0

How to calculate this integral $W=\int_0^{2\pi}\dfrac{6{\epsilon}{\mu}{\omega}{(R/C)^2}\cdot\left({\epsilon}\cos\left({\theta}\right)+2\right)\sin\left({\theta}\right)}{\left({\epsilon}^2+2\right)\left({\epsilon}\cos\left({\theta}\right)+1\right)^2}{\sin(\theta}){d{\theta}}$ with $P(\theta)=\dfrac{6{\mu}{\omega}{(R/C)^2}{\epsilon}\cdot\left({\epsilon}\cos\left({\theta}\right)+2\right)\sin\left({\theta}\right)}{\left({\epsilon}^2+2\right)\left({\epsilon}\cos\left({\theta}\right)+1\right)^2}$ and $\begin{cases} P(\theta=0)=0,\\ P(\theta=2\pi)=0,\\ \end{cases}$ The value of the integral is supposed to be $W=\dfrac{12{\pi}LR^2{\epsilon}{\mu}{\omega}}{C^2\sqrt{1-{\epsilon}^2}\left({\epsilon}^2+2\right)}$ but somehow I keep getting a zero

I.ham
  • 11
  • 1
    A whole bunch of coefficients can be eliminated here; the $\frac{6\epsilon}{\epsilon^2+2}$ can be eliminated entirely and the other $\epsilon$ in the numerator can potentially be removed as well (though I'm not sure whether it helps to do that splitting or not). Doing this removal of irrelevant coefficients is part of making the math equivalent of a MWE in a software question. – Ian May 21 '22 at 13:55
  • 1
    Relatedly you have a whole bunch of constants in your supposed final answer that haven't been defined in your question. – Ian May 21 '22 at 13:57
  • Anyway, as written the answer is indeed zero, because replacing $\theta$ by $2\pi-\theta$ switches the sign. So you have made a mistake in setup. – Ian May 21 '22 at 14:11
  • I've edited my post I had some missing coefficients, this one is a physic problem P is the pressure and W is the load in the case of a long journal bearing, every textbook I used as a reference are finding the same result I wrote above, but none of them had the steps so I need them to make sure, so is there any way I could get that result based on that integral ? – I.ham May 21 '22 at 15:33
  • You have that extra $\sin(\theta)$, one in the numerator and another outside? That breaks the symmetry that I described. – Ian May 21 '22 at 15:54
  • It's not an extra one it's in the load's equation, basically, the load is $W=\int_0^{2{\pi}}p(\theta)sin({\theta}){d{\theta}}$ – I.ham May 21 '22 at 16:07
  • I understand, but I don't think you had it there in the old version. And it not being there would explain why you were getting zero before. – Ian May 21 '22 at 16:15
  • I am pretty sure I had it, I even did the integral on integral-calculator.com and still get a zero, So I posted here hoping that someone would try to calculate it and see what they get to understand what I am doing wrong. – I.ham May 21 '22 at 16:40
  • 1
    Well, if I missed it then I apologize. Anyway with it there you shouldn't get zero anymore. Maybe someone can help you with the calculation if you show your attempt. (You will be more likely to get helped if you strip out the mathematically irrelevant constants, however; the only one that matters mathematically is the $\epsilon$'s directly multiplying with the cosines.) – Ian May 21 '22 at 16:44
  • It's alright you helped me without even knowing it, I just put my whole problem from where I got stuck as it is, so I give a wider look to the problem, and I'll try to do that, thank you. – I.ham May 21 '22 at 16:50

3 Answers3

1

We see that \begin{align} \int_{0}^{2\pi} \frac{(2+\varepsilon\cos(x))\sin^2(x)}{(1+\varepsilon\cos(x) )^2}\,\mathrm{d}x & \overset{u = x-\pi}{=} 2 \int_{0}^{\pi}\frac{(2-\varepsilon\cos(u) )\sin^2(u)}{(1-\varepsilon\cos(u) )^2}\,\mathrm{d}u\\ & =2 \int_0^{\pi} \frac{1 - \cos(2u)}{(1-\varepsilon\cos(x) )^2}\,\mathrm{d}u - \frac{\varepsilon}{2}\int_0^{\pi} \frac{\cos(u) -\cos(3u)}{(1-\varepsilon\cos(x) )^2}\,\mathrm{d}u \end{align} The problem thus reduces to solving $ \int_{0}^{\pi} \frac{\cos(mx)}{(1-\varepsilon\cos(x))^2}\, \mathrm{d}x $ for integers $m = 0,1,2,3$. But since from this answer we know that

$$ \int_0^\pi\frac{\cos(mx)}{(p-q\cos (x))^2}\ \mathrm{d}x = \frac{\pi \left(p - \sqrt{p^2 - q^2}\right)^m \left(p + m \sqrt{p^2 - q^2}\right)}{q^m\left(p^2 - q^2\right)^{\frac{3}{2}}} \qquad \text{for} \quad |q|<p $$ plugging in $p=1$ and $q = \varepsilon$ for each integral solves the problem \begin{align} \int_{0}^{2\pi} \frac{(2+\varepsilon\cos(x))\sin^2(x)}{(1+\varepsilon\cos(x) )^2}\,\mathrm{d}x & = \frac{ 2\pi}{\left(1 - \varepsilon^2\right)^{\frac{3}{2}}}-\frac{2\pi \left(1 - \sqrt{1 - \varepsilon^2}\right)^2 \left(1 + 2 \sqrt{1 - \varepsilon^2}\right)}{\varepsilon^2\left(1 - \varepsilon^2\right)^{\frac{3}{2}}}\\ & \quad - \frac{\varepsilon}{2}\frac{ \pi\left(1 - \sqrt{1 - \varepsilon^2}\right) \left(1 + \sqrt{1 - \varepsilon^2}\right)}{\varepsilon\left(1 - \varepsilon^2\right)^{\frac{3}{2}}}+\frac{\varepsilon}{2}\frac{ \pi\left(1 - \sqrt{1 - \varepsilon^2}\right)^3 \left(1 + 3 \sqrt{1 - \varepsilon^2}\right)}{\varepsilon^3\left(1 - \varepsilon^2\right)^{\frac{3}{2}}}\\ & = \pi \left[\frac{ 2 - \frac{\varepsilon^2}{2}}{\left(1 - \varepsilon^2\right)^{\frac{3}{2}}}-\frac{3}{2}\frac{ \left(1 - \sqrt{1 - \varepsilon^2}\right)^2 \left(1 + \sqrt{1 - \varepsilon^2}\right)^2}{\varepsilon^2\left(1 - \varepsilon^2\right)^{\frac{3}{2}}}\right]\\ & = \boxed{\frac{2\pi}{\sqrt{1-\varepsilon^2}}} \end{align}

