Non-linear black box System identification

Question

I am trying to do a non linear black box model identification based on a dataset an input, output

I tried using an ARX polynomial type method, but nothing come out of it.

Then I tried soemthing similar to this: non-linear grey box System identification with Matlab

Any other idea?

Answer 1

You have the identification data something like this:

d_id_w=iddata(w_id,u_id,dt)

And the validation data :

d_vd_w=iddata(w_vd,u_vd,dt)

The ARX model: First, you create it:

m_w_arx=arx(d_id_w,[1,1,1]);

Then you validate the model:

resid(m_w_arx,d_vd_w)

You also may do a comparison between the output of the model and its measured output:

compare(d_vd_w,m_w_arx)