Title: The development of a generic scheduling approach for the PLANWISE optimizer
Abstract: In this assignment, we focus on creating scheduling approaches for the optimizer in
PLANWISE, a scheduling system created by ORTEC that supports the user in scheduling
operations on different resources, such as machines and operators. The scheduling
approaches must be able to solve a scheduling problem specified by the optimizer. In other
words: The approaches must be able to create a schedule in which a set of operations is
scheduled on a set of resources.
The scheduling approaches have to meet two requirements:
1. The approaches should be generic and be able to handle different scheduling
situations that occur in practice.
2. The approaches need to be able to handle manual adjustments made by users. More
specifically: The approaches must be able to solve scheduling problems that contain
fixed operations. These operations have already been scheduled on a set of
resources and they may not be rescheduled: i.e. they may not be reassigned to
different resources and they may not obtain different starting and completion times.
To create scheduling approaches that meet these requirements, we first determine which
scheduling problem the approaches should be able to solve. After considering the scheduling
situations at different PLANWISE customers and potential customers, we create a general
scheduling problem that contains the constraints and objectives mentioned most often by the
PLANWISE customers. In the resulting scheduling problem, we have a set of jobs that each
consist of one or more operations. These operations each need to be scheduled on a set of
resources for their processing. The objective is now to schedule the operations on the
available resources such that the total tardiness of the jobs in the schedule is minimized. The
main constraints that we consider in our scheduling problem are general finish-start
precedence relations, unique resources, preemptive and non-preemptive downtimes, setup
times, routing constraints, resource-dependent processing times, and fixed operations.
Subsequently, we determine which scheduling approaches we use to solve our scheduling
problem. After performing a literature study on different scheduling approaches, we create
two approaches to solve the general scheduling problem: One approach based on an
integrated architecture and the regret-based random sampling algorithm and one approach
based on the hierarchical architecture and local search algorithms. These approaches are
able to handle the following constraints of the general scheduling problem: Finish-start
precedence relations, unique resources, preemptive and non-preemptive downtimes, and
setup times. The integrated approach is also able to handle routing constraints. In addition,
we expect that the approaches are also able to handle fixed operations.
We test both approaches on a set of scheduling instances and we subsequently compare the
performance of the approaches to each other. In this comparison, we use the integrated
approach as the benchmark approach, because ORTEC currently uses this approach in
certain PLANWISE optimizers. Based on the test results, we can make a number of
conclusions: First, on the tested instances the hierarchical approach generally performs
better than the integrated approach with respect to tardiness minimization. Among the
different versions of the hierarchical approach, we obtain the best results when we use
simulated annealing to solve the assignment problem. In addition, a good initial solution and
frequent iteration between the assignment and sequencing steps are also beneficial for the
performance of the approach.
4
However, improvements can still be made to both approaches with respect to both tardiness
minimization and calculation time. Based on test results, we are able to conclude that the
characteristics of a scheduling instance have a lot of influence on the performance of the
integrated approach. Specifically, the characteristics of the jobs in the scheduling instance
have a lot of influence on the decision sets created to choose a next operation to schedule
and the operation chosen from this decision set. Therefore, further research is required to
fine tune the integrated approach to the scheduling instance. Also, further tests need to be
done to determine the influence of the initial solution and communication scheme on the
performance of the hierarchical approach. Finally, both approaches need to be adjusted so
they are able to handle all constraints of the general scheduling problem: First, both
approaches need to be adjusted so they can handle resource-dependent processing times.
In addition, at present the approaches can only solve scheduling problems where all
resources have the same intervals of preemptive downtime. In practice, however, resources
often have different intervals of preemptive downtime. Finally,
Publication Year: 2008
Publication Date: 2008-01-01
Language: en
Type: dissertation
Access and Citation
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot