simulation on wall-mounted squared cylinder
Hi everyone, I am simulating the vortex flow around a squared cylinder mounted a surface using Nektar++ incompressible solver. For this layout (see attached 1), the vortex wakes will become unstable and be shedding as Re >150. However, the current problem is that the computation cannot get the unsteady vortex shedding, it always converges to steady solutions even at Re=500. I even tried giving an incoming disturbance, but the flow fields still cannot converge to vortex shedding. The steady drag coefficients at various Re agree well with the time-averaged results in previous publications, but no vortex shedding occurs. I have simulated the vortex flow past a 2D cylinder, and vortex shedding can be obtained. But for the 3D case, the results are strange. I attached the layout and grid file, and anyone can give some suggestions. Many thanks, Baofeng Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn Tel: 0086 (0)10 82338344
Dear Baofeng, I think the session files looks OK. How long have you run for. I note you have a very symmetric mesh and so in 2D it can take quite a while (O(100)) for the flow to break symmetry and then shed. I am not sure if this will be the case in a Hex mesh such as yours. Also the properties of shedding near a wall can be quite different to a cylinder away from a wall. Although I would have expected Re=500 to be sufficient for it to shed. So you could think of adding an asymmetric blowing onto one of the lateral walls and if you run that for a 1000 time steps and then turn it off it might kick start the shedding. Cheers, Spencer. [cid:F38B9E49-2732-40FC-ADA9-B372CF995775@vlan147.poli.usp.br] On 15 Aug 2017, at 08:28, BF MA <bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn>> wrote: Hi everyone, I am simulating the vortex flow around a squared cylinder mounted a surface using Nektar++ incompressible solver. For this layout (see attached 1), the vortex wakes will become unstable and be shedding as Re >150. However, the current problem is that the computation cannot get the unsteady vortex shedding, it always converges to steady solutions even at Re=500. I even tried giving an incoming disturbance, but the flow fields still cannot converge to vortex shedding. The steady drag coefficients at various Re agree well with the time-averaged results in previous publications, but no vortex shedding occurs. I have simulated the vortex flow past a 2D cylinder, and vortex shedding can be obtained. But for the 3D case, the results are strange. I attached the layout and grid file, and anyone can give some suggestions. Many thanks, Baofeng Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn> Tel: 0086 (0)10 82338344 <layout.jpg><cuboid_4w.xml> Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052
Dear Spencer, Many thanks for your reply. I run the simulation for more than 10000 step with 0.001 time step, even longer. The shedding frequency St in this case is about 0.11, so the running time should be more than one period. In 2D circular cylinders, the mesh is also symmetric, so it is also difficult to break symmetry at an uniform incoming stream. If I run an asymmetric upstream condition for a while, and then turn it off, similarly as you suggested, it can start the shedding. But for the 3D case, even adding an asymmetric condition, it is unable to shed. I am not sure if it will be better if I run it for longer time. In addition, the side force Cz monitored during the simulation will be significantly fluctuating as the time step is enlarged (e.g. setting it as 0.01), probably owing to numerical instability. The time step 0.001 look fine, but it is so small, and the convergence is slower. Cheers, Baofeng -----Original Messages----- From: "Sherwin, Spencer J" <s.sherwin@imperial.ac.uk> Sent Time: Tuesday, August 15, 2017 To: "BF MA" <bf-ma@buaa.edu.cn> Cc: nektar-users <nektar-users@imperial.ac.uk> Subject: Re: simulation on wall-mounted squared cylinder Dear Baofeng, I think the session files looks OK. How long have you run for. I note you have a very symmetric mesh and so in 2D it can take quite a while (O(100)) for the flow to break symmetry and then shed. I am not sure if this will be the case in a Hex mesh such as yours. Also the properties of shedding near a wall can be quite different to a cylinder away from a wall. Although I would have expected Re=500 to be sufficient for it to shed. So you could think of adding an asymmetric blowing onto one of the lateral walls and if you run that for a 1000 time steps and then turn it off it might kick start the shedding. Cheers, Spencer. t On 15 Aug 2017, at 08:28, BF MA <bf-ma@buaa.edu.cn> wrote: Hi everyone, I am simulating the vortex flow around a squared cylinder mounted a surface using Nektar++ incompressible solver. For this layout (see attached 1), the vortex wakes will become unstable and be shedding as Re >150. However, the current problem is that the computation cannot get the unsteady vortex shedding, it always converges to steady solutions even at Re=500. I even tried giving an incoming disturbance, but the flow fields still cannot converge to vortex shedding. The steady drag coefficients at various Re agree well with the time-averaged results in previous publications, but no vortex shedding occurs. I have simulated the vortex flow past a 2D cylinder, and vortex shedding can be obtained. But for the 3D case, the results are strange. I attached the layout and grid file, and anyone can give some suggestions. Many thanks, Baofeng Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn Tel: 0086 (0)10 82338344 <layout.jpg><cuboid_4w.xml> Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk +44 (0) 20 759 45052 -- Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn Tel: 0086 (0)10 82338344
