IFAC PapersOnLine 50-1 (2017) 7953–7958
ScienceDirect ScienceDirect
Available online at www.sciencedirect.com
2405-8963 © 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2017.08.896
© 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Keywords: Schedule robustness; Production planning and control under uncertainty; Job and activity
scheduling
1. INTRODUCTION
The specification of efficient production schedules is
important for job shop manufacturing systems execution.
Each order has to be assigned to an available machine so that
specific performance measurements are improved. However,
the increasing customization of products requires that
production handles higher numbers of product variants along
with decreasing lot sizes. Furthermore, to ensure process
quality and speed, several specialized machines are
employed. Hence, job shop manufacturing systems become
more complex and the efficient management of these systems
requires reliable and robust scheduling methods. In general,
the computation of schedules employs optimization
techniques, such as mixed-integer programming (Mula et al.,
2010) but do not consider schedule stability or robustness.
Despite the increasing power of modern solvers, most real
manufacturing systems are still too complex to be properly
represented, modeled and solved without ample
simplifications (they belong to the class of NP-hard
optimization problems). As a consequence, the optimal
solution often cannot be computed or requires long
computation times. Therefore, frequently heuristic methods
are applied. These methods cannot guarantee optimality but
are often able to generate near-optimal solutions in relatively
short computation times (Papadimitriou, 2003). One of the
simplest heuristic scheduling approaches is the use of
dispatching rules, assigning a specific priority value to each
job (or order) in the queue of a machine according to some
predefined criteria, such as the remaining time until its due
date (Pickardt and Branke, 2012; Rajendran and Holthaus,
1999). Whenever a suitable machine is available, the job with
the highest priority is chosen for the next operation. Due to
the low implementation effort, dispatching rules are often
used. A more sophisticated approach, which offers more
flexibility, is the use of meta-heuristics, such as genetic
algorithms. Genetic algorithms are able to compute solutions
also for larger instances of combinatorial problems than exact
optimization methods. However, they also feature limitations,
such as the dependence on the choice of several parameters
and the possibility to converge towards local extrema
(Jungwattanakit et al., 2008). Moreover, both, exact
optimization methods as well as meta-heuristics are not able
to deal with two particular characteristics of production
systems: dynamics and randomness (stochasticity), which can
greatly compromise the initial production schedule.
During production, it is well known that random events occur
affecting the adherence to a production schedule. Indeed,
production systems can be regarded as dynamic systems that
are subject to internal as well as external dynamic and
unexpected factors, such as: (i) changes and delays in
raw/component materials arrival from suppliers; (ii)
unacceptable quality for raw/component materials from
suppliers or during production, leading to reprocessing,
reworks and reordering, and consequently to delays; (iii)
unexpected production resources (machines, tools, castings
etc.) failures, which require stops for maintenance and repair;
(iv) workers absenteeism; (v) need for extra cleaning of
production resources and other unexpected setup activities
(Vieira et al., 2003; Vieira et al., 2009a; Vieira et al., 2009b;
Scholz-Reiter et al., 2002).
*Industrial and Systems Engineering Department, Federal University of Santa Catarina, Florianopolis, Brazil
**BIBA - Bremer Institut für Produktion und Logistik GmbH at the University of Bremen, Bremen, Germany
***Department of Production Engineering, University of Bremen, Bremen, Germany
(e-mail:
1.
g.vieira@ufsc.br,
2.
r
.
kue@biba.uni-bremen.de,
3.
enzo.frazzon@ufsc.br,
4.
fre@biba.uni-bremen.de)
Abstract: Complex stochastic job-shop scheduling problems can be handled by simulation-based
optimization (SBO), combining the optimization capabilities of meta-heuristics with the system
representativeness of simulation models. In order to explore the potential of coupling optimization and
simulation techniques in different job shop scheduling scenarios, this paper presents some of the ideas on
an ongoing research project developing an SBO strategy coupling genetic algorithm and discrete-event
simulation. Furthermore, this paper describes an approach to aid in the analysis of computed schedule
feasibility subject to stochastic behavior, which is the case for most of the real world industries. One of
the research goals is to provide an efficient and effective way to evaluate schedule robustness and to find
robust schedules. The research may significantly contribute to businesses where scheduling changes are
expensive, like in airline and train companies and automakers industries and suppliers.
Guilherme Ernani Vieira*
,1
, Mirko Kück**
,2
, Enzo Frazzon*
,3
, Michael Freitag**
,
***,
4
Evaluating the Robustness of Production Schedules using Discrete-Event
Simulation