Project Aims
Poor production scheduling in large scale process
industries such as steel mills, aluminum processing plants, etc., are
responsible for a high proportion of non value adding waste in those industries
in the form of excess inventories and lost sales due to late deliveries.
Generic scheduling systems have had comparatively little success in solving
this problem due to
- the large amounts of domain-specific and sometimes subjective human
knowledge that is usually involved in constructing a satisfactory schedule
which cannot be represented generically
- the lack of attention generally paid to the reactive as opposed to the
predictive aspects of schedule construction
- the inadequate ability of the user to participate in the schedule
construction process and to interact with the schedule in a satisfactory way to
input their subjective preferences for such factors as technological sequencing
constraints, due date preferences etc., and to explore interactively the
possible trade-offs between different schedule effectiveness
criteria.
Advances in object oriented
programming techniques, with the associated concepts of extendible, re-usable
objects and frameworks are starting to provide potential ways of overcoming the
problems of configuring a generic scheduling system to specific industrial
environments. Additional research on the specific problems of reactive
scheduling, including the use of fuzzy temporal constraint representation to
assess reasonable cut-off limits for the extent to which unexpected events will
propagate down a schedule are likewise providing possibilities for developing
scheduling systems that have a substantive capability for reaction to
contingencies. Research on the representation, manipulation and combination of
subjective preferences, combined with graphical methods of exploring multiple
criteria trade-offs are allowing the possibility of more effective forms of
user interface to the schedule.
Technical Objectives
To extend the existing framework with the following
components:
- a fuzzy temporal constraint representation and propagation component for
constraining the computations on the effects of schedule disturbance to zones
where event timings can be predicted with sufficient precision , rather than
over the whole schedule,
- a component for filtering and monitoring the relevant data of the
application / simulation,
- a component for evaluating schedules against multiple performance criteria
including robustness, stability, and nervousness,
- a repair-based scheduling algorithm to react in a generic way to
unanticipated events on the shop floor in real-time, and
- a graphical user interface for exploring the trade-offs between different
criteria and monitoring the dynamic production process.
To develop a generic simulation package with the following
components:
- a language interpreter to describe a wide range of scheduling problems
- a component to generate benchmarks randomly
- a simulation capability allowing the exploration of the effectiveness of
different types of scheduling policies in alternative future production
scenarios,
- communication capability for a wide range of reactive scheduling
systems.
To develop a prototype scheduling
system with the framework for the specified scenario, and to compare this system
- with dynamic dispatching rules and
- with the OPIS-system.
Project Results
The output of this project will be:
- an extended scheduling software framework that can be easily adapted to
specific plant requirements in terms of the different types of managerial and
technological preference constraints, and which can use a wide range of
different scheduling techniques.
- a simulation tool that can be used to evaluate the realized scheduling
system as well as other reactive scheduling systems.
- a set of performance benchmarks for reactive scheduling of a wide range of
flow-shop scenarios. This set will facilitate evaluation of scheduling
strategies in the simulation package.
- a scheduling system communicating with the simulation model and solving
the reactive scheduling benchmarks.
- a reuse manual describing how the existent framework can be reused in
other applications.
The scheduling framework and
systems built with the framework are likely to result in a long term increase
in the competitiveness of the industries which use it, particularly with regard
to reduced throughput times, inventory reductions, and better customer due date
performance.