3
$\begingroup$

Suppose that we have a sufficiently thin, flexible cylindrical rod of length $L$ made from a homogeneous, isotropic material, and that initially [at rest?] the central axis of the rod is a straight line segment.

We select $n$ points, $\mathbf x_1,\ldots,\mathbf x_n$ so that the sum of the distances between consecutive points is no greater than $L$ - i.e.

$$\sum_{i=1}^{n-1}\Vert\mathbf x_i-\mathbf x_{i+1}\Vert\le L$$

Assuming that energy (due to tension, stress, and such) is minimized, if the rod is deformed so that its cenral axis intersects each of the points $\mathbf x_1,\ldots,\mathbf x_n$ (and neither endpoint is located at one of the points, probably), what is the curve formed by the central axis of the rod?


This is not a homework question, it's for a gardening project. I just figure that, since structural engineering and differential calculus have simultaneously existed for at least two centuries, this exact problem - or a nearly identical problem - has probably been extensively studied already.

$\endgroup$
4
  • $\begingroup$ For search purposes, this is also known as the open spline or elastica. $\endgroup$ Commented May 19, 2022 at 18:01
  • $\begingroup$ You may need to add a further constraint of minimizing length. If I wrap the rod around the world the answer is different from if I have only 4 meters of rod. Further, if you have experience with spline fits, the polynomial order and even specific point placements have a considerable effect on the 'best' spline fit. $\endgroup$
    – Jon Custer
    Commented May 19, 2022 at 18:09
  • $\begingroup$ The answer is mathematically complicated. One is essentially trying to minimize integrated energy $E=\int \kappa^2\,ds$ where $\kappa$ is curvature and $s$ is arc-length, while holding the total length $L=\int ds$ fixed, and subject to the conditions which you have stated. In 2-dimensions, the exact curve will essentially be given by a piecewise collection of complicated functions involving elliptic integrals. See this reference. In 3-dimensions, I'm sure the answer will be very complicated. $\endgroup$ Commented May 19, 2022 at 19:52
  • $\begingroup$ @ArturodonJuan For 3-dimensions, maybe we can try averaging 2-dimensional solutions over projections to the tangent planes of a sphere containing the system? I mean, it won't account for torsion, but it might be a start. $\endgroup$
    – R. Burton
    Commented May 20, 2022 at 1:11

1 Answer 1

0
$\begingroup$

The shape is that of a cubic spline. Each segment between two points is a 3rd order polynomial with matching deflection and slope at the nodes.

Between nodes $i$ and $i+1$ with distance $h_i = {\bf x}_{i+1} - {\bf x}_i$ the shape is described by a parameter $t=0\ldots 1$

$$ y(t) = (1-3 t^2+2t^3) {\bf y}_i + t^2 (3-2 t) {\bf y}_{i+1} + h_i \left( t (1-t)^2 {\bf y'}_i + t^2 (t-1) {\bf y'}_{i+1}\right) $$

The nodal displacements are zero ${\bf y}_{i}=0$ and ${\bf y}_{i+1}=0$, and the nodal slopes are unknown ${\bf y'}_i$ and ${\bf y'}_{i+1}$.

To actually deform the rod, you need some kind of loading, and since the nodes are fixed, you need to apply torques at each node to deform it. The torque might be zero in the middle and some value at one or both ends, or it might be defined at every node.

For nodes $i=1\ldots n-1$ the nodal torque $\tau_i$ relates to the slopes with

$$ \frac{\tau_i}{E I} = \frac{2}{h_i^2} \left( 3 ( {\bf y}_{i+1} - {\bf y}_i) - h_i ( {\bf y'}_{i+1} - 2{\bf y'}_i ) \right) $$

for the last segment, the torque is

$$ \frac{\tau_n}{E I} = \frac{2}{h_{n-1}^2} \left( 3 ( {\bf y}_{n-1} - {\bf y}_n) + h_{n-1} ( {\bf y'}_{n-1} + 2{\bf y'}_n ) \right) $$

Where $E\,I$ is the flexural rigidity of the rod.

The above is a system of $n$ equations to be solved for the $n$ unknown slopes. Note that typical static equilibrium requires the net torque to be zero $\sum_i M_i = 0$

$\endgroup$

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