Hi Baofeng, There are some influences of near wall effects that can stop shedding (or at least cause the critical Reynolds number for shedding to be higher). I guess you could try increasing the Reynolds number to see what happens? Regarding time step it is certainly not unusual to require a time step of 0.001. You can turn on the CFL_STEP to get an estimate of this. Cheers, Spencer. On 15 Aug 2017, at 21:17, Baofeng Ma <bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn>> wrote: Dear Spencer, Many thanks for your reply. I run the simulation for more than 10000 step with 0.001 time step, even longer. The shedding frequency St in this case is about 0.11, so the running time should be more than one period. In 2D circular cylinders, the mesh is also symmetric, so it is also difficult to break symmetry at an uniform incoming stream. If I run an asymmetric upstream condition for a while, and then turn it off, similarly as you suggested, it can start the shedding. But for the 3D case, even adding an asymmetric condition, it is unable to shed. I am not sure if it will be better if I run it for longer time. In addition, the side force Cz monitored during the simulation will be significantly fluctuating as the time step is enlarged (e.g. setting it as 0.01), probably owing to numerical instability. The time step 0.001 look fine, but it is so small, and the convergence is slower. Cheers, Baofeng -----Original Messages----- From: "Sherwin, Spencer J" <s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk>> Sent Time: Tuesday, August 15, 2017 To: "BF MA" <bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn>> Cc: nektar-users <nektar-users@imperial.ac.uk<mailto:nektar-users@imperial.ac.uk>> Subject: Re: simulation on wall-mounted squared cylinder Dear Baofeng, I think the session files looks OK. How long have you run for. I note you have a very symmetric mesh and so in 2D it can take quite a while (O(100)) for the flow to break symmetry and then shed. I am not sure if this will be the case in a Hex mesh such as yours. Also the properties of shedding near a wall can be quite different to a cylinder away from a wall. Although I would have expected Re=500 to be sufficient for it to shed. So you could think of adding an asymmetric blowing onto one of the lateral walls and if you run that for a 1000 time steps and then turn it off it might kick start the shedding. Cheers, Spencer. <untitled.png>t On 15 Aug 2017, at 08:28, BF MA <bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn>> wrote: Hi everyone, I am simulating the vortex flow around a squared cylinder mounted a surface using Nektar++ incompressible solver. For this layout (see attached 1), the vortex wakes will become unstable and be shedding as Re >150. However, the current problem is that the computation cannot get the unsteady vortex shedding, it always converges to steady solutions even at Re=500. I even tried giving an incoming disturbance, but the flow fields still cannot converge to vortex shedding. The steady drag coefficients at various Re agree well with the time-averaged results in previous publications, but no vortex shedding occurs. I have simulated the vortex flow past a 2D cylinder, and vortex shedding can be obtained. But for the 3D case, the results are strange. I attached the layout and grid file, and anyone can give some suggestions. Many thanks, Baofeng Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn> Tel: 0086 (0)10 82338344 <layout.jpg><cuboid_4w.xml> Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052 -- Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn> Tel: 0086 (0)10 82338344 Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052
