Finite Capacity Scheduling
Finite capacity scheduling is so-called because it takes capacity
into account from the very outset. The schedule is based on the
capacity available. Infinite capacity scheduling - the approach
used in MRP II - schedules using the customers' order due date
and then tries to reconcile the result with the capacity available.
There is no single accepted way to carry out Finite Capacity
Scheduling, and of the various approaches that exist, some are
proprietary secrets.
It is however possible to define certain approaches, or types
of scheduler:
Electronic scheduling board
The simplest scheduler is the electronic scheduling board,
which mimics the old fashioned card-based loading boards, but
the system calculates times automatically and will warn of any
attempt to load two jobs on the same machine. There is no scheduling
algorithm as such involved.
Order Based Scheduling
In Order Based Scheduling the tasks are scheduled on the basis
of order priority. The sequence at individual resources is determined
by the overall priority of the order for which the parts are
destined. It is a distinct improvement on infinite capacity schedulers
but its biggest drawback is that it allows gaps to appear on
resources. Some schedulers allow the process to be iterated to
try and reduce gaps and therefore reduce the time through the
system. This iteration can be very time consuming.
Constraint based schedulers, Synchronised Manufacturing
With the Constraint based schedulers, also known as Synchronised
Manufacturing, the idea is to locate the bottleneck in the line
and ensure that it is always loaded. The assumption is that non-bottlenecks
can take everything thrown at them, and this allows them to be
synchronised to the bottleneck through the Master Production
Schedule (MPS). The MPS is generated by loading the orders onto
the bottleneck and thus determining when they will be ready.
This system is inclined to produce gaps and is also very sensitive
to small changes such as a customer wanting to reschedule an
order.
Discrete Event Simulation
In Discrete Event Simulation the simulation loads all resources
at a point of time. When all contentions and queues are resolved
it moves on to the next set of events. Because the simulation
moves from one set of events to the next, there are far fewer
gaps in schedules produced this way and they are far more stable.
The problems with simulations are that they are: laborious ;and
also difficult to incorporate into other systems such as data
feedback from the shop floor.
Algorithms, Genetic algorithms
Algorithms usually suffer from being highly mathematical and
therefore user unfriendly, however more recently a new approach
has emerged under the general title of `genetic algorithms'.
These use a 'fitness' criterion. A typical example would be to
minimise the total time for jobs to stay in production. The procedure
starts with a schedule or family of schedules. The idea is to
try and improve them using a selection mechanism akin to natural
selection. 'Children' (new schedules) are bred using characteristics
(such as sequences of work) from parent schedules. If the new
child shows improved fitness i.e. is faster than the parents,
it replaces the worst schedule. While the approach looks promising
it is still in the early stages.
There remains the question of how these new approaches fit
with existing schedulers, particularly MRP
in which companies have invested vast sums. In the first three
cases they tend to replace the scheduling heart of the MRP system
while leaving the rest unchanged. To that extent the MRP system
acts like a database manager.
References
- Harrison. M., "MRP II & Finite Capacity Scheduling
- a combination for the 90's", Works Management, December
1991.
- Kirchmier. W., "Finite capacity Scheduling", Proceedings
of the 37th International Conference APICS, Falls Road, VA, 1994
|
