Analyze the flow of a fluid during injection
into a long vertical channel, assuming that the flow
is modeled by the boundary value problem
We use a
-stage collocation method to formulate
this problem as an optimization problem with a constant merit
function and equality constraints representing the solution
of .
We use a uniform partition with
subintervals of
, and the standard [2, pages 247-249]
basis representation,
The constraints in the optimization problem are
the initial conditions in , the continuity conditions,
and the collocation equations.
There are
initial conditions.
The continuity equations
are a set of
linear equations.
The collocation equations are a set of
nonlinear equations obtained by requiring that
satisfy at the
collocation points
for
and
.
The collocation points
are the
roots of the
th degree Legendre polynomial.
The parameters in the optimization problem are the
parameters
and
in the representation of
.
Data for this problem appears in Table 7.1.
| Variables | |
| Constraints | |
| Bounds | 0 |
| Linear equality constraints | |
| Linear inequality constraints | 0 |
| Nonlinear equality constraints | |
| Nonlinear inequality constraints | 0 |
|
Nonzeros in
|
0 |
|
Nonzeros in |
|
Results for the AMPL implementation with
and
are summarized in Table 7.2.
The starting point is the function
evaluated at the mesh points.
Solutions for several
are shown in
Figure 7.1.
This problem is easy
to solve for small Reynolds numbers but becomes increasingly difficult
to solve as
increases.
| Solver | |
|
|
|
| LANCELOT | |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| iterations | |
|
|
|
| LOQO | 1.55 s | 2.59 s | 7.03 s | 22.54 s |
|
|
1.00000e+00 | 1.00000e+00 | 1.00000e+00 | 1.00000e+00 |
|
|
5.1e-12 | 1.1e-11 | 2.9e-11 | 1.9e-11 |
| iterations | 32 | 25 | 29 | 35 |
| MINOS | 1.09 s | 3.25 s | 11.15 s | 32.65 s |
|
|
1.00000e+00 | 1.00000e+00 | 1.00000e+00 | 1.00000e+00 |
|
|
3.8e-13 | 2.4e-13 | 2.0e-13 | 3.8e-07 |
| iterations | 5 | 5 | 5 | 5 |
| SNOPT | 2.14 s | 7.47 s | 27.91 s | 98.5 s |
|
|
1.00000e+00 | 1.00000e+00 | 1.00000e+00 | 1.00000e+00 |
|
|
6.1e-05 | 4.6e-05 | 4.1e-05 | 3.9e-05 |
| iterations | 3 | 3 | 3 | 3 |
|
|
||||
LANCELOT is unable to solve even simple versions of the problem, advancing very slowly toward the solution (as judged from the value of the merit function).