MTK programmatic populating initial guess values

Specific Domains Modelling & Simulations
1

Hi all, this should be really easy but I can’t quite figure it out.
When you create an MTK system, and want to solve it as a NonlinearProblem, ODEProblem, or whatever, you often need provide an initial guess or initial conditions.

Not a MWE here, just some generic whiteboard code

sys = mtkcompile(_sys)
guess = [ pump.port_b.m_dot => -100.0, pipe.port_a.p => 10.0, some_valve.port_a.m_dot => 1.0]
# and so on with more guess values

prob = NonlinearProblem(sys, guess)
sol = solve(prob)

what I’d like to do is be able to iterate in a loop building the guess vector from strings, as I know what is in the system. Something like this:

for i in N_valves
push!(guesses, Pair(Symbol("valve_$i.port_a.m_dot"), 33)) 
push!(guesses, Pair(Symbol("valve_$i.port_a.p"), 42)) 
# exact guess values don't matter here
end

However, the type of the guess array is

typeof(guesses)
Vector{Pair{Num, Float64}}

The type of the guess pair vectors doesn’t say anything about a Symbol, and the above for loop doesn’t work. Num is supposed to be an abstract type, although displaying the guess vector shows the names and guess value

julia> guesses
3-element Vector{Pair{Num, Float64}}:
   pump₊port_b₊m_dot(t) => -100.0
   pipe₊port_a₊p(t) => 10.0
 some_valve₊port_a₊m_dot(t) => 1.0

So that is puzzling.

2

@Oscar_Smith @cryptic.ax may have comments here.