1

I want an equation $y=a\cosh(\frac{x-b}{a})+c$ for a catenary opening upward that passes through $(x_1,y_1)$ and $(x_2,y_2)$. WLOG, assume $x_2>x_1$.

Let $L$ be the length of the catenary between $(x_1,y_1)$ and $(x_2,y_2)$

Simon M
  • 657
  • If $L$ is greater than the distance between the two points, then the low point of the catenary (assuming it opens up, so $a > 0$), which occurs at $x = b$, lies between $x_1$ and $x_2$. Thus $(x_1 - b)/a < 0$. But $\cosh^{-1}$ only returns positive values. Thus $b = x_1 + a\cosh^{-1}((y-c)/a)$, not "$-$" as you claim. – Paul Sinclair Aug 28 '23 at 17:36
  • @PaulSinclair The low point of the catenary (where the tangent line has slope zero) can occur anywhere with respect to $x_1$ and $x_2$. $x_1$ and $x_2$ are just two points that reside somewhere on the catenary. I did not specify that the low point of the catenary would have to occur at an x value between $x_1$ and $x_2$. I left the order intentionally ambiguous to allow for greater generality. – Simon M Sep 12 '23 at 16:44
  • You've ignored the condition that I lead with: $L$ is greater than the distance between the two points. However, I will grant you that this limit was not fully sufficient for the claim. It does require a somewhat greater length to ensure the low point is between $x_1$ and $x_2$. But In cases where the length is long enough, the low point will occur between $x_1$ and $x_2$, and then your formula gives the wrong result. You did not allow for greater generality. What you did was ignore the fact that $\cosh$ is not a 1-1 function, so only has a limited inverse. – Paul Sinclair Sep 12 '23 at 17:22
  • @PaulSinclair When I looked back over this post I did realize that I had unwittingly proceeded with my formula without acknowledging that since $cosh$ can't pass the vertical line test its inverse will be, as you say, limited. What I'll try to do now is create the generality I desire by the appropriate placements of ± when b is in a given interval with respect to $x_1$ and $x_2$ – Simon M Sep 12 '23 at 17:32
  • @PaulSinclair and I'll also be trying to derive a formula for the length needed in order to ensure that the low point occurs between the two specified points. I imagine such a formula has long since existed in mathematics and been used in physics calculations. – Simon M Sep 12 '23 at 17:39
  • horizontal* line test – Simon M Sep 12 '23 at 17:56
  • 1
    See this question and answer of mine: https://math.stackexchange.com/questions/3557767/how-to-construct-a-catenary-of-a-specified-length-through-two-specified-points – marty cohen Sep 12 '23 at 23:36
  • See https://physics.stackexchange.com/a/279247/392 for the equations – John Alexiou Sep 14 '23 at 00:03

3 Answers3

1

Take the horizontal separation (the span) $S = x_2-x_1$ and the vertical separation $h = y_2-y_1$.

But you also need the catenary constant $a$ which describes the curvature of the curve (in units of length). If total length $L$ is given then one needs to solve for $a$ in

$$ L = a \left( \sinh \left( \frac{ x_2 - x_c}{a} \right) - \sinh \left( \frac{x_1-x_c}{a} \right) \right) \tag{1} $$

Where the lowest point is $(x_c,y_c)$ that is evaluated with the following exact equations.

$$ x_c =x_1 + \frac{S}{2} +a\, \sinh^{-1} \left( \frac{h \hat{e}^{S/(2a)}}{a(1-\hat{e}^{S/a})} \right) \tag{2}$$

and

$$y_c = y_1 -a \left( \cosh \left( \frac{x_c-x_1}{a} \right) -1 \right) \tag{3}$$

So if $L$ is given, one must solve (1) and (2) simultaneously which is not possible to do analytically because (1) is of the form of $y = x \exp(x)$ which is unsolvable for $x$.

