How to solve these two first order non-linear coupled differential equations that represents the velocity field of a vortex?

91 Views Asked by At

enter image description here What is the solution of these coupled differential equations ?

$\frac{dx}{dt}=-sin(2\pi(y-\alpha))sin^2(\pi (x-\beta)) $ eqn 1

$\frac{dy}{dt}= \;\; sin(2\pi(x-\beta))sin^2(\pi (y-\alpha))$ eqn 2

$\alpha$ and $\beta$ are constants. These are the equations of a vortex velocity field. Figure shows the result that I got when I tried to numerically follow the path for intial cordinates (x,y) = (6.5,6.25) and $(\alpha,\beta) = (5.76,6.01)$ with $dt = 1/12$ for 30 million iterations. enter image description here

1

There are 1 best solutions below

8
On BEST ANSWER

$$\frac{dx}{dt}=-2 \sin(\pi(y-\alpha))\cos(\pi(y-\alpha)) \sin^2(\pi (x-\beta))\tag 1$$
$$\frac{dy}{dt}= 2\sin(\pi(x-\beta))\cos(\pi(x-\beta))\sin^2(\pi (y-\alpha)) \tag 2$$ $$\frac{dy}{dx}=-\frac{\cos(\pi(x-\beta))\sin(\pi (y-\alpha)) }{\cos(\pi(y-\alpha)) \sin(\pi (x-\beta))} \tag 3$$ The ODE $(3)$ is separable. $$\frac{\cos(\pi (y-\alpha)) }{ \sin(\pi(y-\alpha)) }dy=-\frac{\cos(\pi(x-\beta)) }{\sin(\pi (x-\beta))}dx$$ $$\ln|\sin(\pi(y-\alpha))| = -\ln|\sin(\pi(x-\beta))|+\text{constant}$$ General equation of the trajectories : $$\boxed{\sin(\pi(y-\alpha))\sin(\pi(x-\beta))=C}\tag 4$$ Each trajectory depends on $C$ which is defined from the initial point $(x_0\;,\;y_0)$

$$C=\sin(\pi(y_0-\alpha))\sin(\pi(x_0-\beta)) \tag 5$$

One put $\sin(\pi(y-\alpha))=\frac{C}{\sin(\pi(x-\beta))}$ from $(4)$ into $(1)$ in order to eliminate $y$. $$\frac{dx}{dt}=-2 \frac{C}{\sin(\pi(x-\beta))} \sqrt{1-\frac{C^2}{\sin^2(\pi(x-\beta))}} \sin^2(\pi (x-\beta))$$ $$\frac{dx}{dt}=-2C \sqrt{\sin^2(\pi(x-\beta))-C^2}$$ $$t=-\frac{1}{2C}\int_{x_0}^x \frac{d\chi}{\sqrt{\sin^2(\pi(\chi-\beta))-C^2}}$$ $$t=-\frac{1}{2C}\int_{x_0}^x \frac{d\chi}{\sqrt{1-\sin^2(\frac{\pi}{4}-\pi(\chi-\beta))-C^2}}$$ $$t=-\frac{1}{2C\sqrt{1-C^2}}\int_{x_0}^x \frac{d\chi}{\sqrt{1-\frac{1}{1-C^2}\sin^2(\frac{\pi}{4}-\pi(\chi-\beta))}} \tag 6$$ $$t=-\frac{1}{2\pi C\sqrt{1-C^2}}\int_{\pi(x_0-\frac{1}{4}+\beta)}^{\pi(x-\frac{1}{4}+\beta)} \frac{d\xi}{\sqrt{1-\frac{1}{1-C^2}\sin^2(\xi)}}$$

$$t(x) = -\frac{1}{2\pi C\sqrt{1-C^2}}\left( \text{F}\left(\pi(x-\frac{1}{4}+\beta)\:,\:\sqrt{\frac{1}{\sqrt{1-C^2}}}\right) -\text{F}\left(\pi(x_0-\frac{1}{4}+\beta)\:,\:\sqrt{\frac{1}{\sqrt{1-C^2}}}\right)\right)$$ $\text{F}$ denotes the Elliptic Integral of the first kind : http://mathworld.wolfram.com/EllipticIntegraloftheFirstKind.html

For the inverse function $x(t)$ one can numerically integrate $(6)$ from $x_0$ up to the value of $x$ such as the value of $t$ be reached.

Proceed similarly from equations $(2)$ and $(4)$.

COMMENT :

The above result, analytical solution $(4)$ , is in contradiction with the numerical result from the OP computation. The analytical result is a periodic function which graphical representation is a closed curve.

My own numerical solving doesn't agree with the OP. On the figure below the curve from equation $(4)$ is drawn in red. The numerical calculus gives the curve drawn in black. They are undistinguishable one from the other. This confirmes that my analytical and numerical calculus are consistent.

enter image description here

I think that the increment of $t$ which was chosen $=1/12$ by the OP is much too large. The small numerical errors accumulate and cause the deviation from a closed curve to something like a spiral. I think that the increment of $t$ must be much smaller, for example $0.00001$ or less.

The above curve seems close to a circle because the chosen initial point $(x_0,y_0)$ is close to $(\alpha,\beta)$.

Another example below shows the result with initial point less close to $(\alpha,\beta)$ :

enter image description here