Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have a vector field (velocity) and I want to integrate around a rectangle. If I had a function for the field it would be easy and I know how to do it. However, I only have the coordinates. Is there an easy way of doing this?

(http://i.stack.imgur.com/soxfY.png)

I thought about how I'm going to solve it, started writing the steps for the solution: parametrise each line, find the derivative of the parametrisation. However, I got stuck because in the integral, the field has to be evaluated at the parametric function.

$$\begin{align} \int_C \vec F \cdot d\vec r &= \int_C \vec F \cdot \vec T \: ds \\ &= \int_a^b \vec F\left(\vec r(t)\right) \cdot \frac{\vec r'(t)}{\left\lVert \vec r'(t) \right\rVert} \left\lVert \vec r'(t) \right\rVert \: dt \\ &= \int_a^b \vec F\left(\vec r(t)\right) \cdot \vec r'(t) \: dt \end{align}$$

The first part of the last line is where I got stuck, t would be involved and I don't know how to do it numerically.

* Side note, I'm trying to calculate circulation which can also be found by calculating the surface integral of the curl of velocity (vorticity), but I guess that would be even harder numerically

I use MatLab but anything that would get me started would do

Thanks


Edit: The above equation was copied from a website, let me use my own notation:

I have a vector field $\vec V = \vec V(x,y)$ which has an x component $u$ and a y component $v$.

$\Rightarrow \vec V(x,y) = u(x,y) \; \vec i + v(x,y) \; \vec j $

I want to integrate along C which is a rectangle so I will have to divide the integration into 4 parts (4 smooth curves).

The vertices of the rectangle are $(0.0274,0.0386)\: (0.0926,0.0386) \: (0.0926,0.1184) \: (0.0274,0.1184)$

So the four curves are: $$\begin{align} C_1: \: \: r_1(t) = t \; \vec i + 0.0386 \; \vec j \hspace{2cm} for\hspace{1cm} 0.0274 \le t \ge 0.0926 \\ C_2: \: \: r_2(t) = 0.0926 \; \vec i + t \; \vec j \hspace{2cm} for\hspace{1cm} 0.0386 \le t \ge 0.1184 \\ -C_3: \: \: r_3(t) = t \; \vec i + 0.1184 \; \vec j \hspace{2cm} for\hspace{1cm} 0.0274 \le t \ge 0.0926 \\ -C_4: \: \: r_4(t) = 0.0274 \; \vec i + t \; \vec j \hspace{2cm} for\hspace{1cm} 0.0386 \le t \ge 0.1184 \end{align}$$ and the derivatives: $$\begin{align} r_1'(t) &= 1 \; \vec i + 0 \; \vec j \\ r_2'(t) &= 0 \; \vec i + 1 \; \vec j \\ r_3'(t) &= 1 \; \vec i + \; \vec j \\ r_4'(t) &= 0 \; \vec i + 1 \; \vec j \end{align}$$ then: $$\begin{align} \Gamma = \int_C \vec V(x,y) \cdot d\vec r &= \\ &= \int_{C_1} \vec V(x,y) \cdot d\vec r + \int_{C_2} \vec V(x,y) \cdot d\vec r - \int_{-C_3} \vec V(x,y) \cdot d\vec r - \int_{-C_4} \vec V(x,y) \cdot d\vec r \\ &= \int^{b_1}_{a_1} \vec V(\vec r_1(t)) \cdot r_1'(t) dt + \int^{b_2}_{a_2} \vec V(\vec r_2(t)) \cdot r_2'(t) dt \\ &- \int^{b_3}_{a_3} \vec V(\vec r_3(t)) \cdot r_3'(t) dt -\int^{b_4}_{a_4} \vec V(\vec r_4(t)) \cdot r_4'(t) dt \end{align}$$

This is where I got to, I know that $\vec V(\vec r(t)) = \vec V(x(t),y(t))$, here I cannot think of a way of integrating numerically (mainly the $t$ term in $r(t)$ confuses me). $x$ and $y$ are 2-D grids (a 41 by 35 matrix) where the columns in x are equal and the rows in y are equal. x-component $u(x,y)$ is also a 41 by 35 matrix and so is the y-component $v(x,y)$

share|improve this question
    
I can't post images yet! here are the two images: i.stack.imgur.com/soxfY.png i.stack.imgur.com/CxJUj.gif –  pythonista Feb 24 '12 at 20:03
    
So the function $F$ is sampled on an evenly spaced grid? Do you have an analytic expression for $C$? If this is the case, fit an interpolating polynomial to $F(r(t))$ and compute $r'(t)$ analytically, then integrate the product of these two functions numerically along the interval $[a,b]$. –  dls Feb 24 '12 at 20:26
    
I added the image and the equation to your post. How exactly do you have your vector field $F$? Is it a pair of rectangular arrays for the $x$ and $y$ components? –  Rahul Feb 24 '12 at 20:26
    
Thanks for editing the question. The vector F is as @dls wrote; it is sampled on an evenly spaced grid. The x and y are given as if produced by the meshgrid function (Rectangular grid in 2-D and 3-D space) While the components of the vector are ud and vd, the first is the x component and the second is the y component. *C is the rectangle, and I do have the vertices. –  pythonista Feb 24 '12 at 23:45
    
i am stuck with same proble. can you share the code or algo –  user77073 May 9 '13 at 19:45

1 Answer 1

up vote 1 down vote accepted

Your work looks pretty good, but you can actually take it a step further since you know $r_i'(t)$. When you substitute in this information, each integral depends only on one component of $\vec V$, but not both. For instance $$\int^{b_1}_{a_1} \vec V(\vec r_1(t)) \cdot r_1'(t) \; dt =\int^{b_1}_{a_1} u(\vec r_1(t)) \; dt$$

The next task is to write a routine to implement the function $\vec V$, that is, you want a function from $\mathbb{R}^2\rightarrow\mathbb{R}^2$. Use something like interp2 (though this may not be the most efficient). Once you have this accomplished, simply evaluate the now 1-d integrals using the quad function, for instance.

share|improve this answer
    
Thanks a lot. I used interp2 but I had no idea how to use its output. This lead me to something easier: the surface fitting tool. I fitted the surface then used feval to evaluate the fit at $r(t)$ and finally used trapzto evaluate the integral. However, feval returns lots of NaN, when I increase the number of evaluation points it gets better. Anyway, as you said, it's not efficient. I did it the way I thought it would be harder; surface integrals. The sfit object can be integrated using quad2d. So I integrated the curl of $\vec V$ over the rectangle and I got better results. –  pythonista Feb 26 '12 at 2:33

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.