to measure the accuracy of results by comparing them with other tools. Rather, it is better suited to understand the trend of results by changing different parameters like geometry, boundary conditions, etc which will help you to get an initial guess of results to meet your simulation objective. You can later use Discovery AIM to get accurate results.

pierro42 posted this 29 May 2019

Hi !


This test was done with a RTX 2070 (8Go GDDR6 RAM) with latest CUDA drivers.

I've tried with a bigger volume to let the vortices fully develop, and with a smaller one to virtually increase the cell count per cm. The result was the same.


The goal here was to show the impact of a rear spoiler. But the simulation doesnt give any airflow in the region where the wing should be.

So in this case i can't use Discovery Live to give a tendency.


I think i should try Discovery AIM, but i wasn't sure what were the difference between both for aerodynamics tests.



pierro42 posted this 29 May 2019

I've done a new simulation using AIM.


LICENSE MANAGER: Not enough ANSYS Neutral Parallel licenses 8/2.

Exception from ServeOneRequest: Low-level communication error -19: Connection is dirty


My computer is a fresh W10 install ( less than a month ) and was free of any ANSYS product.
Firewall is disabled, and AIM has Admin privileges.

This is upsetting...


nmpatre posted this 30 May 2019

Someone with knowledge about licensing should be able to comment on the license issue that you are facing.

pierro42 posted this 03 June 2019


for Discovery AIM, it worked with 1 core only vs 8 available. But i had troubles importing the Discovery Live case. I'll figure it out later.

Could you check the formation of a buble on the back of a Ahmed Body for example ?


It should look like this --> Ahmed_Body


On my Ansys DISCOVERY LIV Esimulation, there is a separation as soon as there's the angle of the back plane.


Have you done a model validation with Ahmed Body ?



pierro42 posted this 05 June 2019



See the article with now flow separations on ANSYS Fluent :



pierro42 posted this 21 June 2019

Anyone ?

abenhadj posted this 21 June 2019

What is now your question? We do not download attachments or read external links.

Best regards, Amine

pierro42 posted this 05 January 2020


Sorry it's been a while.

Here's what i'm getting with Ansys Live on the maximum fidelity :


Ahmed's body


Here's what i should get :

Ahmed's body Ansys

pierro42 posted this 07 January 2020

So the question is : Can you help me to find what i've done wrong ?

Thanks a lot !

senowm posted this 08 January 2020

check the fluid density


pierro42 posted this 08 January 2020

Hi !

Re should be around 4.5e6 for both simulations, to match real data.


u=40 m/s

pierro42 posted this 11 January 2020

Anyone ?

pierro42 posted this 14 January 2020

Anyone ?