DifferentialEquations.jl Mass Matrix DAE Explicit

Specific Domains Modelling & Simulations
1

Are any of the methods listed at Mass Matrix and Fully Implicit DAE Solvers · DifferentialEquations.jl explicit in time, and therefore not require a Jacobian?

Reason for asking is that solving time-dependent Stokes flow using Ferrite renders use of AD to generate the Jacobian hard.

A MWE is at finite_element_electrical_engineering/project-based-assignment/metal-hydride-storage/notebooks/stokes_2d_channel_quad_transient.ipynb at main · ziolai/finite_element_electrical_engineering · GitHub

2

No. Explicit methods are incapable of solving DAEs.

1 Like
3

Clear. Thx. I guessed so much. Can we plunk in an alternative to AD (finite differences?) to compute the derivatives?

4

yeah, instead of sol = solve(prob_mm, Rodas5(), reltol = 1e-8, abstol = 1e-8), you want

using ADTypes
sol = solve(prob_mm, Rodas5P(autodiff=AutoFiniteDiff()), reltol = 1e-8, abstol = 1e-8)
1 Like