In the thrust of nations towards ever higher competitive
advantage,
container terminals play a pivotal and indispensable role. Technological
advances and competition amongst them, have forced terminals to raise their
efficiency to remarkable levels. This was not always the case: there were
times, during the protectionist years following WWII, when inefficient ports
were tacitly encouraged by local exporting interests, seen as protection from foreign imports. These were the days of the general cargo freighter,
often known to spend half of her time in port, waiting to berth, unload and
load.
Seafaring was fun during those days. Today, the ship is turned around in two
days and the terminal may be 50 kms or more from the city. Even if public
transport did exist, the youngster would rather relax in his airco berth, or by
the pool, or playing a game of snooker with his mates. At any rate, he would
again be back home in a couple of weeks.
Simply put, efficiency means two things: Either we
strive to achieve a certain output (i.e., number of containers handled per
annum, or ships hosted in our berths) with as low a cost as possible; or, given
a certain endowment of port resources (i.e., cost), we struggle to maximize the port’s
output. The methodology commonly employed in this type of efficiency
assessments is known as Data Envelopment Analysis (DEA): a mathematical
programming approach, producing ‘frontiers’ of best practice (i.e., top
efficiency), against which all other firms (in our case port terminals) can benchmark themselves; a procrustean bed, so to speak.
Whatever the case, cost control is the paramount
consideration of port management. One of the ways to achieve this is to
minimize the movement of containers and their handling-equipment in and around
the terminal. A few examples might suffice: Minimize the turnaround time of
ships (how many ship-to-shore cranes can I deploy on a large ship before I
start realizing diseconomies?); Minimize container rehandles in the stacking
yard (if a container departs in two hours, you may not stack other containers
on top of it); minimize the distance between the berth and the
stacking yard; minimize the distance an external truck must travel between the
gate and the place where it must drop its export container; minimize the time a
truck must wait in the parking lot before it can enter the terminal to pick up its
container; and so on.
If one wants, and one must want, problems could become even
more challenging: Stack containers in the yard according to the stowage-plan of
calling ships; i.e., optimize ship stowage-planning and yard-planning
simultaneously and well in advance, adjusting terminal-planning according to
ship operations in previous ports. Yard-planning is an operations research
(OR) challenge but, even in the largest of terminals, like Shanghai, Rotterdam,
Singapore or Los Angeles, the problem has been efficiently solved through the
development of advanced IT software. But if yard-planning is a challenge, stowage-planning
may be an even bigger one: A stowage plan needs to take into account not only
the ship’s port-rotation, but also a) stability considerations during loading
(the clearance between the ship’s keel and the seabed, these days, may be less
than half a meter, and if loading along rows is not even, touching the seabed
could be disastrous); b) alliance-members’ dedicated bays on the ship; c) crane
density; d) different sizes of containers; e) dangerous goods, and more. If stowage-planning
and yard-planning are ‘challenges’ in themselves, trying to optimize them simultaneously is
an OR nightmare. It is not by accident that my brightest students (in math and
OR) work at PSA, DP World, Hutchinson and other global terminal operators.
But still we haven’t explained what are ‘dual transactions’
in container terminals.
There are two types of external trucks visiting a container terminal: those who
bring export containers from the hinterland, to be loaded on arriving ships;
and those who come to pick up import containers, already unloaded and waiting
in the container yard. Usually, in both cases, one of the two legs of the trip
is unproductive: the ballast leg, as we would say in shipping. A truck
drops the container at the terminal and returns empty; another goes empty to
the terminal to pick up an import container. This type of inefficiency -if we
could call it that- not only leads to higher transport costs but, these days,
it causes things even more important: these
are the negative externalities of land infrastructure use, i.e., pollution,
congestion, and road accidents. It would be interesting at this point to make a
small diversion.
The drive to efficiency, as we said above, has to do with
maximizing output (e.g., number of containers handled) given a certain
endowment of resources (cranes, land, people). Today, however, there is a new factor entering
the efficiency calculation: This concerns the minimization of negative externalities from
port operations, such as sea and air pollution, noise, disturbances of sea
ecosystems, accidents, impacts on local communities and on commercial
activities (e.g., fishing, aquacultures, etc.), conflict with urban development
plans, road congestion around the port and so on; the list goes on.
All these are called ‘negative output’ of port operations and reducing them is
equivalent (or it should be seen as equivalent) to increasing ‘output’.
The question of course here, as in all cases involving negative externalities,
is how to price them, who should pay for them and how, and what would be the
impact of higher prices on trade and welfare. But let us finish our diversion
here and return to our dual transactions.
Can a hinterland consignee know which trucks take export
containers to the port so that he can ‘book’ one to pick up his waiting import
container and bring it to him? And can the truck going to the port to pick up
an import container know who, in the hinterland, needs a truck to carry his
export container to the terminal? Technically, this exchange of information
shouldn’t be too difficult to organize, and a simple APP could take care of it.
The impact of dual transactions on terminal management requirements, however,
is considerable, and this is the problem we have tried to solve in this research, the
development of which took us more than three years. This is why:
The terminal management system we have described above,
i.e., stacking-berthing-gate (etc.) operations, now needs to be modified to
accommodate the following dual transaction considerations: a) An incoming dual
transaction (DT) truck cannot wait at the gate and needs to jump the queue; b)
the truck cannot wait either at the queue of the export block to drop its
container; priority should be given to it over the operations of internal
terminal trucks and other handling-equipment; c) when the truck is ready to
move to the import stack to pick up a container, the handling-equipment (e.g.,
bridge cranes, straddle carriers; reach-stackers, forklifts, etc.) should be
ready and waiting and, ideally, the availability of the equipment should have
been planned in advance. So, in short, if optimizing shore and yard operations
jointly (we have carried out research where even gate operations are included
in this optimization) in a nightmare, the inclusion in the problem of dual
transactions makes the problem apocalyptic. This, because now one needs to
develop a heuristic algorithm that jointly optimizes: gate; berth; stowage;
yard; export/import blocks and handling-equipment deployment.
In an effort to address these issues, we have developed a bi-objective
mixed integer programming model that optimizes the allocation of
appointment quotas simultaneously with the deployment of (yard) cargohandling
equipment. The model addresses the challenges posed by the different types of truck
movement in the terminal, i.e., delivery, pickup, and dual transaction. These
require different handling-equipment, involving various deadlines, and multiple
priorities. To estimate the queuing length of external trucks in single or dual
transactions (as well as that of internal trucks), we have set up a novel three-level
vocation queuing model. For the bi-objective optimization, we propose a revised
non-dominated genetic algorithm, to obtain the approximate optimal
solution. Experimental results have proven the efficiency and effectiveness of
our method, which outperforms all similar algorithms. We show that our vocation
queuing model can estimate the prioritized queuing process more effectively in
three respects: a) the 3-level queuing; b) discrete truck arrivals in the
queuing system; c) non-interruption of servers. Our quota optimization design improves
the model’s applicability to real cases, especially in the case of dual
transactions. We finally demonstrate that the method proposed here, if adopted,
could help terminal operators allocate quotas and simultaneously match the
capacity of yard-handling, thus improving truck services, cost reductions and
environmental impacts. The benefits to be enjoyed by port users, because of
higher terminal efficiency, are only too obvious to be discussed.-
HH, May 2022.
