Facing problems with Sundials.jl (ARKODE) with discrete callback

ChrisRackauckas · September 12, 2024, 9:08am

Is that ad-hoc one with error control? My guess is that with the large system, it might be stiff and the adaptivity could be turning the dt down?

There is a CashKarp in the ExplicitRK form. ExplicitRK(tableau=constructCashKarp()) (requires using DiffEqDevTools).

github.com

SciML/DiffEqDevTools.jl/blob/v2.45.0/src/ode_tableaus.jl#L1227-L1248


      
          """
          constructCashKarp()
          
          Constructs the tableau object for the Cash-Karp Order 4/5 method.
          """
          function constructCashKarp(T::Type = Float64)
              A = [0 0 0 0 0 0
                   1//5 0 0 0 0 0
                   3//40 9//40 0 0 0 0
                   3//10 -9//10 6//5 0 0 0
                   -11//54 5//2 -70//27 35//27 0 0
                   1631//55296 175//512 575//13824 44275//110592 253//4096 0]
              c = [0; 1 // 5; 3 // 10; 3 // 5; 1; 7 // 8]
              α = [37 // 378; 0; 250 // 621; 125 // 594; 0; 512 // 1771]
              αEEst = [2825 // 27648; 0; 18575 // 48384; 13525 // 55296; 277 // 14336; 1 // 4]
              A = map(T, A)
              α = map(T, α)
              c = map(T, c)
              αEEst = map(T, αEEst)
              return (ExplicitRKTableau(A, c, α, 5, αEEst = αEEst, adaptiveorder = 4,

This file has been truncated. show original

For examples of ExplicitRK in use with chosen tableaus, you can see for example:

github.com

SciML/OrdinaryDiffEq.jl/blob/v6.89.0/test/interface/ode_initdt_tests.jl#L1-L10


      
          using OrdinaryDiffEq, DiffEqDevTools, Test
          import ODEProblemLibrary: prob_ode_linear, prob_ode_2Dlinear
          
          prob = prob_ode_linear
          sol = solve(prob, Rosenbrock32())
          dt₀ = sol.t[2]
          
          prob = prob_ode_2Dlinear
          sol = solve(prob, ExplicitRK(tableau = constructBogakiShampine3()))
          dt₀ = sol.t[2]

Though I can almost guarantee you that it’s not a tableau thing and is likely more to do with adaptivity trying to use error controls and your adhoc version not being error corrected. But you can give this a try and see.

Topic		Replies	Views
How to use DiscreteCallbacks with sundials.jl IDA solver? How to implement reinit!? Numerics question , diffeq , differentialequation , diffeqcallbacks	8	1071	February 3, 2022
When to use a discrete callback in DifferentialEquations.jl General Usage	2	1101	December 10, 2020
ModelingToolkit: Building a DAEProblem and solving it with IDA with KLU & Jacobian usage Modelling & Simulations modelingtoolkit	3	1001	March 20, 2022
Discrete callback + Saving callback for changing parameters in DifferentialEquations doesn't work as intended Numerics question , diffeq , differentialequation	11	120	September 12, 2024
DiscreteCallback takes a break: discrete callback shows counterintuitive behaviour in DifferentialEquations.jl General Usage question , numerics , differentialequation , diffeqcallbacks	2	195	April 11, 2024

Facing problems with Sundials.jl (ARKODE) with discrete callback

Related topics