Transcript Linear Programming - Georgia Institute of Technology

Linear Programming
Why talk about Linear Programming (LP)?
•
LP is simpler than NLP, hence, good for a foundation
•
linearity has some unique features for optimization
•
a lot of problems are or can be converted to a LP
formulation
•
some NLP algorithms are based upon LP simplex
method
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
57
Bolted Joint Design
Given
At - tensile strength diameter, function of d
Db - bolt circle diameter
Pt - total load
Question:
C - joint constant
Is this a linear or
Fi - preload (= 0.75 Sp At)
nonlinear model?
Find
N - number of bolts, Sp - proof strength, d - diameter
Satisfy
3d   Db / N
good wrench rule
 Db / N  6d
good seal rule
C Pt / N  Sp At - Fi
static loading constraint
Fi  (1 - C) Pt / N
joint separation constraint
Minimize Z= [ f1(N, d, Sp), f2(N, d, Sp), ..]
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
58
Bolted Joint Design (2)
Given
d - diameter
At - tensile strength diameter, function of d
Db - bolt circle diameter
Pt - total load
Question:
C - joint constant
Is this a linear or
nonlinear problem?
Fi - preload (= 0.75 Sp At)
Find
N - number of bolts, Sp - proof strength
Satisfy
3d   Db / N
 Db / N  6d
C Pt / N  Sp At - Fi
Fi  (1 - C) Pt / N
good wrench rule
good seal rule
static loading constraint
joint separation constraint
Minimize Z= [ f1(N, Sp), f2(N, Sp), ..]
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
59
Historical Perspective
1928 – John von Neumann published related central theorem of game
theory
1944 – Von Neumann and Morgenstern published Theory of Games and
Economic Behavior
1936 – W.W. Leontief published "Quantitative Input and Output
Relations in the Economic Systems of the US" which was a linear
model without objective function.
1939 – Kantoravich (Russia) actually formulated and solved a LP
problem
1941 – Hitchcock poses transportation problem (special LP)
WWII – Allied forces formulate and solve several LP problems related to
military
A breakthrough occurred in 1947...
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
60
SCOOP
•
•
•
US Air Force wanted to investigate the feasibility of applying
mathematical techniques to military budgeting and planning.
George Dantzig had proposed that interrelations between activities of
a large organization can be viewed as a LP model and that the optimal
program (solution) can be obtained by minimizing a (single) linear
objective function.
Air Force initiated project SCOOP (Scientific Computing of Optimum
Programs)
NOTE:
SCOOP began in June 1947 and at the end of the same summer, Dantzig
and associates had developed:
1) an initial mathematical model of the general linear programming
problem
2) a general method of solution called the simplex method
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
61
Simplex Today
•
Other (polynomial time) algorithms have been developed
for solving LP problems:
• Khachian algorithm (1979)
• Kamarkar algorithm (AT&T Bell Labs, mid 80s)
BUT,
none of these algorithms have been able to beat Simplex in
actual practical applications
HENCE,
Simplex is and will most likely remain the most dominant
LP algorithm for at least the near future
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
62
Fundamental Theorem
Extreme point (or Simplex filter) theorem:
If the maximum or minimum value of a
linear function defined over a polygonal
convex region exists, then it is to be found
at the boundary of the region.
Convex set:
A set (or region) is convex if, for any two points (say, x1
and x2) in that set, the line segment joining these points
lies entirely within the set.
A point is by definition convex.
See also pages 664 and 666 in your book.
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
63
What does the extreme point theorem imply?
•
A finite number of extreme points implies a finite number of
solutions
•
Hence, search is reduced to a finite set of points
•
However, a finite set can still be too large for practical purposes
•
Simplex method provides an efficient systematic search
guaranteed to converge in a finite number of steps.
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
64
Basic Steps of Simplex
1. Begin the search at an extreme point (i.e., a basic feasible
solution).
2. Determine if the movement to an adjacent extreme can improve on
the optimization of the objective function. If not, the current solution
is optimal. If, however, improvement is possible, then proceed to
the next step.
3. Move to the adjacent extreme point which offers (or, perhaps,
appears to offer) the most improvement in the objective function.
4. Continue steps 2 and 3 until the optimal solution is found or it can
be shown that the problem is either unbounded or infeasible.
(see also page 152 in your book)
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
65
Step 0 – Obtain Canonical Form
IMPORTANT: Simplex only deals with equalities
General Simplex LP model:
min (or max) z =  ci xi
s.t.
Ax=b
x0
In order to get and maintain this form, use
• slack, x  b -> x + slack = b (also called negative deviation)
• surplus, x  b -> x - surplus = b (also called positive deviation)
• artificial variables (sometimes need to be added to ensure all
variables  0)
(see also Section 4.3)
Compare constraint conversion with goal
conversions using deviation variables
Georgia Institute of Technology
Optimization in Engineering Design
Systems Realization Laboratory
66
Different "components" of a LP model
LP model can always be split into a basic and a non-basic part.
"Transformed" or" reduced" model is a good way to show this.
This can be represented in mathematical terms as well as in a LP
or simplex tableau.
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
67
Movement to Adjacent Extreme Point
Given any basis we move to an adjacent extreme point
(another basic feasible solution) of the solution space by
exchanging one of the columns that is in the basis for a
column that is not in the basis.
Two things to determine:
1) which (nonbasic) column of A should be brought into
the basis so that the solution improves?
2) which column can be removed from the basis such that
the solution stays feasible?
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
68
Entering and Departing Vector (Variable) Rules
General rules:
•
The one nonbasic variable to come in is the one which
provides the highest reduction in the objective function.
•
The one basic variable to leave is the one which is
expected to go infeasible first.
NOTE: THESE ARE HEURISTICS!!
Variations on these rules exist, but are rare.
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
69
Simplex Variations
Various variations on the simplex method exist (see
handouts):
•
"regular" simplex (see Example 4.6)
•
two-phase method: Phase I for feasibility and Phase II for
optimality (see Section 4.4.5)
•
condensed/reduced method: only use the nonbasic columns
to work with (see handout)
•
(revised) dual simplex (see Section 4.8), etc.
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
70
Computational Considerations
•
Unrestricted variables (unboundedness)
•
Redundancy (linear dependency, modeling errors)
•
Degeneracy (some basic variables = 0)
•
Round-off errors
(see also page 162-163)
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
71
Limitations of Simplex
•
•
Inability to deal with multiple objectives
Inability to handle problems with integer variables
Latter has resulted in:
•
•
Cutting plane algorithms (Gomory, 1958)
Branch and Bound (Land and Doig, 1960)
However,
solution methods to LP problems with integer or Boolean
variables are still far less efficient than those which include
continuous variables only
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
72
Example Problem
Maximize Z = 5x1 + 2x2 + x3
subject to
x1 + 3x2 - x3 ≤ 6,
x2 + x3 ≤ 4,
3x1 + x2
≤ 7,
x1, x2, x3 ≥ 0.
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
73
Simplex and Example Problem
Step 1. Convert to Standard Form
a11 x1 + a12 x2 + ••• + a1n xn ≤ b1,
a11 x1 + a12 x2 + ••• + a1n xn + xn+1 = b1,
a21 x1 + a22 x2 + ••• + a2n xn ≥ b2,
a21 x1 + a22 x2 + ••• + a2n xn - xn+2 = b2,