The combined (1) and (2) equations are

$$ L = 2 a \sinh \left( \frac{S}{2 a} \right) \sqrt{1 + \left( \frac{h \hat{e}^{S/(2a)}}{a(1-\hat{e}^{S/a})} \right)^2} $$

which must be solved for $a$ when $L$ is given. But in many cases, the horizontal tension $H$ and the unit weight of the cable is given, providing with a direct estimation of $a = \frac{H}{w}$.

The equation for the curve is

$$ \boxed{ y(x) = y_1 + a \left( \cosh \left( \frac{x-x_c}{a} \right)- \cosh \left( \frac{x_1-x_c}{a} \right) \right) } \tag{4}$$

You can try for yourself to see that $y(x_1) = y_1$ and $y(x_2)=y_2$

An example of the equations above is shown below

fig1

taken from my answer(s) in SE #1 and SE #2

Example is $(x_1,y_1) = (50,15)$ and $(x_2,y_2) = (450, 45)$ with a catenary constant $a = H/w = 300$

  • These points result in

    $$S=450-50 = 400$$

    and

    $$h = 45-15 = 30$$

  • The catenary constant

  • The lowest point is

    $$x_c = 50 + \frac{400}{2} + 300 \sinh^{-1} \left( \frac{30 \exp(400/(2\,300)}{300 (1-\exp(400/300))} \right) = 50 + 200 - 20.89897 = 229.101028 $$

    $$y_c = 15 + 300 (1-1.183563) = -40.06883 $$

  • The shape for this example is

    $$ y=15+300 \left( \cosh \left( \frac{x}{300}-0.7636700940207649 \right)-\cosh \left(-0.5970034273540983\right) \right) $$

    fig2

You can delve into C# code to handle catenaries in my GitHub - CatenaryCalculator.cs

/// <summary>
/// Calculate the lower point on the catenary (center point)
/// </summary>
/// <param name="span">The span</param>
/// <param name="w">The unit weight</param>
/// <param name="H">The horizontal tension</param>
/// <returns>A position vector in (x,y) coordinates</returns>
public static Vector2 CenterPosition(Vector2 span, double w, double H)
{
    double a=H/w;
    double η=span.X/(2*a);
double xc=span.X/2+a*DoubleEx.Asinh((span.Y*Math.Exp(η))/(a*(1-Math.Exp(2*η))));
double yc=-a*(Math.Cosh(-xc/a)-1);

return new Vector2(xc, yc);

}

A Desmos implementation of the above linked. Change the parameter $t=\frac{S}{2a}$ to get the length required.

