Hi.I have been trying a lot to "find" the answer. I still can not get it.

I have a MATLAB function which uses the FIND command, and when I use the same function in Simulink I am getting an error. Can someone please tell me how to implement FIND command in Simulink?

[Geoff Hayes]

Sanjal - please post the code that you have written (that calls find) and post the full error message (all the red text).

[Sanjal C C]

x1=(0:0.001:50 )'; 
x1= round(x1,3 ); 

lperc=find(x1==px);

I want to model this in Simulink but the find is not working exactly as that in the script. I want to find the value px in the corresponding x1 array. So could someone please help me how to do it in Simulink?

[Adam]

You still haven't posted the error

[Guillaume]

Unrelated to your question (which I can't answer since I don't know simulink), but ultimately probably more important, it looks like you need to learn about floating point comparison. Read this to learn more.

A more reliable code would be:

x1 = (0:0.001:50 )';
lperc = find(abs(px - x1) <= 1e-8);  %arbitrary small value much smaller than the precision of x1.

Answer 1

[John D'Errico]

The error? Yours, in making an assumption about floating point arithmetic, and how numbers are stored.

x1=(0:0.001:50 )'; 
x1 = round(x1,3 );

Now, lets look at x1(2), for example. The same will apply to nearly every other element in x1.

x1(2)
ans =
                   0.001

It looks like 0.001. But is it?

sprintf('%0.55f',x1(2))
ans =
    '0.0010000000000000000208166817117216851329430937767028809'

Almost all such floating point numbers cannot by exactly stored, because you are creating decimal fractions, but storing them in binary form.

Just as you cannot represent 1/3 exactly as a decimal number, you cannot represent 0.001 exactly in binary form, using a finite number of binary bits. While it looks like 1/1000 to you, in binary form, it will be a infinitely repeating string of binary bits.

That rounding operation did not in fact round numbers to exactly 3 places.

So NEVER test for exact equality of floating point numbers. Use a tolerance instead. (If you know what you are doing, and know when it is safe, integers can be tested for equality, even when the integers are doubles, within limits.)