am1 x1 + am2 x2 + ••• + amn xn ≤ bm,
am1 x1 + am2 x2 + ••• + amn xn + xn+k = bm,
In our example problem:
x1 + 3x2 - x3 ≤ 6,
x2 + x3 ≤ 4,
3x1 + x2
≤ 7,
x1, x2, x3 ≥ 0.
Optimization in Engineering Design
x1 + 3x2 - x3 + x4
x2 + x3
3x1 + x2
= 6,
+ x5
= 4,
+ x6 = 7,
x1, x2, x3, x4, x5, x6 ≥ 0.
Georgia Institute of Technology
Systems Realization Laboratory
74
Simplex: Step 2
Step 2. Start with an initial basic feasible solution (b.f.s.) and set up the
initial tableau.
In our example
Maximize Z = 5x1 + 2x2 + x3
x1 + 3x2 - x3 + x4
x2 + x3
3x1 + x2
= 6,
+ x5
= 4,
+ x6 = 7,
x1, x2, x3, x4, x5, x6 ≥ 0.
cB
0
0
0
Optimization in Engineering Design
Basis
x4
x5
x6
c
row
cj
5 2 1 0
x1 x2 x3 x4
1 3 -1 1
0 1 1 0
3 1 0 0
5 2 1 0
Constants
0 0
x5 x6
0 0
1 0
0 1
0 0
6
4
7
Z=0
Georgia Institute of Technology
Systems Realization Laboratory
75
Step 2: Explanation
Adjacent Basic Feasible Solution
If we bring a nonbasic variable xs into the basis, our system changes from
the basis, xb, to the following (same notation as the book):
x1
+ ā1sxs= b1
x = b a
for i =1, …, m
i
xr
+ ārsxr= b
r

