structural_simplify removes variable it needs later
#3159
Comments
|
While the error message is a little misleading and should be fixed, this isn't a bug. In the first ("not working") system, julia> sys_simplified = structural_simplify(sys)
Model sys with 1 equations
Unknowns (1):
S_sep_out2(t)
Parameters (2):
S_in [defaults to 1]
q_in [defaults to 1]
julia> observed(sys_simplified)
5-element Vector{Equation}:
S(t) ~ q_in
q(t) ~ S_in
q_sep_out2(t) ~ 0.5
S_sep_out1(t) ~ S_sep_out2(t)
q_sep_out1(t) ~ -q_sep_out2(t) + q(t)Since julia> prob = ODEProblem(sys_simplified, [], (0, 1); guesses = [S_sep_out2 => 0.1])
ODEProblem with uType Vector{Float64} and tType Int64. In-place: true
timespan: (0, 1)
u0: 1-element Vector{Float64}:
0.0
julia> sol = solve(prob, Rodas5())
┌ Warning: Rosenbrock methods on equations without differential states do not bound the error on interpolations.
└ @ OrdinaryDiffEqCore ~/Julia/SciML/OrdinaryDiffEq.jl/lib/OrdinaryDiffEqCore/src/solve.jl:108
retcode: Success
Interpolation: specialized 4rd order "free" stiffness-aware interpolation
t: 8-element Vector{Float64}:
0.0
1.0e-6
1.1e-5
0.00011099999999999999
0.0011109999999999998
0.011110999999999996
0.11111099999999996
1.0
u: 8-element Vector{Vector{Float64}}:
[1.0]
[1.0]
[1.0]
[1.0]
[1.0]
[1.0]
[1.0]
[1.0]The solution is constant because this is a system of nonlinear equations, not a DAE/ODE; there are no differential equations. As for your "working" example, julia> sys_simplified = structural_simplify(sys)
Model sys with 0 equations
Unknowns (0):
Parameters (2):
S_in [defaults to 1]
q_in [defaults to 1]And creating the |
|
Thanks a lot for the quick and detailed reply! It is very helpful and makes sense to me. Indeed the error message is misleading (especially also since the wrong variable is mentioned, However, now, I'm running into another problem, and this is that the simulation seems to get stuck in the initialization (takes over 1h and I put in a progress printing callback which gets never called until then). I'm still working on it to make a MWE and might come back if I manage to create one. But a general question regarding guesses: Can it be a problem if too many guesses are provided? Or are the additional (for variables removed during simplification) just ignored? |
|
Too many guesses shouldn't be a problem. As for the infinite loop, I fixed an infinite loop in [email protected]. Does updating to that version (or later) help? |
|
I just updated and tested now with [email protected] and [email protected] and I'm still facing the infinite loop. Now, however, once I canceled now after ~20min it showed an |
|
Ok nevermind. It seems that I just selected especially bad initial guesses which apparently made solving the initialization system taking a very long time (~30min, thus before it was most likely indeed the one you fixed in Symbolics and now it was just taking a very long time). Picking better initial conditions now yielded a faster response and a physically sensible result. Thanks a lot for your help! Shall I close this issue then? |
That's probably the maxiters on the recursion detection. 30 minutes would be too high 😅, we should probably make the maxiters a bit lower. |
|
I agree that this is a very long time. But I guess this depends on the number (and size) of loops that are there. I work with a system that has ~450 unknowns after simplification and in there are 6 loops which are interconnected. +1 additional loop over all of these 6 together. So I guess I have quite some loops involved here 😅 (Although only few of them are algebraic, I guess the only case where substitutions are involved) I guess how many iterations there are made, depends on how the maxiters are implemented: if they are applied to all substitutions together (i.e. it is an absolute limit on substitutions) or per-iteration (i.e. limit on applying the same substitution circle, thus in each iteration of the outer loop it could make maxiters iterations on the inner ones). |
Describe the bug 🐞
Building an ODEProblem from an ODESystem throws an error indicating missing variables, that were optimized away during
structural_simplify.By manually combining two of the equations, I can get it to work it for the MWE below.
Expected behavior
Building the ODEProblem should work no matter how the equations are specified.
Minimal Reproducible Example 👇
Failing version:
Working version:
Error & Stacktrace⚠️
Environment (please complete the following information):
using Pkg; Pkg.status()using Pkg; Pkg.status(; mode = PKGMODE_MANIFEST)versioninfo()Additional context
I'm currently working on a library building on top of ModelingToolkit which aims to simulate biochemical reactions (especially for wastewater treatment). Applying it to one of these plants, this error occurred (in a much more complex model). Doing the rewriting myself would not be feasible due to the complexity of the overall model.
The text was updated successfully, but these errors were encountered: