
[Sponsors] 
transient vs steady state solution for multiphase flows 

LinkBack  Thread Tools  Search this Thread  Display Modes 
February 12, 2021, 04:53 
transient vs steady state solution for multiphase flows

#1 
New Member
Mehran Janghorbani
Join Date: Feb 2021
Posts: 10
Rep Power: 5 
Dear All
Please forgive the newbie question, but I am trying to simulate the accumulation of sand grains carried by water (both laminar and turbulent flows) in a horizontal pipe using an EulerEuler approach. The boundary conditions of the problem do not change with time (constant flow of water and sand into the domain) but am I right in assuming that since the concentration of sand in the domain changes, the transient solver should be used? I ask this because many authors have used a steady state solver (I am using ANSYS) and their results vary considerably from mine. I do not understand how an S.S solver can be used in this case and how such a solver would solve the transient terms. With best regards Mehran 

February 12, 2021, 09:57 

#2 
Senior Member
Kira
Join Date: Nov 2020
Location: Canada
Posts: 435
Rep Power: 8 
Hello Mehran,
It is indeed very strange that the literature you have read was using a steady state solver for this problem. I believe you are correct in using a transient solver, as the backwashing process of the quartz/sand filter layer is both a dynamic and a stable process. As well, sand deposition and transport is a transient phenomenon in nature, as the sand particles tend to change concentrations with time and along the flow path. Could you post a link to the papers using a steady state solver? What was their reason for adopting a steady state solver? 

February 14, 2021, 22:24 

#3 
New Member
Mehran Janghorbani
Join Date: Feb 2021
Posts: 10
Rep Power: 5 
The authors have unfortunately not given any reasons for their choice, however I have listed a few of them for your reference:
1) Bilgesu, H.I., Ali, M.W., Aminian, K. and Ameri, S., 2002, January. Computational Fluid Dynamics (CFD) as a tool to study cutting transport in wellbores. In SPE Eastern Regional Meeting. Society of Petroleum Engineers. 2) Amanna, B. and Movaghar, M.R.K., 2016. Cuttings transport behavior in directional drilling using computational fluid dynamics (CFD). Journal of Natural Gas Science and Engineering, 34, pp.670679. 3) Moraveji, Mostafa Keshavarz, Mohammad Sabah, Ahmad Shahryari, and Ahmadreza Ghaffarkhah. "Investigation of drill pipe rotation effect on cutting transport with aerated mud using CFD approach." Advanced Powder Technology 28, no. 4 (2017): 11411153. 4) Mme, Uduak, and Pål Skalle. "CFD calculations of cuttings transport through drilling annuli at various angles." International Journal of Petroleum Science and Technology 6, no. 2 (2012): 129141. 5) Heydari, Omid, Eghbal Sahraei, and Pål Skalle. "Investigating the impact of drillpipe's rotation and eccentricity on cuttings transport phenomenon in various horizontal annuluses using computational fluid dynamics (CFD)." Journal of Petroleum Science and Engineering 156 (2017): 801813. These are only a few, I have studied over 70 such publications and the authors have either not mentioned the solution method (in which case I would also assume that they used SS) or mentioned that they used SS. Regs Mehran 

