Ways to reduce the calculation time

[LiuPengPeter]

Hello,

In my model, the time_scale is 1.0E+9 (from s to ns). The length dt=1000. So 1.0E+6 of steps have to be used to get the result of 200s. So huge!

What's more, the situation won't be better if I change the time_scale and dt. (Larger dt will be adapted to around 1000 finally.)

Is there any other methods recommended to reduce the calculation time? Will scaling factor help?

Thank you a lot!

My execution is in the following:

[Executioner]
  type = Transient
  scheme = bdf2
  solve_type = 'PJFNK'
  #########PETSC_OPTIONS###################very_important###################
  petsc_options_iname = '-pc_type -sub_pc_type   -sub_pc_factor_shift_type'
  petsc_options_value = 'asm       ilu            nonzero'
  l_max_its = 50
  l_tol = 1e-4
  nl_max_its = 20
  nl_rel_tol = 1e-10
  nl_abs_tol = 1e-10
  end_time = 2.00E+11
[./TimeStepper]
    type = IterationAdaptiveDT
    dt = 1.00E+03
    cutback_factor = 0.8
    growth_factor = 2.0 
    optimal_iterations = 7
 [../]
[]
[Preconditioning]
  active = 'full'
  [./full]
    type = SMP
    full = true
  [../]
  [./mydebug]
    type = FDP
    full = true
  [../]
[]

[aeslaughter]

@fdkong Can you help here?

[fdkong]

It is often problem dependent.

1. Run the code in parallel may be able to make the code faster

2. If the Jacobian matrix is accurate, use solve_typ=newton could speedup a bit

3. If the number of linear iterations is not small, could increase the fill-in levels -sub_pc_factor_levels 2

[LiuPengPeter]

It runs very quickly, about 5 seconds for one time step.
I set the end_time = 1.0E+9, (200 s), but after one day only get the result of 1.0E+7. So maybe 100 days are needed...

The unit of time is ns (10E-9 s) and the simulation won't converge well if I change it. Why is it so small?

Thank you for your help!

