2D arrays with ModelingToolkit

Specific Domains Modelling & Simulations
, ,
22

I also implemented that in the following Julia package: https://github.com/ufechner7/KiteModels.jl/blob/main/src/KPS4.jl , but in this package I was doing these calculations numerically and not symbolically. It uses a lookup table:

const SPRINGS_INPUT = [0.    1.  150.
                       1.    2.   -1. # s1, p7, p8
                       4.    2.   -1. # s2, p10, p8                        
                       4.    5.   -1. # s3, p10, p11
                       3.    4.   -1. # s4, p9, p10
                       5.    1.   -1. # s5, p11, p7
                       4.    1.   -1. # s6, p10, p7
                       3.    5.   -1. # s7, p9, p11
                       5.    2.   -1. # s8, p11, p8
                       2.    3.   -1.] # s9, p8, p9

to define which springs are connecting which point massesā€¦

I donā€™t see any issue to do something similar symbolicallyā€¦ But it would take a lot of time, it is very easy to make mistakes and they are hard to debugā€¦

In other words, perhaps the better approach with ModelingToolkit is to create components and connect them, but I donā€™t have any experience with this approachā€¦

23

Yes I think it is the way to go too. I wonder, if this necessitates connectors, or if a more straightforward approach like in my previous post is enough.

Another thing. I get some errors related to symbolic vectors when calling structural_simplify without specifying only scalar parameters in ODESystem. Such as the following.

julia> sys = structural_simplify(model)
ERROR: MethodError: no method matching hasmetadata(::Vector{SymbolicUtils.BasicSymbolic{Real}}, ::Type{Symbolics.VariableDefaultValue})

Did you get such errors ?

@ChrisRackauckas, is this still a missing functionality, should I file an issue ?

EDIT : So I found a way to circumvent this problem by (without much surprise) iterating through the symbolic vectors. So for instance, if you have an ODESystem such as

@named _model = ODESystem(eqs, t)

where some symbolic parameters are arrays, one way to make it work is to do

@named _model = ODESystem(eqs, t, [<list of all symbolic variables of interest, even arrays>], [mysymbolicparameterarray...])

With the same logic, getting the default value of a given parameter array dos not work so something like [ModelingToolkit.getdefault(d) for d in sys.mysymbolicparameterarray] seems to be the only solution.

24

For the record it seems such an implementation is presented in Exposing More Parallelism By Tearing Algebraic Equations in ODESystems Ā· ModelingToolkit.jl

<...>
eqs = [
    D(E) ~ sum(((i, sys),) -> getproperty(sys, Symbol(:resistor, i)).h.Q_flow,
        enumerate(rc_systems)),
]
<...>