Você está na página 1de 5

A Simulation to Represent the Mapping of Systems Measure

of Performances to Single System Measure of Effectiveness

Lt.cdor.assoc.prof. PhD eng. Ghiţă Bârsan*


Lt.col. PhD eng. Paul Bechet**
Assoc.prof.PhD eng. Lavinia Gavrilă***
Carmen Vatamanu*
“Nicolae Bălcescu” Land Forces Academy, Sibiu*
Informatics and Electronic Warfare Application School, Sibiu**
“Henri Coandă” Air Forces Academy, Braşov***

Abstract
This paper presents the fact that complex systems require a simulation to
represent the mapping of systems measure of performances to single system measure
of effectiveness.
All military technical systems might be considered a complex system of systems.
From this point of view the system of systems requires a simulation to represent the
mapping of systems measure of performances to single system measure of
effectiveness.
Optimization of each system does not guarantee overall system of systems
optimization.
1. The Theoretical Framework of the Complex System of
Subsystems
The basic principle behind systems analysis is that complex systems
or problems can be decomposed into simpler sub-systems (or sub-
problems), which themselves may be subdivided into even simpler sub-
sub-systems, until a level is reached where the component parts may be
treated as elementary.
In this problem, three main ideas are important:
– the system of systems is considered as a whole entity;
– the individual systems are unfunctional when viewed from the
system of systems perspective;
– the process of determination an optimal upgrade includes:
– operational avalability analysis;
– cost modeling;
– linear or nonlinear optimization;
– modeling and simulation.
These assumptions are used in order to achieve overall mission or
objective.
The process of P.P.B.E.S. (Planning, Programming, Budgeting and
Estimation Systems) has developed a new method for non-material
solution by using analysis of alternatives.
This procedure has risen a question: how best to upgrade on existing
system of systems. By treating each acquisition system as a subsystem of
the larger entity doesn’t guarantee the increasing of the operational
availability of whole system.
The objective of this paper is to use modeling and simulation to
estimate a mathematical model for upgrading system performances.
A non-material solution implies the solving of one of two problems:
– to maximize the system of systems performance subjected to a cost
constraints;
– to minimize the additional cost under performance constrains.
The result of solving these two kinds of problems could be:
1. a new type of system must be added to the original system of
systems;
2. an existing subsystem must be replaced to increase performances.
2. Steps to Represent the Mapping of Systems Measure of
Performances to Single System Measure of Effectiveness
The model which will be proposed requires some steps:
– define the overall mission or missions of the system;
– define the overachieving measure of effectiveness (MOE) for the
system of systems that represents the goal of the model;
– define the characteristic MOE for each system and show how it
contributes to overachieve the MOE goal;
– represent the maps of system measure of performances (MOPs);
– specify the constraints: cost constraints for each component systems
and the overall system of systems, MOPs constraints. These constraints
depend on the type of problem we solve: to maximize performances under
cost constraints or to minimize additional cost under performance
constraints;
– formulate a model that will capture the map from component system
MOPs to system MOEs for overachieving MOE;
– solve an optimization problem under the constraints imposed;
– evaluate the solution and analyze alternatives.
3 The Performance Model for System of Systems
Mathematically, complex systems are characterized by multiple non-
linearities.
Physically, these formal properties are associated with connections
between subsystems and their environment.
Let consider n types of systems S n , that comprise a system of systems
S (figure no.1).
System S

