why do we get different responses when simulating the same ODE model with matlab and simulink, even when using the same initial values and parameters.

problem with model calibration

[Sara]

Are you using the same ode solver with the same options?

Answer 1

[rafik]

Hi, Yes, i'm using the same ODE solver "ODE 45", with the same parameters

the system of nonlinear ordinary differential equations is given by the following equations :

Knowing that :

For initial conditions equals respectively to : [1,05 ; 0,001; 2, 0] for t=1 :1 :48 And parameter values equals to : Mu=0,6 ;K=3,6 ;Delta=0,087 ; Sigma=2 ; S-inf=0,35 ; beta=0,22 ; Y=0,63 ;phi=0,146 ; v1,36= . The problem i face is an inconsistence in results given when i model the system using matlab command line, and when modeling it with simulink, the inconsistencies rellies on differences observed when i compare my results with those obtained by the author of the system which are for N given by a sigmoidal curve starting from 0 and reaching a steady state at a value of 1 or a bit more, and from 0 to 2,5 or 3 for P . For S we get a sigmoidal decay curve starting from 2 and finiching at 0 . And finaly, for Q we have a bell curve reaching top at value equal to 0,3 ; 0,4 and decreasing until it reach zero . Thank you for your help.

[Sara]

Was the simulink model implemented by someone else? You may want to try to use odeset/odeget to check with tolerance is set in the two systems (simulink vs. command line)

[rafik]

Hi, Sara no one has implemented the simulink model except me, and you may be right as i didn't use any ODE options on the command line code when simulating the model. on the other hand when running the simulink model i had to use some parametrization as i faced a Singularity problem so i had to put a lower step size and tolerance. in sum, the closest best result when compared to the published results was the command line code implementation. i'would be thank full if you could check the correctness of the two models, i could probably made a mistake some where !!. Thank you.

[Sara]

I don't have the simulink so I can't test it, unfortunately. I would recommend setting the same options in both the simulink and the matlab to be sure that they match first. Can you attach the published results? I could take a look at the command line implementation if that could help.