17-03-2014, 07:12 PM
Introduction
Factory management in today’s frequently changing technical and
marketing environments calls for complex decisions and analysis.
These relate to the matters such as production processes, incoming
orders, cost accounting, and so on, and are based on a wide variety
of information about manufacturing systems. In steel manufacturing,
for example, the market now demands high-quality, valueadded
products for specified purposes. Small-volume, high-variety
production is therefore required. Consequently, the amount of
equipment used in steel-making processes is increasing, and the
processes themselves are becoming even more complicated. It is
therefore necessary to develop and analyze computerized supporting
tools for the efficient scheduling and control of production lines.
There are three major requirements for a system to be applied to an
actual scheduling environment:
Representation of the problems
Resources and organization are frequently changing. Therefore,
environments and constraints should be easy to represent
and modify.
0 Various evaluation criteria
A schedule is evaluated according to many conflicting criteria.
These vary from case to case, and therefore trade-offs between
them are difficult, and require expert knowledge.
Efficient generation of a solution
For short-range scheduling, such as daily scheduling, it is necessary
to get a feasible solution in a reasonable time. A real-time
rescheduling capability is also essential, so that mechanical
problems can be dealt with.
To meet these requirements, we propose a new approach called
cooperative scheduling, in which three different factors, namely, procedures,
rules, and the user, cooperate to make a schedule. Since
interaction with the user is included in the system architecture, the
purpose of this approach is not to replace the user and automatically
produce a solution to a problem that the user has formulated,
but to collaborate with the user in designing a solution.
In this approach, the final evaluation of the schedule is left to the
user: It
might seem too simple a way, but it is practical because the system
can be easily introduced into an actual environment where an
expert manually makes a schedule. The reason for this is that the
fully-automated scheduling system often produces a schedule that is
completely different from the manual one, and consequently the
human expert does not trust it because he cannot evaluate it.
Scheplan is a scheduling environment for steel-making processes
that employs cooperative scheduling architecture. It is designed to
assist in the process of daily scheduling, which is currently done
manually by a limited number of experts. The system was developed
as a result of a joint study with Nippon Kokan Co., Ltd.
(NKK) and NK-EXA.
The rest of this paper is divided into five sections. Section two
overviews OR and AI approaches to solving scheduling problems.
Section three introduces the steel-making processes and classifies
their constraints as scheduling problems. Section four proposes
if the schedule satisfies the user, it becomes a solution.
cooperative scheduling and explains its architecture and the flow of
the scheduling. Implementation of the architecture is described in
section five. Finally, section six summarizes our work and discusses
its future development.
Factory management in today’s frequently changing technical and
marketing environments calls for complex decisions and analysis.
These relate to the matters such as production processes, incoming
orders, cost accounting, and so on, and are based on a wide variety
of information about manufacturing systems. In steel manufacturing,
for example, the market now demands high-quality, valueadded
products for specified purposes. Small-volume, high-variety
production is therefore required. Consequently, the amount of
equipment used in steel-making processes is increasing, and the
processes themselves are becoming even more complicated. It is
therefore necessary to develop and analyze computerized supporting
tools for the efficient scheduling and control of production lines.
There are three major requirements for a system to be applied to an
actual scheduling environment:
Representation of the problems
Resources and organization are frequently changing. Therefore,
environments and constraints should be easy to represent
and modify.
0 Various evaluation criteria
A schedule is evaluated according to many conflicting criteria.
These vary from case to case, and therefore trade-offs between
them are difficult, and require expert knowledge.
Efficient generation of a solution
For short-range scheduling, such as daily scheduling, it is necessary
to get a feasible solution in a reasonable time. A real-time
rescheduling capability is also essential, so that mechanical
problems can be dealt with.
To meet these requirements, we propose a new approach called
cooperative scheduling, in which three different factors, namely, procedures,
rules, and the user, cooperate to make a schedule. Since
interaction with the user is included in the system architecture, the
purpose of this approach is not to replace the user and automatically
produce a solution to a problem that the user has formulated,
but to collaborate with the user in designing a solution.
In this approach, the final evaluation of the schedule is left to the
user: It
might seem too simple a way, but it is practical because the system
can be easily introduced into an actual environment where an
expert manually makes a schedule. The reason for this is that the
fully-automated scheduling system often produces a schedule that is
completely different from the manual one, and consequently the
human expert does not trust it because he cannot evaluate it.
Scheplan is a scheduling environment for steel-making processes
that employs cooperative scheduling architecture. It is designed to
assist in the process of daily scheduling, which is currently done
manually by a limited number of experts. The system was developed
as a result of a joint study with Nippon Kokan Co., Ltd.
(NKK) and NK-EXA.
The rest of this paper is divided into five sections. Section two
overviews OR and AI approaches to solving scheduling problems.
Section three introduces the steel-making processes and classifies
their constraints as scheduling problems. Section four proposes
if the schedule satisfies the user, it becomes a solution.
cooperative scheduling and explains its architecture and the flow of
the scheduling. Implementation of the architecture is described in
section five. Finally, section six summarizes our work and discusses
its future development.