Ilian, thank you for the clear ansver!

You could put in a Print statement in the function and see what values NMinimize is calling. Perhaps it's passing the function values outside the range you expect.

Looks like the iterator error message is from the function model.

Did you test krit to see if it works as expected?

Frank, ListConvolve in Compile is not the issue.

With compiled function odezva it also works

odezva = Compile[{{k1o, _Real}, {k2o, _Real}, {n1o, _Real}, {n2o, 
_Real}, {d1o, _Integer}, {d2o, _Integer}, {e1o, _Real}, {e2o, _Real}, 
{Xo, _Real, 1}, {Vo, _Real, 1}},
  Module[{k1 = k1o, k2 = k2o, n1 = n1o, n2 = n2o, d1 = d1o, d2 = d2o, 
    e1 = e1o, e2 = e2o, X = Xo, V = Vo, Xu, Vu, jadroX, jadroV, 
    odezva, Fmean},
   jadroX = k1*(Table[(1 - i/d1)^n1, {i, 0, d1}]);
   jadroV = k2*(Table[(1 - i/d2)^n2, {i, 0, d2}]);
   Xu = Sign[X]*Abs[X]^e1;
   Vu = Sign[V]*Abs[V]^e2;

   odezva = 
    ListConvolve[jadroX, Xu, 1, 0] + ListConvolve[jadroV, Vu, 1, 0] ;
   odezva
   ]
  ]

with a message

CompiledFunction::cfsa: Argument k1 at position 1 should be a machine-size real number. >>

but cca 25% faster..

Do I have a bad day? This my older code is working....

In[1]:= ClearAll["Global`*"]
SetDirectory[NotebookDirectory[]];

In[3]:= data = 
  ReadList["test66-2-0000000949-MOD2.out", Real, RecordLists -> True];

In[4]:= X = data[[All, 2]];
V = data[[All, 3]];
Fexp = data[[All, 4]];
data =.;

In[8]:= odezva[k1o_, k2o_, n1o_, n2o_, d1o_Integer, d2o_Integer, e1o_,
   e2o_, Xo_, Vo_] :=
 Module[{k1 = k1o, k2 = k2o, n1 = n1o, n2 = n2o, d1 = d1o, d2 = d2o, 
   e1 = e1o, e2 = e2o, X = Xo, V = Vo, Xu, Vu, jadroX, jadroV, 
   odezva, Fmean},
  jadroX = k1*(Table[(1 - i/d1)^n1, {i, 0, d1}]);
  jadroV = k2*(Table[(1 - i/d2)^n2, {i, 0, d2}]);
  Xu = Sign[X]*Abs[X]^e1;
  Vu = Sign[V]*Abs[V]^e2;

  odezva = 
   ListConvolve[jadroX, Xu, 1, 0] + ListConvolve[jadroV, Vu, 1, 0] ;
  odezva
  ]

In[9]:= krit[k1o_, k2o_, n1o_, n2o_, d1o_Integer, d2o_Integer, e1o_, 
  e2o_, Fo_, Xo_, Vo_] :=
 Module[{k1 = k1o, k2 = k2o, n1 = n1o, n2 = n2o, d1 = d1o, d2 = d2o, 
   e1 = e1o, e2 = e2o, F = Fo, X = Xo, V = Vo, kriterium, o, Fmean},

  Fmean = Mean[Abs[F]];

  o = odezva[k1, k2, n1, n2, d1, d2, e1, e2, X, V];
  kriterium = 100.*Mean[Abs[o - F]]/Fmean;
  kriterium
  ]

In[10]:= krit[10000., 3000., 0.5, 0.01, 3000, 2000, 3.5, 1.8, Fexp, \
X, V]

Out[10]= 58.8654

In[11]:= res = 
 NMinimize[{krit[k1, k2, n1, n2, d1, d2, e1, e2, Fexp, X, V], 
   k1 > 150 && k2 > 10  && 
    d1 > 10 && d2 > 10  &&
     n1 > .01 && n2 > .01 && 
    e1 > .01 && e2 > .01}, 
  {k1, k2, n1, n2, d1 \[Element] Integers, d2 \[Element] Integers , 
   e1, e2}
  , MaxIterations -> 5
   ]

Speak["Calculation finished!"]



During evaluation of In[11]:= NMinimize::cvmit: Failed to converge to the requested accuracy or precision within 5 iterations. >>

Out[11]= {20.3068, {k1 -> 150.043, k2 -> 10.9215, n1 -> 1.94636, 
  n2 -> 0.712322, d1 -> 12, d2 -> 13, e1 -> 0.679406, e2 -> 0.95152}}

Does the code run if you don't compile the module?

Thank you, Frank, but it is still not clear for me...

I tried a very simple compiled function - NMinimze works with it. The non-compiled version of my function - also works.

It means, that some compiled functions cannot be manipulated symbolically (and cannot by used with NMinimize -- for example my function above), even his non-compiled version can be manipulated symbolically (and can by used with NMinimize) -- whereas other functions can by used with NMinimize even compiled?