Optimization
The heart of APS systems are
advanced mathematical
algorithms and logic. The
mathematics formulate business
problems and solve them to a
selected objective function.
The challenge is to
formulate the problem in
discrete terms that
characterize the reality of
the supply
chain's physical attributes
and capability then solve it
in a reasonable time
frame. The answer must
satisfy all
constraints (capacity,
labor, inventory storage,
lead
times) while using available
variables.
Optimization Objectives
Business optimization
objective functions can be
lowest cost, highest profit,
greatest return on
assets/investment,
or highest customer service.
Different objectives are
picked for strategic,
tactical, and operational
planning.
Problem Size
The problem size grows very
large because of the multiple
effect of combining the
number of products, constraints,
variables, work centers, and
time buckets. Older
technologies were selected
because they were able to
generate solutions within a
reasonable time frame. Basically,
they employed simplifying
methods to increase
solve time. This
simplification made the
solutions less
accurate, less optimal.
However, new mathematical
methods have greatly
improved the solve times.
That,
in combination with faster
computers, has now made
accurate representation of
the problem possible without
the earlier simplification
sacrifices.
Optimization Technologies
Now and into the future
technologies will be
combined to
accomplish three things:
better representation of
reality, faster solve times,
and an easier way for the
user to abstractly
formulate the business
problems in the models.
Finite Capacity Scheduling
The area that will benefit
the most from enhancement
in these technologies is
finite capacity scheduling.
These
problems are the most
complex mathematical
problems
known to man.
Ever-tightening constraints,
increasing
customer expectation, and
shorter lead times will
ultimately
put more pressure on
technologies to solve
tougher scheduling problems.
Discrete Representation
Bottlenecks in manufacturing
float because different
rocks pop to the surface at
different times. If
constraining
resources are not
represented in the model,
then
there is no visibility of
the bottleneck. Future
models
will include more
constraints that would have
overcomplicated
the models in the past,
which possibly made
them impossible to solve.
Simultaneously Solve
One of the classic
techniques used to solve
complex
models is to treat certain
variables as constants and
solve
for one variable at a time,
causing the user to manually
iterate the problems to
solve for "real" variables.
It is
not uncommon to have three
or more variables that exist
in a scheduling problem.
Future PCS tools will solve
all existing variables
simultaneously. Four key
ones are lot
size, sequence, multilevel
coordination, and sourcing.
These are all necessarily
interdependent.
Alphabet Soup
APS models will take a
broader look at the supply
chain,
as well as automate the
solution of many parts at
the
same time. This will
minimize the whipsaw affect
illustrated
by the Beer Game.
Continued