Subsystem S1 Subsystem S 2 … Subsystem S n

Unit Unit Unit Unit Unit Unit … Unit Unit Unit

Subunit Subunit … Subunit Subunit

Subunit Subunit

Fig. no. 1 System of systems


S = S1 ∪ S 2 ∪ ... ∪ S n (1)
There are mi systems of type i and the total number of systems are:
n
m = ∑ mi (2)
i =1
Each system type has a set of MOPs, pi:
{
pi = p1i , p 2i ,... p mii } (3)
where each p ji , j = 1, mi has dimension ri
Each system type has one overall measure of effectiveness (MOE):
MOEi = fi (m, p1,..., pn ) (4)
which reflects how it contributes to the overachieving MOE for system S .
The whole system MOE is:
MOE = f (m, MOE1, MOE2 ,..., MOEn ) (5)
Each system’s MOPs are constrained at lower or highest
performances:
Pilow ≤Pi ≤Pi high (6)
The highest performances could be imposed.
Each system S cost unit is a function of performance expressed in
terms of MOPs:
ci = fi ( pi ) (7)
mlow < m ≤ m high (8)
The total cost is:
n
C = ∑ Ci (9)
i =1
Another constraint could be the number of component systems.
Now, with this model, we can formulate the cases of interest:
– first problem:
Maximize S = S1 ∪ S 2 ∪ ... ∪ S n system of systems performance
subjected to performance and cost constants:
max MOE = f (m, MOE1 , MOE2 ,..., MOEn )
with MOEi = fi (m, p1 ,..., pn ) (10)
subjected to:
mlow ≤ m ≤ m high
pilow ≤ p ≤ pihigh
(11)
Ci ≤ Cihigh
C ≤ C high
– the second problem:
Minimize S = S1 ∪ S 2 ∪ ... ∪ S n system of systems cost subjected to
individual cost and performance constraints:
n
min C = ∑ Ci (12)
i =1
subjected to:
MOE = f (m, MOE1 , MOE2 ,K, MOEn ) > MOE
With:
MOEi = fi (m, p1, p2 K pn )
mlow ≤ m ≤ m high
Ci ≤ Cihigh (13)
pilow ≤ pi ≤ pihigh
4. The Optimization Method
Optimization tools and techniques are designed, under strong
assumptions about the character of the system, to find the “best”
combination of characteristics for a system. For most complex systems,
this means the outcome of these tools and techniques will usually be a set
of “better” characteristics, rather than a “best”.
One of the most common and effective nonlinear programming
algorithms is known as the Davidon-Davies method or the Davidon-
Fletcher-Powell method with linear constraints.
The Davidon-Davies method is a gradient search method that can
handle linear quality and inequality constraints.
Another method is used MATLAB software. This program uses the
method of sequential quadratic programming to solve the fully general
nonlinear programming problem in which objective (goal) and constraint
function can be nonlinear. The particular routine is called “CONSTR” and
is contained is the “Optimization Toolbox”.
Basically, the method formulates a sequence of quadratic
programming subproblems based on a quadratic approximation of the
Lagrangean function and by linearizing the nonlinear constraints about the
current iterate. The simpler quadratic programming subproblem is solved
very similar to the Davidon Davies method in order to provide a search
direction for a line search procedure that provides the next iterate.
The original nonlinear function and constraint sits are then
approximated about the new iterate and the sequence is repeated until
convergence criteria are satisfied.
The objective function MOE is not always feasible.
5. Conclusions
Complexity is a fundamental concept when a system is to be modeled.
We have seen that there are many face of complexity for the complex
system of systems and these kind of systems require a simulation to
represent the mapping of systems measure of performances to single
system measure of effectiveness.
A good testing program is essential for reliable operational
performance, and will significantly reduce support and maintenance costs.

Bibliography

1. Bentley. P.J., Evolutionary Design by Computers, San Francisco,


Morgan Kaufmann Publishers Inc, 1999
2. Burrough, Peter A., Dynamic modeling and geocomputation. In:
“Geocomputation: a primer”. Eds. Longley, P., Brooks, S., Macmillan, B.,
and McDonnell, R., Ch. 9., New York: Wiley, 1989
3. Casti, John L., Alternative Realities Mathematical Models of Nature
and Man, New York, John Wiley & Sons, Inc., 1989
4. Diamantis, Z.G., Tsahalis D.T., Evolutionary Algorithms in
Engineering and Computer Science, West Sussex, John Wiley and Sons,
LTD., 1999, UK
5. Mann, H., Course on multipole modeling, simulation and analysis
of multidisciplinary systems, http://icosym.cvut.cz/course/

Você também pode gostar