Hi Spencer, Many thanks. I’ll try the case with higher Re. I have another question on higher order grids. I made the grid with Gmsh where higher order grids can be generated. If I understand it correctly, the higher order grid should be used to dealt with curvature boundaries. Therefore, for our present case without curvature boundaries, it is no need to create higher order grids. I just need a first order grid, and set higher order expansions in Nektar++. Is it correct? By contrast, if there are curvature boundaries, a higher order grid is necessary in order to obtain higher order accuracy. Does the order number of grids need the same with the one of polynomial expansions? Cheers, Baofeng Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn Tel: 0086 (0)10 82338344 From: Sherwin, Spencer J [mailto:s.sherwin@imperial.ac.uk] Sent: 2017年8月20日 8:09 To: Baofeng Ma Cc: nektar-users Subject: Re: simulation on wall-mounted squared cylinder Hi Baofeng, There are some influences of near wall effects that can stop shedding (or at least cause the critical Reynolds number for shedding to be higher). I guess you could try increasing the Reynolds number to see what happens? Regarding time step it is certainly not unusual to require a time step of 0.001. You can turn on the CFL_STEP to get an estimate of this. Cheers, Spencer. On 15 Aug 2017, at 21:17, Baofeng Ma <bf-ma@buaa.edu.cn <mailto:bf-ma@buaa.edu.cn> > wrote: Dear Spencer, Many thanks for your reply. I run the simulation for more than 10000 step with 0.001 time step, even longer. The shedding frequency St in this case is about 0.11, so the running time should be more than one period. In 2D circular cylinders, the mesh is also symmetric, so it is also difficult to break symmetry at an uniform incoming stream. If I run an asymmetric upstream condition for a while, and then turn it off, similarly as you suggested, it can start the shedding. But for the 3D case, even adding an asymmetric condition, it is unable to shed. I am not sure if it will be better if I run it for longer time. In addition, the side force Cz monitored during the simulation will be significantly fluctuating as the time step is enlarged (e.g. setting it as 0.01), probably owing to numerical instability. The time step 0.001 look fine, but it is so small, and the convergence is slower. Cheers, Baofeng -----Original Messages----- From: "Sherwin, Spencer J" <s.sherwin@imperial.ac.uk <mailto:s.sherwin@imperial.ac.uk> > Sent Time: Tuesday, August 15, 2017 To: "BF MA" <bf-ma@buaa.edu.cn <mailto:bf-ma@buaa.edu.cn> > Cc: nektar-users <nektar-users@imperial.ac.uk <mailto:nektar-users@imperial. ac.uk> > Subject: Re: simulation on wall-mounted squared cylinder Dear Baofeng, I think the session files looks OK. How long have you run for. I note you have a very symmetric mesh and so in 2D it can take quite a while (O(100)) for the flow to break symmetry and then shed. I am not sure if this will be the case in a Hex mesh such as yours. Also the properties of shedding near a wall can be quite different to a cylinder away from a wall. Although I would have expected Re=500 to be sufficient for it to shed. So you could think of adding an asymmetric blowing onto one of the lateral walls and if you run that for a 1000 time steps and then turn it off it might kick start the shedding. Cheers, Spencer. <untitled.png>t On 15 Aug 2017, at 08:28, BF MA <bf-ma@buaa.edu.cn <mailto:bf-ma@buaa.edu.cn> > wrote: Hi everyone, I am simulating the vortex flow around a squared cylinder mounted a surface using Nektar++ incompressible solver. For this layout (see attached 1), the vortex wakes will become unstable and be shedding as Re >150. However, the current problem is that the computation cannot get the unsteady vortex shedding, it always converges to steady solutions even at Re=500. I even tried giving an incoming disturbance, but the flow fields still cannot converge to vortex shedding. The steady drag coefficients at various Re agree well with the time-averaged results in previous publications, but no vortex shedding occurs. I have simulated the vortex flow past a 2D cylinder, and vortex shedding can be obtained. But for the 3D case, the results are strange. I attached the layout and grid file, and anyone can give some suggestions. Many thanks, Baofeng Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: <mailto:bf-ma@buaa.edu.cn> bf-ma@buaa.edu.cn Tel: 0086 (0)10 82338344 <layout.jpg><cuboid_4w.xml> Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk <mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052 -- Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn <mailto:bf-ma@buaa.edu.cn> Tel: 0086 (0)10 82338344 Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk <mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052
HI Baofeng, I have another question on higher order grids. I made the grid with Gmsh where higher order grids can be generated. If I understand it correctly, the higher order grid should be used to dealt with curvature boundaries. Therefore, for our present case without curvature boundaries, it is no need to create higher order grids. I just need a first order grid, and set higher order expansions in Nektar++. Is it correct? This is indeed correct! By contrast, if there are curvature boundaries, a higher order grid is necessary in order to obtain higher order accuracy. Does the order number of grids need the same with the one of polynomial expansions? For curved surfaces the grid expansion does not need to be the same as the polynomial expansion and generally it is lower (sub parametric). There is a cost of having curved surfaces and it can lead to a reduction in the accuracy if the domain becomes