John Alexiou
  • 13,816
  • Do these formulas hold if I construct an example where the lowest point of the catenary occurs at an x value less than $x_1$ and less than $x_2$? – Simon M Sep 14 '23 at 01:03
  • Yes. By increasing the catenary constant, for example, it will move the lowest point to the left past the support to cause tower uplift. This happens when $x_c < x_1$. Do you want to see the rest of the equations describing the length, tension, sag for an uneven span? – John Alexiou Sep 14 '23 at 12:00
  • I do, but now I'm a bit confused. In the problem that I originally described, we're dealing with an abstract, platonic, mathematically ideal string. It's my understanding that no physics variables ought to be involved. In fact, the only variables that should have any impact on the equation we end up with are the 2 chosen points as well as the length of catenary between the two points. If you first choose the 2 points and then you vary the length of the string, the length will determine where $x_c$ ends up with respect to $x_1$ and $x_2$. That's why my answer needs only the points and length L – Simon M Sep 14 '23 at 15:28
  • The catenary constant $a$ is a purely geometrical constant with units of length. If you know the total length $L$ you need to find the corresponding catenary constant $a$, but there is no analytical solution for this. For the flat case this is $$ L = 2 a \sinh \left( \frac{S}{2 a} \right) $$ – John Alexiou Sep 14 '23 at 15:31
  • Oh, ok. That makes sense. What I'm still wondering is why my answer involves so few variables. My answer must be wrong, then? I've provided the equation solved for $y$ for each of the 5 possible cases. You could then take that equation and derive the corresponding equation for the arc length from $x_1$ to $x_2$. Then, set that equal to $L$ to obtain an equation that relates the length $L$ to the free variable that I've called $c$. Solving for $c$ given $L$ would not be analytical, but it would be all you need to do in order to obtain the right equation for your catenary – Simon M Sep 14 '23 at 15:42
  • My original hope when beginning this problem was that closed form equations for $a$, $b$, and $c$ in terms of length $L$ could be obtained. I see that that's not possible. Nonetheless, I still firmly believe that after you choose your two points, the only other thing you need to choose in order to determine a unique catenary is the length $L$. The unit weight of the cable and the horizontal tension would not be appropriate to take into consideration. – Simon M Sep 14 '23 at 16:02
  • @SimonM after you have the endpoints, what you need to do is define the constant $a$ for a unique catenary. Remember that $a$ has a geometric meaning, and it is related to the directrix of the curve. The length, curve, slope etc are all a function of $a$. – John Alexiou Sep 14 '23 at 16:04
  • I prefer to think of defining the length $L$, and then $a$ is a function of $L$ – Simon M Sep 14 '23 at 16:06
  • I think in order to clear up my confusion I would like to know what parts of my answer are wrong, if any – Simon M Sep 14 '23 at 16:09
  • @SimonM - there is no function of $a$ in terms of $L$. Only a function of $L$ in terms of $a$. That is the nature of the problem. I have worked with the sag of transmission lines for many years and it always involves numerical methods to find a solution. – John Alexiou Sep 14 '23 at 16:10
  • I misspoke. What I mean to say is that $a$ can be found in terms of $L$. – Simon M Sep 14 '23 at 16:11
  • After I have chosen the endpoints, would it suffice to simply choose a length $L$ in order to determine a unique catenary? Instead of choosing a value for the catenary constant? – Simon M Sep 14 '23 at 16:40
  • @SimonM - as you know even the end points are not needed. To parameterize a catenary you need three values. The (x,y) coordinates of the lowest point and the catenary constant. If you are asking if you can derive these three parameters from the end points and the length, yes you can as an approximation. For exact values you need the end points and the constant. – John Alexiou Sep 14 '23 at 18:52
  • There is another reason not to pick the length $L$ as a parameter because the allowed values are limited and it needs to be specified very precisely. The range is $\sqrt{S^2+h^2} < L < \infty$ but only values near the lower limit make much sense. On the other hand, picking $a$, or in fact picking $\eta = \frac{S}{2 a}$ as a parameter is much better because it is dimensionless, it range is all positive real numbers and the smaller the number near zero the more straight the catenary is. – John Alexiou Sep 15 '23 at 01:14
  • Look at this Demsos solver where you manipulate one parameter (called $t$ here) to set the length. Without loss of generality, you can put the coordinate origin at the left support and specify only $S$ and $h$ the span length and height difference. – John Alexiou Sep 15 '23 at 01:19
  • 1
    I've saved a modified copy of the desmos solver you linked, where instead of fixing the left endpoint to the origin you can choose each of $x_1$, $x_2$, $y_1$ and $y_2$, keeping in mind that we assume $x_2>x_1$. Here it is https://www.desmos.com/calculator/pd3pgnerks – Simon M Sep 15 '23 at 17:29
0

WLOG, assume $x_2>x_1$.

Let $L$ be the length of the catenary between $(x_1,y_1)$ and $(x_2,y_2)$

Your particular choices of $x_1$, $x_2$, $y_1$, $y_2$ and $L$ will determine which of the following 5 cases applies to you and your catenary.

