Beruflich Dokumente
Kultur Dokumente
Multi-stage processes
OSA Tübingen / Teacher 1999 J. Bemetz / Martin-Heidegger-Gymnasium Meßkirch
J. Bemetz / materials to multistage processes
Page 1
Preface • The Lesson Multistage processes are matrices and vectors from its prev
ious close ties to solve geometric problems. In this expanded view is nothing ot
her than rectangular matrices and vectors are paying schemes are tuples. Both co
mputational objects can be interpreted as the content very wide, which tasks are
possible, which extend beyond the geometric aspects of linear algebra. Multista
ge processes are characterized in that one described by a state vector start sit
uation is transformed step by step with the help of transition matrices in succe
ssion situations. This can be done by such a transfer from stage to stage variou
s matrices (eg material linkages) or by applying the repeated and the same matri
x (eg, population trends). The three examples discussed in detail this collectio
n of material are selected so that will be developed with their help, the main c
hallenges and issues related to multilevel processes with the students can. Cond
itions are the matrix notation of linear equations (A ⋅ x = b) and the Gaussian
algorithm (cf. Curriculum Unit 5: Linear Systems). The exercises align themselve
s directly to the issues at the examples and extend them. Due to the variety of
application areas arise depending on the domain name different ways. The most im
portant are: - state vectors are also called distribution vectors, population ve
ctors, start vectors, output vectors need vectors, etc. - transition matrices ar
e also called (material) Verflechtungsmatrizen, input OutputMatrizen, etc. - The
elements of the transition matrix are called correspondingly transitional facto
rs, survival rates, death rates, transition probabilities, shares, etc. - the gr
aph to represent the transitions are transition graphs, arrow diagrams, flowgrap
h, Gozintographen, etc. mentioned. In dealing with the teaching unit Multistage
processes, the use of a CAS to be very helpful. Thus, for example the impact of
changing rapidly starting vectors or modified transition matrices are investigat
ed. The MAPLEWorksheets (MAPLE V Release 5) to the examples 2 and 3 are added to
these materials.
•
•
•
•
J. Bemetz / materials to multistage processes
Examples
Page 2
Example 1: Material interdependence
(Definition of matrix multiplication)
In a production process used to manufacture two intermediates Z1 and Z2 three di
fferent resources R1, R2, R3 and needs. From the two intermediates then create t
hree different end products E1, E2 and E3. The figure below can be found, how ma
ny units of the raw materials for the various intermediates and how many units o
f the intermediates for the respective end products are needed. Wanted is the ra
w material needed for the various end products. R1 R2 2 3 3 4 4 1 R3
Z1
Z2
3 E1
2
2 E2
1
3
3 E3
The needs of intermediates for the final products and the demand for raw materia
ls for the intermediates is often provided in the form of tables: Z1 Z2 2 1 3 2
3 3 3 3 R 1 R 2 R 3 4 2 1 4
E1 E2 E3
Z1 Z2
Representation of these tables as LGS: E1 = E2 = Z2 + 2 3Z1 2Z1 + 3Z1 + 3Z2 Z2 =
E3 In matrix notation: E1 3 2 Z1 E 2 = ⋅ 2 1 (1) E
3 3 Z2 3 R1 3 Z1 4 1 = Z 4 3 2 ⋅ R2 R
2 3 3 Z1 = R1 + R2 + R3 4 Z2 = 3 R1 + 2R2 + 4R3
(2)
J. Bemetz / materials to multistage processes
Examples
Page 3
Inserting (2) in (a) provides the desired relationship between the finished prod
ucts and raw materials: E1 2 3 = E2 e 3 3 2 R
1 1 3 4 ⋅ 1 R2 ⋅ 3 2 4 R 3
B
(3)
A
This connection between the end products and the raw materials required, however
, can also describe a single matrix C: c E1 11 c E 2 = 21
E c31 3 c12 c13 R1 c 22 c 23 ⋅ R2 R3 c32 c33
C
(4)
In this example, spends c21 on how many units of the raw material for the finish
ed product R1 E2 are required (E 2 = c 21 + c 22 ⋅ R1 ⋅ R2 ⋅ R3 + c 23).For the
calculation of c21 you need the numbers necessary for the final product Zwische
nprodukteinheiten E2 (2 units of Z1 and Z2 of a unit: second row of the matrix A
) and the intermediate units contained in each of the raw material R1 (3 units o
f R1 in Z1 and 3 units of R1 in Z2: first column of the matrix B). and c21 = 2 ⋅
1 ⋅ 3 + 3 = 9 = a 21 c21 b11 ⋅ ⋅ + a 22 b 21
General: To receive the item cik, multiply the elements of the i-th row of A in
sequence with the corresponding elements of the k-th column of B and adds the pr
oduct (scalar product of the i-th row of A with k th column of B). The so-define
d shortcut that maps the two matrices A and B is a matrix C, matrix multiplicati
on. We write:
C = A ⋅ B
Definition: A ⋅ B multiplication of two matrices A and B is defined if and only
if the number of columns of A is the number of lines of B the same. If A is a (m
, n)-matrix and B an (n, p) matrix, then C = A ⋅ B a (m, p) matrix for the eleme
nts cik is: ai1b1k cik = ai + b2 2 k + .... Ain + bnk (scalar product of i-th ro
w of A with k-th column of B)
J. Bemetz / materials to multistage processes
Examples
Page 4
Notes: • The matrix multiplication is not commutative. • Identity element of mat
rix is 1 0 ... 0 0 1 ... 0 = unit matrix En, a (n, n)-matrix, in whic
h all elements of the ... ... ... ... 0 0 ... 1 main diagonal equal to
1 and all others are 0. If there is no confusion over n, we write E instead oft
en En particular, is: E ⋅ x = xx: column vector
With the scheme of the matrix Falk1 can be calculated in a clear way: Pnnm
B C = a.b
A
For our example is: R1 R2 3 3 15 9 18 4 2 16 10 18 results R3 1 4 11 6 13
Z1 Z2 E1 E2 E3 Z1 Z2 3 2 3 2 1 3
E2 for the final product, for example, you need that is 9 units from R1, R2, and
10 units of six units of R3. Notes on the content of the right combination of t
he matrices in material linkages
1
Sigurd Falk, a professor at the Technical University of Braunschweig
J. Bemetz / materials to multistage processes
Examples
Page 5
The matrices are always correct links, if all sub-processes and the overall proc
ess have the same orientation, ie, if the need for higher production level eithe
r in all matrices (even in the Product Matrix!) From the lines or in all matrice
s from the columns can read. Example: For a two-stage production process, the ma
terial linkages are given by the following tables: E1 E2 Z1 Z2 1 3 2 4 5 7 Z1 Z2
R1 R2 6 8
The following two models are correct content marked links: Z1 Z1 Z2 12 34 R1 R2
5 7 19 43 6 8 22 50 Z1 Z2 Z1 57 68 E1 E2 1 2 19 22 3 4 43 50
E1 E2
R1 R2
All matrices are line-
All matrices are column-
Mathematical background is: C = A ⋅ B applies, then: CT = BT AT ⋅ The transpose
of a product is equal to the product of the transpose in the reverse order of th
e factors.
J. Bemetz / materials to multistage processes
Examples
Page 6
Example 2: beetle population
(Matrizenpotenzen, cyclic matrices)
A beetle female lays 80 eggs and dies soon after. From the concomitant developin
g larvae (grubs) survive only a quarter of the following year. In the second yea
r only one quarter of the larvae survive. In the third year, the larvae pupate a
nd from a fifth of them develop in the years following beetle females lay the eg
gs again 80.
80
1 larvae
0.25
Larvae 2
0.25
Larvae 3
0.2
Beetle females
We study the development of a start population of 6000 larvae 1, 2000 larvae 2,
300 and 500 larvae three female beetles. State vector of the starting population
: l1 (0) 6000 l 2 (0) 2000 po = = l3 (0) 300 k (
0) 500 For the distribution after one year is clear: l1 (1) = 80 ⋅ k (
0) l 2 (1) = 0.25 ⋅ l1 (0) l 3 (1) = 0.25 ⋅ l 2 (0) k (1) = 0.2 l ⋅ 3 (0) LGS i
n matrix notation: l1 (1) 2 l (1) l (a) = 3 k (1) 0 0 80
l1 (0) 0 0 0 0 l 2 (0) ⋅ 0.25 0 0.25 0 0 l 3 ( 0)
0 0 0.2 0 k (0) 0 0 80 0 0 0 0 0.25 with the transition matrix
U = 0 0, 25 0 0 0 0 0.2 0
p1 = U ⋅ p 0
J. Bemetz / materials to multistage processes
Examples
Page 7
Starting from the initial vector p 0 we can therefore now the populations for th
e following years: route 40,000 1500 p1 = U ⋅ p0 = 500 60 8
000 1200 p3 ⋅ U = p 2 = 2500 75 4800 10 000 p 2 =
U ⋅ p1 = 375 100 6000 p 2000 4 = U ⋅ p3 = p0 = 300
500
In a cycle of four years, thus returns to its initial population.
Development of the beetle population
Such four-year cycles observed in beetle populations in the country!
J. Bemetz / materials to multistage processes
Examples
Page 8
If we define the powers of square matrices, one might then be the cause for this
cyclic behavior quickly see through: Definition: Under the n-th power An (n ∈ N
) of a square matrix A is understood as the n-fold product of A with itself . ar
e also to be: A0 = E (unit matrix) with the help of Matrizenpotenzen we can writ
e now: p1 = U ⋅ p 0 p 2 = U ⋅ p1 = U ⋅ U ⋅ p 0 = U 2 ⋅ p 0 p3 = U ⋅ p 2 = U ⋅ U
2 ⋅ p 0 = U 3 ⋅ p 0 p 4 = U ⋅ p3 = U ⋅ U 3 ⋅ p 0 = U 4 ⋅ p 0 Thus the four-year
cycle is characterized in that applies to the transition matrix U: U4 = E recalc
ulating supplies for our example now: 0 16 0 0 0 0 0 0 20 0 3 U2 =
U = 0.0625 0 00 0 0 0 , 0125 0.05 0 0 4 0 0 0 0 4 0 0 0 0
5:00 0 1 0 0 U4 = 0 0 1 0 0 0 0 1 0 0 0 = E 0 1
We call matrices that lead to such cycles, cyclic matrices. Definition: A square
matrix is called cyclic if there exists k ∈ N, so that Ak = E. Notes: • For the
occurrence of the four-year cycle is crucial that the product is made of the pr
opagation rate v and the three survival a1, a2, and a3 equal to 1. 0 a1
For U = 0 0 0 0 0 0 0 0 a2 a3 v 0 0 0 2 a1 a a3 v 0 a1 a3 a 2 v 0
0 0 4 gives U = 0 0 0 2 a1 a a3 v 0 0 0 0 0 a1 a 2 a 3 v U 4
= E for a 2 a1 a3 v = 1
J. Bemetz / materials to multistage processes
Examples
Page 9
•
Is a 2 a1 a3 v> 1, then the beetle population to grow long term, however, is a 2
a1 a3 v <1, so extinction of the population. Examples: I. Doubles are for examp
le v due to changes in environmental conditions at constant survival a1, a2 and
a3, the rate of increase, it follows that a 2 a1 a3 v = 2. 2 0 0 0 0 2
0 0 4 This follows: U = = 2E (doubling once every four years) 0 0 2 0 0 0
0 2
Unlimited growing population
II is half the growth rate at constant survival rates, however, we have: a1 a3 a
2 v = 0.5. 0 0.5 0 0 0 0 0.5 0 4 This leads to U = = 0.5 E (halved a
fter 4 years) 0 0 0.5 0 0 0 0 0.5
Endangered Population Endangered Population
J. Bemetz / materials to multistage processes
Examples
Page 10
the computational cost should be kept low, it lends itself to the introduction o
f cyclic matrices in the following example with a (3,3) - Transition Matrix: Dev
elopment of a fictitious beetle population A beetle lays so many eggs that devel
op from it in the next 10 years larvae. Soon after he dies. A fifth of these lar
vae survived the first year of the second year, half of the larvae pupate and ar
e in the third year in a beetle. Investigate the development of a one-year from
60 initial larvae (L1), 30-year larvae (L2) and 18 beetles existing population.
10
L1
0.2
L2
0.5
K
Transition matrix:
0 10 0 U = 0.2 0 0 0.5 0 0
U3 = E results in the development of this population a cycle of three years. Map
:
J. Bemetz / materials to multistage processes
Examples
Page 11
Example 3: Inheritance of characteristics
(Stationary distribution, stochastic matrices)
A population of insects contains animals with two different characters A and B (
eg, color). Observations show a long time that insects with characteristic A 30
to 70% offspring with feature A and feature B% are those with. Insects with feat
ure B to have offspring with 80% again this characteristic, such as 20% is chara
cteristic of A. The multiplication rate is not influenced by the features. Überg
angsgraph:
0.3 0.7
A characteristic feature of B
0.8
0.2
xA (0) be the proportion of insects with a feature to start watching, xB (0) in
accordance with the one who is as characteristic for the distribution of charact
eristics in the next generation then xA (1) = 0.7 xA (0) + 0.2 xB (0) xB (1) = 0
.3 xA (0) + 0,8 xB (0) 0.7 0.2 0.3 0.8 U = x A (0) With the help
of the initial vector x0 = x (0) and the transition matrix U can be the dis
tribution B The transition matrix is thus given: in the nth generation (see
Example 2): xn = n U can ⋅ x0 By calculating the following Matrizenpotenzen we a
re in an overview of the development the distribution of the gain characteristic
s: 0.55 0.30 0.45 0.70 U2 = 0.420 0.388 0.582 0.613 U5 =
0.475 0.350 U3 = 0.525 0.650 0.410 0.394 0.591 0.607 U6 =
0.438 0.375 0.563 0.625 U4 = U12 = 0.401 0.400 (U6) 2 = 0
.601 0.601
Notes: • The fact that the column totals are not exactly one is due to rounding
errors. • If you want to go quickly to high Matrizenpotenzen, so we recommend th
e following procedure: A 2 = A ⋅ A, A 4 = A 2 ⋅ A 2; A8 = A 4 A 4 ⋅, ⋅ A16 A8 =
A8, etc.