very curved within a single element. Cheers, Spencer. On 20 Aug 2017, at 04:25, BF MA <bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn>> wrote: Hi Spencer, Many thanks. I’ll try the case with higher Re. I have another question on higher order grids. I made the grid with Gmsh where higher order grids can be generated. If I understand it correctly, the higher order grid should be used to dealt with curvature boundaries. Therefore, for our present case without curvature boundaries, it is no need to create higher order grids. I just need a first order grid, and set higher order expansions in Nektar++. Is it correct? By contrast, if there are curvature boundaries, a higher order grid is necessary in order to obtain higher order accuracy. Does the order number of grids need the same with the one of polynomial expansions? Cheers, Baofeng Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn> Tel: 0086 (0)10 82338344 From: Sherwin, Spencer J [mailto:s.sherwin@imperial.ac.uk] Sent: 2017年8月20日 8:09 To: Baofeng Ma Cc: nektar-users Subject: Re: simulation on wall-mounted squared cylinder Hi Baofeng, There are some influences of near wall effects that can stop shedding (or at least cause the critical Reynolds number for shedding to be higher). I guess you could try increasing the Reynolds number to see what happens? Regarding time step it is certainly not unusual to require a time step of 0.001. You can turn on the CFL_STEP to get an estimate of this. Cheers, Spencer. On 15 Aug 2017, at 21:17, Baofeng Ma <bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn>> wrote: Dear Spencer, Many thanks for your reply. I run the simulation for more than 10000 step with 0.001 time step, even longer. The shedding frequency St in this case is about 0.11, so the running time should be more than one period. In 2D circular cylinders, the mesh is also symmetric, so it is also difficult to break symmetry at an uniform incoming stream. If I run an asymmetric upstream condition for a while, and then turn it off, similarly as you suggested, it can start the shedding. But for the 3D case, even adding an asymmetric condition, it is unable to shed. I am not sure if it will be better if I run it for longer time. In addition, the side force Cz monitored during the simulation will be significantly fluctuating as the time step is enlarged (e.g. setting it as 0.01), probably owing to numerical instability. The time step 0.001 look fine, but it is so small, and the convergence is slower. Cheers, Baofeng -----Original Messages----- From: "Sherwin, Spencer J" <s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk>> Sent Time: Tuesday, August 15, 2017 To: "BF MA" <bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn>> Cc: nektar-users <nektar-users@imperial.ac.uk<mailto:nektar-users@imperial.ac.uk>> Subject: Re: simulation on wall-mounted squared cylinder Dear Baofeng, I think the session files looks OK. How long have you run for. I note you have a very symmetric mesh and so in 2D it can take quite a while (O(100)) for the flow to break symmetry and then shed. I am not sure if this will be the case in a Hex mesh such as yours. Also the properties of shedding near a wall can be quite different to a cylinder away from a wall. Although I would have expected Re=500 to be sufficient for it to shed. So you could think of adding an asymmetric blowing onto one of the lateral walls and if you run that for a 1000 time steps and then turn it off it might kick start the shedding. Cheers, Spencer. <untitled.png>t On 15 Aug 2017, at 08:28, BF MA <bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn>> wrote: Hi everyone, I am simulating the vortex flow around a squared cylinder mounted a surface using Nektar++ incompressible solver. For this layout (see attached 1), the vortex wakes will become unstable and be shedding as Re >150. However, the current problem is that the computation cannot get the unsteady vortex shedding, it always converges to steady solutions even at Re=500. I even tried giving an incoming disturbance, but the flow fields still cannot converge to vortex shedding. The steady drag coefficients at various Re agree well with the time-averaged results in previous publications, but no vortex shedding occurs. I have simulated the vortex flow past a 2D cylinder, and vortex shedding can be obtained. But for the 3D case, the results are strange. I attached the layout and grid file, and anyone can give some suggestions. Many thanks, Baofeng Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn> Tel: 0086 (0)10 82338344 <layout.jpg><cuboid_4w.xml> Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052 -- Baofeng Ma, PH.D Associate Professor Institute of Fluid Mechanics School of Aeronautical Science and Engineering Beihang University Email: bf-ma@buaa.edu.cn<mailto:bf-ma@buaa.edu.cn> Tel: 0086 (0)10 82338344 Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052 Spencer Sherwin FREng, FRAeS McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052
participants (3)
-
Baofeng Ma
-
BF MA
-
Sherwin, Spencer J