Robert Lee
  • 7,233
  • 2
    Just for your curiosity, if $|q|<p\land q \text{Real}\land p \text{Real}$ $$\int_0^\pi \frac{\cos (m x)}{(p-q \cos (x))^2},dx=\frac{\pi }{(p+q)^2},,, _3\tilde{F}_2\left(\frac{1}{2},1,2;1-m,m+1;\frac{2 q}{p+q}\right)$$ – Claude Leibovici May 22 '22 at 05:34
0

Setting aside the constants, we're integrating $$\int_0^{2 \pi} \frac{(2 + \epsilon \cos\theta) \sin^2 \theta \,d\theta}{(1 + \epsilon \cos \theta)^2}, \qquad \epsilon \in (-1, 1) .$$

Since the integrand has period $2 \pi$, we might as well compute the integral of the function over the symmetric interval $[-\pi, \pi]$, which makes the next step easier.

Apply the Weierstrass substitution $\theta = 2 \arctan t$, $d\theta = \frac{2 \,dt}{1 + t^2}$, which yields the improper rational integral $$8 \int_{-\infty}^{\infty} \frac{[(2 - \epsilon) t^2 + (2 + \epsilon)] t^2 \,dt}{(t^2 + 1)^2 [(1 - \epsilon) t^2 + (1 + \epsilon)]^2} .$$ Decomposing the integrand using the method of partial fractions yields an expression with $8$ unknowns, but the evenness of the integrand in $t$ ensures that the coefficients of the four odd terms are all zero, and solving (assuming $\epsilon \neq 0$) gives $$\frac{1}{\epsilon} \int_{-\infty}^{\infty} \left[\frac{2}{1 + t^2} - \frac{4}{(1 + t^2)^2} + \frac{2 (1 - \epsilon)}{(1 - \epsilon) t^2 + (1 + \epsilon)} + \frac{4 (1 + \epsilon)}{[(1 - \epsilon) t^2 + (1 + \epsilon)]^2}\right] dt ,$$ each of the summands of which is standard. In particular, $\int_{-\infty}^\infty \frac{du}{1 + u^2} = \pi$ and $\int_{-\infty}^\infty \frac{du}{(1 + u^2)^2} = \frac{\pi}{2}$, leaving $$\int_0^{2 \pi} \frac{(2 + \epsilon \cos\theta) \sin^2 \theta \,d\theta}{(1 + \epsilon \cos \theta)^2} = \boxed{\frac{2 \pi}{\sqrt{1 - \epsilon^2}}} .$$ A straightforward substitution shows that this formula remains valid for $\epsilon = 0$.

Travis Willse
  • 99,363
0

A more self-contained approach. Denoting $I = \int_{0}^{2\pi} \frac{(2+\varepsilon\cos(x))\sin^2(x)}{(1+\varepsilon\cos(x) )^2}\,\mathrm{d}x$, then, by substitution $x \to x-\pi$ and noticing that $ (2+\varepsilon\cos(x))\sin^2(x) = \sin^2(x) +(1+\varepsilon\cos(x)) \sin^2(x)$ we get $\require{\cancel}$ \begin{align} I &= 2 \int_{0}^{\pi}\sin(x)\frac{\sin(x)}{(1-\varepsilon\cos(x) )^2}\,\mathrm{d}x + 2 \int_{0}^{\pi}\frac{\sin^2(x)}{1-\varepsilon\cos(x) }\,\mathrm{d}x\\ & \overset{\color{purple}{\text{I.B.P.}}}{=}\cancel{\frac{\sin(x)}{\varepsilon(\varepsilon\cos(x) -1)}\Bigg\vert_{0}^{\pi}} + \frac{2}{\varepsilon} \int_{0}^{\pi}\frac{\cos(x)}{1-\varepsilon\cos(x) }\,\mathrm{d}x+ 2 \int_{0}^{\pi}\frac{\sin^2(x)}{1-\varepsilon\cos(x) }\,\mathrm{d}x\\ & \overset{\color{purple}{\sin^2(x) = 1-\cos^2(x)}}{=} 2 \int_0^\pi \frac{\mathrm{d}x}{1-\varepsilon\cos(x)} + \cancel{\frac{2}{\varepsilon} \int_0^{\pi}\cos(x) \,\mathrm{d}x}\\ & \overset{\color{purple}{t = \sqrt{\frac{1+\varepsilon}{1-\varepsilon}}\tan\left(\frac{x}{2}\right)}}{=} \frac{4}{\sqrt{1-\varepsilon^2}} \int_0^{\infty}\frac{\mathrm{d}t}{t^2+1} = \boxed{\frac{2\pi}{\sqrt{1-\varepsilon^2}}} \end{align}

Robert Lee
  • 7,233