J. Bemetz / materials to multistage processes
Examples
Page 12
Assumption: The Matrizenfolge (Un) is a border matrix G is approaching with:
0.4 0.4 n G = lim U = 0.6 0.6 n → ∞ Is this the case, it is to stabiliz
e the distribution to a limit vector x out. With a uniform distribution of chara
cteristics in the starting population (xA (0) = 0.5 and xB (0) = 0.5) we get:
0.4 0.4 0.5 0.4 x = G ⋅ x0 = ⋅ 0.6 0.6 0.5 0.6 =
the long term will be adjusted so a stable distribution of characteristics su
ch that the feature A at 40% and the feature B at 60% of the insects found. The
long-term recruiting stable limit distribution (= stationary distribution) does
not depend on the initial distribution of the features: a Be x0 = 1 - a
with a ∈ [0,1] an arbitrary initial distribution, the result is a stable a
+ 0.4 0.4 (1 - a) 0.4 limit distribution: x = x0 = G ⋅ a + 0.6 0.6 (1 -
a) = 0.6 Assuming the convergence of distributions of xn to a lim
it distribution point x, the limit distribution can also be calculated without k
nowledge of the boundary matrix. The limit vector x is characterized in that it
is no longer under the effect of the transition matrix changed. It is therefore:
U = x ⋅ x Applied to our example: 0.7 0.2 x A x A ⋅ 0.3 0.8 x x
= B B xA and xB to calculate a homogeneous LGS must therefo
re be solved: 0,7 xA xB = xA + 0.2 0.3 + 0.8 xA xB = xA xB -0.3 + 0.2 = 0 0.3 xA
xB - xB = 0 0.2 (a) (2) (a ') (2')
An equation turns out to be unnecessary. We choose t = xA xB = 1.5, and then get
t. The required boundary vector is calculated as the special solution of the ho
mogeneous LGS under the constraint: xA + xB = a (xA and xB specify the proportio
ns that occur with the characteristics in the population, therefore their sum mu
st be equal to 1). t + 1.5 t = 1 ⇒ t = 0.4 0.4 border vector: x = 0.6
0 4 7 3 7
4 = 0 Ug 7 3 7 0 0 4 4 = 7 7 3 3 7 7 0
g 7 3 7
0 1 2 1 2
0 0 2 4 ⋅ 3 3 7 1 3 7
0 4 7 3 7
0 4 7 3 7
J. Bemetz / materials to multistage processes
Page 30
References: [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] Bosch, Karl,
Mathematics for Economists, Oldenbourg Verlag 1999 Hahn / Dzewas, Linear Algebr
a / Geometry, Westermann 1996 Lehmann, Eberhard, linear algebra with vectors and
matrices, JBMetzler 1990 Lehmann, Eberhard, Linear Algebra with the Computer, T
eubner 1983 Kroll, Reiffert, Vaupel, Analytical Geometry / Linear Algebra, Dümml
er 1997 LEU Issue M 36, Amann, elinka, To the application of Mathematics Educat
ion, Issue 3 Linear Algebra and Analytic Geometry, 1994 LEU Issue M 41, How the
teaching of mathematics changed by the use of a computer algebra system?, 1998 R
ényi, Alfred, probability, VEB German publisher of cience, Berlin, 1979 Black,
Jochen, Mathematics for Economists, Verlag Neue Wirtschafts-Briefe, 1996 igma,
analytical geometry, Velcro (out of print) top, Friedmar, including, Mathematic
s IV (Matrices, Linear optimization), book publisher Leipzig-Köln, 1992 Trinkaus
, Hans L. problems? Higher Mathematics!, pringer-textbook, 1993 Vohrer, Thilo,
school vocational schools and specialized secondary schools, publishing T. Vohre
r 1997
A collection of sample tasks of the academy in Karlsruhe on "multi-step processe
s" can be found at: http://www.lehrer.uni-karlsruhe.de/ ~ za242/osa/mstproz/Must
er98.html