6
$\begingroup$

I have some 3D parametric curve par and some line line.

I want to find parameters u,v and t for the intersection points of par and line.

I tried to do that with NSolve in many different ways and always got the following error message:

"This system cannot be solved with the methods available to NSolve."

Below is just an example, but I'm looking for a method, that works for every parametric curve and every line.

Also I don't want to get x,y and z but u,v and t

How to fix it?

Clear[u, v, t];
par = 5 {Cos[u], Cos[v] + Sin[u], Sin[v]};
line = {1, 2, 3} + {4, -5, 6} t;
NSolve[
 par == line
  && 0 < u < 2 \[Pi]
  && -\[Pi] < v < \[Pi],
 {t, u, v}, Reals]
$\endgroup$
  • $\begingroup$ par = 5 {Cos[u], Cos[v] + Sin[u], Sin[v]}; line = {1, 2, 3} + {4, -5, 6} t; Solve[par == line && 0 < u < 2 \[Pi] && -\[Pi] < v < \[Pi], {t, u, v}, Reals, Method -> Reduce] $\endgroup$ – cvgmt Aug 15 '20 at 2:58
5
$\begingroup$

In general, Reduce is the most powerful Mathematica function for solving equations.

Reduce[par == line && 0 < u < 2 Pi && -Pi < v < Pi, {t, u, v}, Reals]
(* (t == AlgebraicNumber[Root[-18291750000 + 235480000*#1 - 237800*#1^2 + 
        48*#1^3 + #1^4 & , 1, 0], {0, 1/725, 0, 0}] && 
    u == 2*ArcTan[AlgebraicNumber[Root[-18291750000 + 235480000*#1 - 
        237800*#1^2 + 48*#1^3 + #1^4 & , 1, 0], 
        {3961/4536, -2479/1315440, 1121/353220000, 83/19073880000}]] && 
    v == 2*ArcTan[AlgebraicNumber[Root[-18291750000 + 235480000*#1 - 
        237800*#1^2 + 48*#1^3 + #1^4 & , 1, 0], {3505/1484, -9221/6455400, 
        17053/9360330000, 197/18720660000}]]) || 
    (t == AlgebraicNumber[Root[-18291750000 + 235480000*#1 - 237800*#1^2 + 
        48*#1^3 + #1^4 & , 2, 0], {0, 1/725, 0, 0}] && 
     u == 2*ArcTan[AlgebraicNumber[Root[-18291750000 + 235480000*#1 - 
        237800*#1^2 + 48*#1^3 + #1^4 & , 2, 0], 
        {3961/4536, -2479/1315440, 1121/353220000, 83/19073880000}]] && 
     v == 2*ArcTan[AlgebraicNumber[Root[-18291750000 + 235480000*#1 - 
         237800*#1^2 + 48*#1^3 + #1^4 & , 2, 0], 
         {3505/1484, -9221/6455400, 17053/9360330000, 197/18720660000}]]) @)

% // N
(* (t == -1.07503 && u == 2.29164 && v == -0.761533) || 
   (t == 0.116633 && u == 1.27311 && v == 2.30858) *)

The intersections can be visualized by

Show[
    ParametricPlot3D[par, {u, 0, 2 Pi}, {v, -Pi, Pi}, 
        PlotStyle -> Opacity[.5], LabelStyle -> {15, Bold, Black}],
    ParametricPlot3D[line, {t, -2, 1}, PlotStyle -> {Black, Thick}],
    ListPointPlot3D[{line /. t -> -1.07503, line /. t -> 0.11663}, PlotStyle -> Red]]

enter image description here

Addendum: Use FindRoot

In the event that Reduce does not provide a solution, FindRoot almost always will, but requires multiple initial guesses to obtain multiple intersections, as is the case here.

FindRoot[par == line, {t, 0}, {u, Pi}, {v, 0}]
FindRoot[par == line, {t, 0}, {u, Pi}, {v, 2}]
(* {t -> -1.07503, u -> 2.29164, v -> -0.761533} *)
(* {t -> 0.116633, u -> 1.27311, v -> 2.30858} *)
$\endgroup$
  • 1
    $\begingroup$ +1 You can tell Solve to use Reduce, e.g., Solve[par == line && 0 < u < 2 Pi && -Pi < v < Pi, {t, u, v}, Reals, Method -> Reduce] $\endgroup$ – Bob Hanlon Jul 27 '20 at 0:57
4
$\begingroup$

In general, first try Reduce as already mentioned. But some more difficult curves will not reduce or Mathematica will hang. For example:

par = {1, Erf[t], Cos[t]};
line = {1, 1/2, 0} + {0, 1, 1} s;
Reduce[par == line && 0 < s < 2 && 0 < t < 2, {s, t}]

Instead you can try minimizing the distance between the line and the curve:

result = NMinimize[{SquaredEuclideanDistance[par, line], 0 < s < 2, 0 < t < 2}, {s, t}]
(* result: {4.42622*10^-19, {s -> 0.399155, t -> 1.1602}} *)
{par,line} /. Last[result]

(* {{1, 0.899155, 0.399155}, {1, 0.899155, 0.399155}} *)

For your example of a line surface intersection NMinimize cam get trapped in a local minimum with the default method:

par = 5 {Cos[u], Cos[v] + Sin[u], Sin[v]};
line = {1, 2, 3} + {4, -5, 6} t;
NMinimize[{SquaredEuclideanDistance[par, line], 0 < u < 2 Pi, -Pi < v < Pi}, {t, u, v}]
(* result: {2.97068, {t -> 0.39428, u -> 5.45283, v -> 0.964654}} *)

Notice the distance 2.97 is too high. Try a different method to get a better result:

NMinimize[{SquaredEuclideanDistance[par, line], 0 < u < 2 Pi, -Pi < v < Pi},
 {t, u, v}, Method -> "DifferentialEvolution"]
(* {1.77932*10^-16, {t -> -1.07503, u -> 2.29164, v -> -0.761533}} *)
$\endgroup$
  • $\begingroup$ With my version 8.0 this works not good with standard method "NelderMead" for NMinimize. Have to choose other methods and in some cases higher WorkingPrecision. Your shown result isn't correct either. $\endgroup$ – Akku14 Jul 27 '20 at 7:53
  • $\begingroup$ I copied the wrong result - I mentioned this in a comment earlier that it would get stuck in a local minimum but I must have forgotten to put it here. I'll edit it. $\endgroup$ – flinty Jul 27 '20 at 11:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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