February 14, 2021, 23:08 

#4  
Senior Member
Kira
Join Date: Nov 2020
Location: Canada
Posts: 435
Rep Power: 8 
Quote:
I took a look through all of the papers. I found an additional one relating to the topic of drill pipe rotation/mud flowing through a pipe. A quote from it is: ... CFD can be used to investigate a 3D fluid problem and obtain a 3D solution, providing pressure and velocity fields. Balance equations for mass, momentum and (depending on the assumptions/required solution) energy are solved numerically on a threedimensional grid of the domain of interest. If a second phase, such as dispersed particles/cuttings, is part of the problem this can be either treated as a second continuum (EulerianEulerian concept) or as individual particles (EulerianLagrangian method). In the latter case more computational effort arises as the particles trajectories are computed for each particle. Mainly depending on the available computational power and the desired accuracy of the result the following concepts can be distinguished: Direct Numerical Simulations (DNS) resolve turbulence on on all length and time scales down to the Kolmogorov length and time scale. Large Eddy Simulations (LES) resolve turbulence on length and time scales larger than grid size and time step (resolving Large Eddies) and modeling turbulence on subgrid scales. ReynoldsAveraged NavierStokes (RANS) approaches entirely model the effect of turbulence on all length and time scales. The respective turbulence model is chosen with regards to the physics of the problem. The latter method is generally being used for CFD cuttings transport modeling purposes as less computational power is required. From Cuttings Transport Modeling  Part 1: Specification of Benchmark Parameters with a Norwegian Continental Shelf Perspective. Available from: https://www.researchgate.net/publica...lf_Perspective So, while, yes, the problem is inherently transient, it seems like most just use steady state simulations for these problems to save on computational power/time. Again, as using a steady state model seems to be accepted in this field (from my observation as well as your information on finding a large amount of papers mentioning a steady state solver was used), it is probably okay to stick with using a steady state solver as well. Looks like it is more important to select the particle tracking scheme. 

February 16, 2021, 01:02 

#5 
New Member
Mehran Janghorbani
Join Date: Feb 2021
Posts: 10
Rep Power: 5 
Thank you very much!
The fact of the matter is that since I am using the student version of ANSYS and therefore I cannot make the mesh small enough to get a converged solution in S.S, therefore, I am using the transient approach (EulerEuler) so that I can reduce the time step size and get convergence. I would be greatly obliged if you (or anyone else) could advise if the S.S pressure based solver in ANSYS has an option to lower the pseudotime step (like the courant number in the density based solver which I cannot use since it does not support multiphase flows) and help converge the solution or are the under relaxation factors the only way to control convergence? (if you think it is a question worth discussing I can start a new thread based on it) With best regards Mehran 

February 20, 2021, 02:14 

#6 
New Member
Mehran Janghorbani
Join Date: Feb 2021
Posts: 10
Rep Power: 5 
I have struggeled with this problem a little more and from what I have read, it seems that a S.S solution can be seen as a transient solution at infinity. If that is true, then in cases where getting a converged steady sate solution is difficult, would it be acceptable to "walk" towards one by doing a transient solution for a sufficiently long enough time? (say 50 complete cycles of the fluid in the fluid domain)


February 20, 2021, 16:05 

#7  
Senior Member
Kira
Join Date: Nov 2020
Location: Canada
Posts: 435
Rep Power: 8 
Quote:
Sorry for the delay, I was working on a project. I think you would be looking at a pseudotransient approach, then. Here's a good thread to go through: Pseudo Transient approach From the thread (and I agree with these points): A transient simulation is to solve the problem in timeaccurate fashion, while pseudo transient is the method to get the steady state solution. The pseudo transient method is an accelerated solver for getting the steady state solution. That means don't apply this for transient problems, only steady state problems. Generally you apply it when you want to get to the solution quicker. Usually this means you have tried SIMPLE and gotten a converged problem but it takes too long. So then maybe you have switch to the COUPLED solver, and here you manage to get a solution slightly quicker but still you want even faster. Then the pseudotransient solver is a good idea. My recommendation is to start with SIMPLE/COUPLED solvers and only have you have a robust experience with these two solvers (for each problem you are dealing with) then to go to the pseudo transient solver. Accelerating the solution generally leads to oscillatory or unstable solutions so you want to make sure that you have a robust strategy in place first. 

Tags 
eulereuler, multiphase, transient 
Thread Tools  Search this Thread 
Display Modes  


Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Galvanizing process  Steady state solution  Ritah  Main CFD Forum  0  June 25, 2019 04:11 
Solver for transonic flow?  Martin Hegedus  OpenFOAM Running, Solving & CFD  22  December 16, 2015 04:59 
Transient Solution looks like Steady State  ljwnow  FLUENT  0  March 26, 2012 01:54 
error message  cuteapathy  CFX  14  March 20, 2012 06:45 
About the difference between steady and unsteady problems  Lisa  Main CFD Forum  11  July 5, 2000 14:37 