i
is
xs = 1
xj = 0
for j=m+1, ..., n and js
xm + āmsxs= b
s

The new value of the objective function becomes:
m
Z
 c (b  a
i
i
is )  c s
i 1
Thus the change in the value of Z per unit increase in xs is
c s = new
value of Z - old mvalue of Z
m
=
 c (b  a
c  c a
i
i 1
=
i
m
s
i is
i 1
Optimization in Engineering Design
is )  c s

c b
i i
i 1
This is the Inner Product rule
Georgia Institute of Technology
Systems Realization Laboratory
76
Simplex: Step 3
Use the inner product rule to find the relative profit coefficients
cB
0
0
0
Basis
x4
x5
x6
c
row
cj
5 2 1 0
x1 x2 x3 x4
1 3 -1 1
0 1 1 0
3 1 0 0
5 2 1 0
Constants
0 0
x5 x6
0 0
1 0
0 1
0 0
6
4
7
Z=0
c j  c j  cB Pj
c1 = 5 - 0(1) - 0(0) - 0(3) = 5 -> largest positive
c2 = ….
c3 = ….
Step 4: Is this an optimal basic feasible solution?
Optimization in Engineering Design
Georgia Institute of Technology
Systems Realization Laboratory
77
Simplex: Step 5
Apply the minimum ratio rule to determine the basic variable to leave the basis.
The new values of the basis variables:
xi = bi
 a is x s
for i = 1, ..., m
 bi 
max x s  min  
a is 0 a is
 
In our example:
cB
0
0
0
Basis
x4
x5
x6
c
row
cj
5 2 1 0
x1 x2 x3 x4
1 3 -1 1
0 1 1 0
3 1 0 0
5 2 1 0
Optimization in Engineering Design
Constants
0 0
x5 x6
0 0
1 0
0 1
0 0
6
4
7
Z=0
Row
1
2
3
Basic Variable
x4
x5
x6
Ratio
6
7/3
Georgia Institute of Technology
Systems Realization Laboratory
78
Simplex: Step 6
Perform the pivot operation to get the new tableau and the b.f.s.
cB
0
0
0
Basis
x4
x5
x6
c
row
cj
5 2 1 0
x1 x2 x3 x4
1 3 -1 1
0 1 1 0
3 1 0 0
5 2 1 0
Constants
0 0
x5 x6
0 0
1 0
0 1
0 0
New iteration:
find entering
variable:
c j  c j  c B Pj
cB = (0 0 5)
c2 = 2 - (0) 8/3 - (0) 1 - (5) 1/3 = 1/3
c3 = 1 - (0) (-1) - (0) 1 - (5) 0 = 1
c6 = 0 - (0) 0 - (0) 0 - (5) 1/3 = -5/3
Optimization in Engineering Design
6
4
7
Z=0
cB
0
0
5
c
Basis
x4
x5
x1
row
cj
5
2
1
x1 x2 x3
0 8/3 -1
0
1
1
1 1/3 0
0 1/3 1
Constants
0
x4
1
0
0
0
0
x5
0
1
0
0
0
x6
0
0
1/3
-5/3
11/3
4
7/3
Z=35/3
Georgia Institute of Technology
Systems Realization Laboratory
79
Final Tableau
cB
0
0
5
c
Basis
x4
x5
x1
row
cj
5
2
1
x1 x2 x3
0 8/3 -1
0
1
1
1 1/3 0
0 1/3 1
Wrong value!
4 should be 11/3
0
x4
1
0
0
0
cB
0
1
5
c
Optimization in Engineering Design
Constants
0
x5
0
1
0
0
0
x6
0
0
1/3
-5/3
11/3
4
7/3
Z=35/3
Basis
x4
x3
x1
row
x3 enters basis,
x5 leaves basis
cj
5
2
x1 x2
0
4
0
1
1 1/3
0 -2/3
1
x3
0
1
0
0
Constants
0
x4
1
0
0
0
0
0
x5 x6
1
0
1
0
0 1/3
-1 -5/3
23/3
4
7/3
Z=47/3
Georgia Institute of Technology
Systems Realization Laboratory
80