# Simulation time for steady state solution using transient solver

 Register Blogs Members List Search Today's Posts Mark Forums Read

 June 26, 2019, 05:03 Simulation time for steady state solution using transient solver #1 New Member   Sebastian Pelletier Join Date: May 2019 Posts: 12 Rep Power: 6 This may be a slightly silly question and I`m probably misunderstanding how transient simulations run, but when I am trying to determine what an acceptable real run time for a transient simulation that is meant to find a final steady state solution. Can I base my assumption on the travel time of the fluid through the domain? For example if the fluid is traveling at 10m/s and the domain is 1m in length is it safe to assume that the approximate length of a single real time period would be .1s? or is the total simulation time not analogous in that way to fluid velocity? If my values of interest (Heat transfer coefficient) appears to have reached steady state would using this estimation be another appropriate check?

June 26, 2019, 05:08
#2
Senior Member

Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,738
Rep Power: 71
Quote:
 Originally Posted by spelletier This may be a slightly silly question and I`m probably misunderstanding how transient simulations run, but when I am trying to determine what an acceptable real run time for a transient simulation that is meant to find a final steady state solution. Can I base my assumption on the travel time of the fluid through the domain? For example if the fluid is traveling at 10m/s and the domain is 1m in length is it safe to assume that the approximate length of a single real time period would be .1s? or is the total simulation time not analogous in that way to fluid velocity? If my values of interest (Heat transfer coefficient) appears to have reached steady state would using this estimation be another appropriate check?

No, you get only an estimation for the convective time in which a fluid particle travels in to the domain.
For laminar steady flow, one can assume the non-dimensional time for a steady state to be proportional to the Re number. But in case of a URANS formulation used to get towards the steady RANS solution this is more complex as the physical viscosity is locally changed by the turbulence model.
The proper check for the steady state is to use a specific norm on the time derivatives of the variable. Theoretically they should vanish, in practice you can set a proper threshold.