A job shop problem is the most common scheduling problem. It is very popular for research because of the problem's commonness and complexity. At first, when scheduling job shop problems, the researchers would minimise makespan, but in the last years total weighted tardiness is becoming much more popular.
Researchers usually focus on modelling the schedule with the help of disjunctive graph. This representation of the model is relatively well researched. Our goal will be to research modelling of the job shop problem with total weighted tardiness with the help of Petri nets.
We also know that for tardiness objectives both local search algorithms and simulated annealing have been reported very efficient. Typically, applying genetic algorithms to scheduling turns out to be very time consuming and if we do not formulate the problem in a right way, the results can be relatively poor in quality. In this work we will try to prepare the algorithm in such way, that we will get good results.
We will also research the effect of schedule type on the quality of the results. In the first part of the work we create only non-delay schedules, where the machine is not allowed to remain idle if there is a job waiting for processing. Those schedules return lesser results. In the second part we create schedules where the machines can stay idle for a certain amount of time. Optimisation of that kind is more time consuming, but gives better results.
|