Case 1:

$y_1<y_2$ and $\sqrt{(x_2-x_1)^2+(y_2-y_1)^2}<L<k\sinh(\frac{x_2-x_1}{k})$, such that $k$ is the unique solution to $$0=x\cosh(\frac{x_2-x_1}{x})+y_1-x-y_2$$

The length $L$ is not enough so that the low point must occur inside of the interval from $x_1$ to $x_2$, and $y_1<y_2$

These two conditions mean that the low point of the catenary occurs at an $x$ value less than $x_1$

Define:

$$g(x)=x_1-x\cosh^{-1}(\frac{y_1-c}{x})$$

$$f(x)=x_1-x_2+x(\cosh^{-1}(\frac{y_2-c}{x})-\cosh^{-1}(\frac{y_1-c}{x}))$$

Then:

$$a=f^{-1}(0)$$

$$b=g(a)$$

Then, define:

$$\alpha(c)=f^{-1}(0)$$

$$\beta(c)=g(\alpha(c))$$

Then an upward-opening catenary through the desired points can be generated by choosing a value of c for the function

$$y=\alpha(c)\cosh(\frac{x-\beta(c)}{\alpha(c)})+c$$


Case 2:

$y_1<y_2$ and $L=k\sinh(\frac{x_2-x_1}{k})$, such that $k$ is the unique solution to $$0=x\cosh(\frac{x_2-x_1}{x})+y_1-x-y_2$$

The length L is exactly enough so that the low point of the catenary must occur at either $x_1$ or $x_2$

These two conditions mean that the low point is at $(x_1,y_1)$

The upward-opening catenary through the desired points can be generated by the function

$$y=k\cosh(\frac{x-x_1}{k})+y_1-k$$

where $k$ is the unique solution to

$$0=x\cosh(\frac{x_2-x_1}{x})+y_1-x-y_2$$


Case 3:

$L>k\sinh(\frac{x_2-x_1}{k})$, such that $k$ is the unique solution to $$0=x\cosh(\frac{x_2-x_1}{x})+y_1-x-y_2$$

The length $L$ is enough so that the low point must occur inside the interval from $x_1$ to $x_2$

These two conditions mean that the low point of the catenary occurs at an $x$ value that satisfies $x_1<x<x_2$

Define:

$$g(x)=x_1+x\cosh^{-1}(\frac{y_1-c}{x})$$

$$f(x)=x_1-x_2+x(\cosh^{-1}(\frac{y_2-c}{x})+\cosh^{-1}(\frac{y_1-c}{x}))$$

Then:

$a$ is either zero of $f$ and

$$b=g(a)$$

Then, define:

$\alpha(c)$ is either zero of $f$ and

$$\beta(c)=g(\alpha(c))$$

Then an upward-opening catenary through the desired points can be generated by choosing a value of c for the function

$$y=\alpha(c)\cosh(\frac{x-\beta(c)}{\alpha(c)})+c$$


Case 4:

$y_1>y_2$ and $L=k\sinh(\frac{x_2-x_1}{k})$, such that $k$ is the unique solution to $$0=x\cosh(\frac{x_2-x_1}{x})+y_2-x-y_1$$

The length L is exactly enough so that the low point of the catenary must occur at either $x_1$ or $x_2$

These two conditions mean that the low point is at $(x_2,y_2)$

The upward-opening catenary through the desired points can be generated by the function

$$y=k\cosh(\frac{x-x_2}{k})+y_2-k$$

where $k$ is the unique solution to

$$0=x\cosh(\frac{x_1-x_2}{x})+y_2-x-y_1$$


Case 5:

$y_1>y_2$ and $\sqrt{(x_2-x_1)^2+(y_2-y_1)^2}<L<k\sinh(\frac{x_2-x_1}{k})$, such that $k$ is the unique solution to $$0=x\cosh(\frac{x_2-x_1}{x})+y_1-x-y_2$$

The length $L$ is not enough so that the low point must occur inside of the interval from $x_1$ to $x_2$, and $y_1>y_2$

These two conditions mean that the low point of the catenary occurs at an $x$ value greater than $x_2$

Define:

$$g(x)=x_2+x\cosh^{-1}(\frac{y_2-c}{x})$$

$$f(x)=x_1-x_2+x(\cosh^{-1}(\frac{y_1-c}{x})-\cosh^{-1}(\frac{y_2-c}{x}))$$

Then:

$$a=f^{-1}(0)$$

$$b=g(a)$$

Then, define:

$$\alpha(c)=f^{-1}(0)$$

$$\beta(c)=g(\alpha(c))$$

Then an upward-opening catenary through the desired points can be generated by choosing a value of c for the function

$$y=\alpha(c)\cosh(\frac{x-\beta(c)}{\alpha(c)})+c$$


Simon M
  • 657
  • How do you know this answer is correct? have you tried an example? – John Alexiou Sep 14 '23 at 18:53
  • Yes. I have provided 5 examples in desmos with the 5 links for the possible types of cases. What I did for these examples is manually find the x-intercept(s) of $f$ and then apply my formulas – Simon M Sep 14 '23 at 19:35
  • @JohnAlexiou [1]: https://www.desmos.com/calculator/xsvg0tatvd [2]: https://www.desmos.com/calculator/odfgmngdww [3]: https://www.desmos.com/calculator/hsq8eeul0q [4]: https://www.desmos.com/calculator/nzglfzjp0o [5]: https://www.desmos.com/calculator/epfhpmrjkb – Simon M Sep 14 '23 at 19:42
  • I see 6 inputs, plus an eyeballing of $a$ to get it to work. You wanted a solution with 5 inputs (two point coordinates + length) and I gave you one with 5 inputs. Plus the complexity of different cases is not needed. One solution works for all cases. – John Alexiou Sep 15 '23 at 00:51
  • @JohnAlexiou The 5 inputs are $x_1$, $x_2$, $y_1$, $y_2$ and $c$, where $c$ is a quantity related to the length $L$. The $k$ you see is actually a leftover bit unrelated to everything else. I had simply forgotten to delete it. Here is the updated graph https://www.desmos.com/calculator/8njkvoicuy – Simon M Sep 15 '23 at 16:38
0

WLOG, assume $x_2>x_1$.

Let $L$ be the length of the catenary between $(x_1,y_1)$ and $(x_2,y_2)$

Define:

$$S=x_2-x_1$$

$$h=y_2-y_1$$

$$A(t)=\frac{S}{2t}$$

$$B(t)=\frac{S}{2}+A(t)\sinh^{-1}\left(\frac{he^{t}}{A(t)(1-e^{2t})}\right)=\frac{S}{2}+A(t)\sinh^{-1}\left(\frac{-h}{2A(t)\sinh\left(t\right)}\right)$$

Such that $t$ is the solution to $$L=2A(t)\sinh\left(t\right)\sqrt{1+\left(\frac{he^{t}}{A(t)\left(1-e^{2t}\right)}\right)^{2}}=A(t)\sinh\left(\frac{S-B(t)}{A(t)}\right)+A(t)\sinh\left(\frac{B(t)}{A(t)}\right)$$

Then the equation for the catenary is

$$ \boxed{ y=A(t)\left(\cosh\left(\frac{x-B(t)-x_{1}}{A(t)}\right)-\cosh\left(\frac{B(t)}{A(t)}\right)\right)+y_{1} }$$

Let k be the solution to $0=x\cosh\left(\frac{x_{1}-x_{2}}{x}\right)-|y_{2}-y_{1}|-x$

If $L>k\sinh\left(\frac{x_{2}-x_{1}}{k}\right)$ then the lowest point of the catenary will occur between $x_1$ and $x_2$

If $L=k\sinh\left(\frac{x_{2}-x_{1}}{k}\right)$ then the lowest point of the catenary will occur at whichever is lower of the two endpoints

If $L<k\sinh\left(\frac{x_{2}-x_{1}}{k}\right)$ then the lowest point of the catenary will occur outside of the interval from $x_1$ to $x_2$

Simon M
  • 657