while algorithms have improved by a factor of 55 000
in the same period. The joint forces result into a
staggering factor of 66 million, or in other words: a
planning model using linear programming that would
have taken two years to solve in the 1990s, could be
solved in just one second today.
Benchmarking
The power and quality of optimisation software is
often evaluated on benchmark data for idealised
planning problems. For transport planning, such an
idealised problem is known as the Vehicle Routing
Problem with Time Windows (VRPTW), where a fleet
of vehicles with limited capacities must supply goods
from a central depot to customer sites within given
time windows. Over the past years, a benchmark
suite prepared by M. Solomon in 1987 that remains
challenging until today has been used to test virtually
all scientific research contributions for solving this
problem. When a core module of Inform’s current
transport optimisation solution was evaluated on this
benchmark set, it always found solutions equal to the
best known solutions from the scientific literature
and in some cases was able to find better solutions,
improving upon the state of the art.
Advanced optimisation approaches can often
benefit from the current trend in computer hardware
to increased parallel processing capability that can,
for example, be seen in multi-core CPUs. To exploit
this hardware feature, a planning task must be
decomposed into parts that can be solved in parallel.
This can lead to a faster computation of plans. A
procedure called FSMAGP developed by Inform for
automatically partitioning planning tasks by solving
the so-called Graph Partitioning Problem, is currently
leading in the scientific benchmarking competition.
With its help, very large transport planning problems
can be solved quickly and efficiently.
Real transport planning tasks are much more
complex than idealised benchmark problems, but
many of the solutions techniques can be transferred
to real world applications. This approach has recently
been successfully implemented at one of Germany’s
largest cement suppliers.
Algorithms make the difference
Table 1 compares three different scenarios for a
cement producer in Europe with an annual transport
volume of about 1.6 million t: no optimisation
software, intelligent optimisation versus Inform’s
advanced optimisation.
With advanced optimisation the number of trucks
can be reduced from 100 to 88, which represents a
productivity increase of 12%. Fixed costs per truck
remain the same (
g
72 000), while variable costs per
truck rise from
g
15 500 to
g
17 360. Overall annual
costs per truck will therefore increase slightly from
g
97 500 to
g
99 360, but since the number of trucks
is reduced by 12, the costs per t drop from
g
5.77 to
g
5.18. This means that the cement producer will save
almost
g
1 million per year.
If standard intelligent optimisation is used, the
number of trucks will only be reduced to 93 (which
represents a productivity increase of 7%). This
translates into annual savings of just
g
0.57 million.
The example shows the enormous potential of
high quality algorithms and optimisation methods:
the difference between intelligent and advanced
optimisation is more than 400 000 per year. So when
assessing planning software, it is always necessary
to ask which optimisation methods are used. If the
answer is advanced optimisation, the return on
investment is radically higher than with standard
optimisation.
Strategic Decisions
The use of intelligent planning software is not
only limited to road transports. Strategic network
planning enables to optimise the distribution of
bulk material in terms of time to market and cost
per unit moved. By using real time optimisation
software for network wide transport planning,
resources can be used on the day in a more efficient
way. In a smart network, ship, rail and road
resources can be shifted where required, preventing
idle times in some locations and work overloads in
others. An integrated approach allows to update all
plans across the business within seconds.
PROCESS SOLUTIONS
Partnerships in Cement
Quality & Environmental Solutions
Cement Distribution Solutions
Plant Automation Services
