Re: How to create restart files
Posted: 21 Dec 2012, 18:26
Hello again,
I am back to the problem with continuation from previous solution. Still I cannot obtain the same solution if I restart from some time step (the difference is small, but there should be no difference). For example, if I find the solutions for time steps from 0 to 10, then if I restart from time step 5, the solutions at time step 6 and the rest are different.
I am solving elasticity problem, thus the equation is of second order. It it written in the ElmerSolver Manual that for second order equation, Time Derivative Order must be set to 2 and that it is always discretized using Bossak method.
If you look in this method, to obtain the solution at time t_n+1, it uses the solution, velocity and the acceleration at time t_n, see eq. (5.7) from ElmerSolver Manual.
Thus, I think that we need to save not only the dispalcement and the velocity, but also the acceleration, in order to continue from previous solution. I did not find anything like Calculate Acceleration = Logical True and probably it uses zero vector for the acceleration. On the other hand, the acceleration at time t_n+1 can be calculated, once the displacement vector and the velocity are known at time t_n+1.
Can someone comment, or give an idea how to continue from previous solution?
Cheers,
Stan
I am back to the problem with continuation from previous solution. Still I cannot obtain the same solution if I restart from some time step (the difference is small, but there should be no difference). For example, if I find the solutions for time steps from 0 to 10, then if I restart from time step 5, the solutions at time step 6 and the rest are different.
I am solving elasticity problem, thus the equation is of second order. It it written in the ElmerSolver Manual that for second order equation, Time Derivative Order must be set to 2 and that it is always discretized using Bossak method.
If you look in this method, to obtain the solution at time t_n+1, it uses the solution, velocity and the acceleration at time t_n, see eq. (5.7) from ElmerSolver Manual.
Thus, I think that we need to save not only the dispalcement and the velocity, but also the acceleration, in order to continue from previous solution. I did not find anything like Calculate Acceleration = Logical True and probably it uses zero vector for the acceleration. On the other hand, the acceleration at time t_n+1 can be calculated, once the displacement vector and the velocity are known at time t_n+1.
Can someone comment, or give an idea how to continue from previous solution?
Cheers,
Stan