Time Step 20412, time = 3.48711e+07, dt = 2063.65
0 Nonlinear |R| = 1.301949e-08
|residual|_2 of individual variables:
w: 1.30195e-08
c: 4.55567e-14
eta1: 9.17821e-19
eta2: 7.08598e-20
eta3: 2.41848e-15
c1: 5.31692e-13
c2: 5.29201e-13
c3: 3.56361e-14
0 Linear |R| = 1.301949e-08
1 Linear |R| = 1.301949e-08
2 Linear |R| = 1.301947e-08
3 Linear |R| = 1.301946e-08
4 Linear |R| = 1.301945e-08
5 Linear |R| = 1.301943e-08
6 Linear |R| = 1.301939e-08
7 Linear |R| = 1.301930e-08
8 Linear |R| = 1.301916e-08
9 Linear |R| = 1.301889e-08
10 Linear |R| = 1.301845e-08
11 Linear |R| = 1.301758e-08
12 Linear |R| = 1.301579e-08
13 Linear |R| = 1.301191e-08
14 Linear |R| = 1.300521e-08
15 Linear |R| = 1.299228e-08
16 Linear |R| = 1.297684e-08
17 Linear |R| = 1.294398e-08
18 Linear |R| = 1.290817e-08
19 Linear |R| = 1.283983e-08
20 Linear |R| = 1.274005e-08
21 Linear |R| = 1.255210e-08
22 Linear |R| = 1.225769e-08
23 Linear |R| = 1.176580e-08
24 Linear |R| = 1.112838e-08
25 Linear |R| = 1.033345e-08
26 Linear |R| = 9.153863e-09
27 Linear |R| = 8.003364e-09
28 Linear |R| = 6.622598e-09
29 Linear |R| = 5.597964e-09
30 Linear |R| = 8.052611e-09
31 Linear |R| = 8.052609e-09
32 Linear |R| = 8.052323e-09
33 Linear |R| = 8.052207e-09
34 Linear |R| = 8.052110e-09
35 Linear |R| = 8.051696e-09
36 Linear |R| = 8.051194e-09
37 Linear |R| = 8.049746e-09
38 Linear |R| = 8.047512e-09
39 Linear |R| = 8.043057e-09
40 Linear |R| = 8.036283e-09
41 Linear |R| = 8.022776e-09
42 Linear |R| = 7.995286e-09
43 Linear |R| = 7.937876e-09
44 Linear |R| = 7.842004e-09
45 Linear |R| = 7.663963e-09
46 Linear |R| = 7.471331e-09
47 Linear |R| = 7.100422e-09
48 Linear |R| = 6.752856e-09
49 Linear |R| = 6.194370e-09
50 Linear |R| = 5.573689e-09
Linear solve did not converge due to DIVERGED_ITS iterations 50
1 Nonlinear |R| = 9.034513e-09
|residual|_2 of individual variables:
w: 8.13015e-10
c: 1.20437e-09
eta1: 3.7189e-15
eta2: 6.26328e-18
eta3: 3.23551e-12
c1: 5.46881e-09
c2: 5.46157e-09
c3: 4.44682e-09
0 Linear |R| = 9.034513e-09
1 Linear |R| = 9.034496e-09
2 Linear |R| = 9.033128e-09
3 Linear |R| = 9.032453e-09
4 Linear |R| = 9.031768e-09
5 Linear |R| = 9.030367e-09
6 Linear |R| = 9.027351e-09
7 Linear |R| = 9.020515e-09
8 Linear |R| = 9.010147e-09
9 Linear |R| = 8.988535e-09
10 Linear |R| = 8.955818e-09
11 Linear |R| = 8.893175e-09
12 Linear |R| = 8.763566e-09
13 Linear |R| = 8.504297e-09
14 Linear |R| = 8.103937e-09
15 Linear |R| = 7.454319e-09
16 Linear |R| = 6.839006e-09
17 Linear |R| = 5.880821e-09
18 Linear |R| = 5.170450e-09
19 Linear |R| = 4.222539e-09
20 Linear |R| = 3.376157e-09
21 Linear |R| = 2.587485e-09
22 Linear |R| = 1.984959e-09
23 Linear |R| = 1.508259e-09
24 Linear |R| = 1.178323e-09
25 Linear |R| = 9.336882e-10
26 Linear |R| = 7.058346e-10
27 Linear |R| = 5.551710e-10
28 Linear |R| = 4.198611e-10
29 Linear |R| = 3.374034e-10
30 Linear |R| = 5.871209e-09
31 Linear |R| = 5.847506e-09
32 Linear |R| = 2.595568e-09
33 Linear |R| = 2.313918e-09
34 Linear |R| = 2.085521e-09
35 Linear |R| = 1.767244e-09
36 Linear |R| = 1.421870e-09
37 Linear |R| = 1.051410e-09
38 Linear |R| = 8.035187e-10
39 Linear |R| = 5.871254e-10
40 Linear |R| = 4.542527e-10
41 Linear |R| = 3.245219e-10
42 Linear |R| = 2.340989e-10
43 Linear |R| = 1.600106e-10
44 Linear |R| = 1.141013e-10
45 Linear |R| = 8.264118e-11
46 Linear |R| = 6.363237e-11
47 Linear |R| = 4.606866e-11
48 Linear |R| = 3.695673e-11
49 Linear |R| = 3.015359e-11
50 Linear |R| = 2.427008e-11
Linear solve did not converge due to DIVERGED_ITS iterations 50
2 Nonlinear |R| = 4.032112e-10
|residual|_2 of individual variables:
w: 3.84205e-12
c: 2.47542e-11
eta1: 1.08331e-13
eta2: 6.64417e-17
eta3: 1.3091e-12
c1: 2.84151e-10
c2: 2.84007e-10
c3: 2.34102e-11
0 Linear |R| = 4.032112e-10
1 Linear |R| = 9.910058e-11
2 Linear |R| = 2.129399e-11
3 Linear |R| = 1.666190e-11
4 Linear |R| = 1.177270e-11
5 Linear |R| = 8.907619e-12
6 Linear |R| = 6.862658e-12
7 Linear |R| = 4.720409e-12
8 Linear |R| = 3.422265e-12
9 Linear |R| = 2.714312e-12
10 Linear |R| = 2.103812e-12
11 Linear |R| = 1.671415e-12
12 Linear |R| = 1.394507e-12
13 Linear |R| = 1.047202e-12
14 Linear |R| = 8.452052e-13
15 Linear |R| = 6.944958e-13
16 Linear |R| = 5.150750e-13
17 Linear |R| = 4.053526e-13
18 Linear |R| = 3.429931e-13
19 Linear |R| = 2.808270e-13
20 Linear |R| = 2.169458e-13
21 Linear |R| = 1.632714e-13
22 Linear |R| = 1.290776e-13
23 Linear |R| = 9.948311e-14
24 Linear |R| = 7.628235e-14
25 Linear |R| = 5.842547e-14
26 Linear |R| = 4.751091e-14
27 Linear |R| = 3.907655e-14
Linear solve converged due to CONVERGED_RTOL iterations 27
3 Nonlinear |R| = 7.619928e-13
|residual|_2 of individual variables:
w: 4.22469e-15
c: 4.47435e-14
eta1: 1.57013e-16
eta2: 8.84063e-20
eta3: 2.382e-15
c1: 5.3707e-13
c2: 5.37337e-13
c3: 3.78424e-14
Nonlinear solve converged due to CONVERGED_FNORM_ABS iterations 3
Solve Converged!

[aeslaughter]

Could the problem have reached a steady state?

[fdkong]

What are the initial residuals for the first several steps?

You could use "steady_state_detection=true" to Executioner block, if the steady state is reached, no nonlinear solver will be involved

[LiuPengPeter]

The initial residuals for the first several steps are:

Steady-State Relative Differential Norm:
0.000996499
1.3265e-05
1.33776e-06
1.02387e-06
6.08638e-07
5.03052e-07

[LiuPengPeter]

Do I need normalization to improve the converge?

[GiudGiud]

You could use scaling. Either manually scale the variables with a scaling = xx in the variable blocks or turn on automatic_scaling = true in the Executioner block