griddata command returns all NaN values

Hi,

I am trying to create a contour plot from X and Y values that correspond to a theoretical temperature (Tt) value. I am trying to use the griddata command to interpolate the data so it can be graphed. But, the command returns NaN for most every value. I could really use some help. Thank you.

W = 6;
L = 5;
X1 = zeros(100,1);
Y1 = zeros(100,1);
Tt = zeros(100,1);
b = 1;

for a = 1:100
	X1(a) = (5/200).*a;
    c1 = rem(a,50);
	if c1 == 0
		Y1(a) = (6/200).*b;
		b = b + 1; 
	else 
		Y1(a) = (6/200).*b;
	end
end
end

if a == 100
    for m = 1:100
        syms m1
         Tt_xy = 425.8343968*symsum((2/pi).*(((-1).^(m1+1)+1)./m1).*sin(m1.*pi.*(X1(m)./L)).*(sinh(m1.*pi.*(Y1(m)./L))./sinh(m1.*pi.*(W./L))),m1,[1 5]);
         Tt(m) = Tt_xy;
    end
end
  

n1 = 100;
[x1,y1] = meshgrid(linspace(.0250,5,n1),linspace(.03, 6,n1));
f2 = griddata(X1,Y1,Tt,x1,y1);
f3 = figure;
[C1, h1] = contour(x1,y1,f2);
clabel(C1,h1);

[Image Analyst]

Do you need symbolic variables? Can't you do it with actual numerical variables?

[V]

No I do not need symbolic variables. I need numerical variables. Is that where I am making a mistake? If so I'm not sure how to fix that? Sorry I'm not the greatest at MatLab.

Answer 1

[Walter Roberson]

for a = 1:100
	X1(a) = (5/100).*a;

X1 is linear in a.

    c = a/50;

a is a double precision integer, 1 to 100. c is the integer divided by 50. c will always be real-valued.

    c1 = isreal(c);
	if c1 == 1

so that branch is always going to be taken

		Y1(a) = (6/100).*b;

Y1 in this branch is linear in b. But this branch is the only one that can get executed, so all Y1 are linear in b.

		b = b + 1;

and b is increment by 1 each time the branch is met. But the branch is met every iteration. So every time, through, b will be the same as a at the time of the calculation of Y1. So X1 and Y1 are both linear in a .

	else 
		Y1(a) = (6/100).*b;

This branch is never executed.

	end
end

Your X1 and Y1 are now the same as if you had done

a = 1:100;
X1 = (5/100) .* a;
Y1 = (6/100) .* a;

This is a straight line of values, not even one value that is off of the line.

Later, you interpolate over an entire grid. But the values to interpolate from are strictly co-linear. There is no information available for sidewise movement, so everything off-line that is interpreted is going to be NaN.

[V]

Wow ok, I made a dumb mistake. I was using "isreal" to see if there was a remainder of the division of "a" into 50. In otherwords I only want b to advance when it has reached a whole number, or every 50 iterations. However, that was the wrong command. Thank you so much for the help.

However, now after correcting the formula and using "rem" to test if the remainder is 0 before iterating my "f2" is still returning mostly NaN values. I am still confused if I'm missing something

for a = 1:200
	X1(a) = (5/200).*a;
    c1 = rem(a,50);
	if c1 == 0
		Y1(a) = (6/200).*b;
		b = b + 1; 
	else 
		Y1(a) = (6/200).*b;
	end
end

if a == 200
    for m = 1:200
        syms m1
         Tt_xy = 425.8343968*symsum((2/pi).*(((-1).^(m1+1)+1)./m1).*sin(m1.*pi.*(X1(m)./L)).*(sinh(m1.*pi.*(Y1(m)./L))./sinh(m1.*pi.*(W./L))),m1,[1 5]);
         Tt(m) = Tt_xy;
    end
end
  

n1 = 50;
[x1,y1] = meshgrid(linspace(.0250,5,n1),linspace(.03, 6,n1));
f2 = griddata(X1,Y1,Tt,x1,y1);
f3 = figure;
[C1, h1] = contour(x1,y1,f2);
axis([0.001 5 0.001 6]);
clabel(C1,h1);

[Walter Roberson]

griddata() was the wrong function for that situation.

W = 6;
L = 5;
X1 = zeros(100,1);
Y1 = zeros(100,1);
Tt = zeros(100,1);
b = 1;

for a = 1:200
	X1(a) = (5/200).*a;
    c1 = rem(a,50);
	if c1 == 0
		Y1(a) = (6/200).*b;
		b = b + 1; 
	else 
		Y1(a) = (6/200).*b;
	end
end
if a == 200
    for m = 1:200
        syms m1
         Tt_xy = 425.8343968*symsum((2/pi).*(((-1).^(m1+1)+1)./m1).*sin(m1.*pi.*(X1(m)./L)).*(sinh(m1.*pi.*(Y1(m)./L))./sinh(m1.*pi.*(W./L))),m1,[1 5]);
         Tt(m) = Tt_xy;
    end
end

n1 = 50;
[x1,y1] = meshgrid(linspace(.0250,5,n1),linspace(.03, 6,n1));
%f2 = griddata(X1,Y1,Tt,x1,y1);
F = scatteredInterpolant(X1, Y1, Tt);
f2 = F(x1, y1);

f3 = figure;
[C1, h1] = contour(x1,y1,f2);
axis([0.001 5 0.001 6]);
clabel(C1,h1);