Hybrid flow shop scheduling with learning effect under uncertain process times

Production scheduling is optimizing workflows by organizing and determining when and where a certain task will be completed. Scheduling problems can be classified by the type of production environment and scheduling for mass production systems is under the topic of flow shop scheduling. Hybrid flow shop scheduling is a special case in flow shop scheduling where in some steps of the system, production line is modified by increasing the number of resources. In addition to hybridity, it is not mandatory for parts to be processed in every stage of the system. Developing a schedule with the objective of minimizing the sum of parts' completion times is aimed. The real-life difficulties during the application phase are considered to be uncertainty of processing times of the new parts included in the system and the learning effect on processing times and both aspects are included in the problem. Learning effect on previously manufactured parts is analyzed and learning effect is seperated to obtain pure processing times. Processing times for new parts are then calculated from pure processing times with statistical learning methods and learning effect is included later. With the calculated processing times, a mixed integer linear mathematical model for hybrid flow shop scheduling with missing operations is proposed. Because of the size of the problem, it was not possible to obtain effective solutions with mathematical model for data sets above a certain size. The Genetic Timetable Gap Filling Algorithm has been developed with the aim of achieving better feasible solutions in a shorter period of time, especially for using in bigger data sets. In the developed heuristic algorithm; selection, crossover and mutation properties of genetic algorithm are adopted. The Genetic Timetable Gap Filling Algorithm simultaneously determines the order in which the jobs will be scheduled and the resources to which the jobs are assigned. Learning effect on process times is calculated based on parts produced in the past and parts that are scheduled before. The schedules obtained by assignments and rankings correspond to the individuals of the population and the sum of parts' completion times are determined as the fitness values of the individuals. Estimation accuracy of the processing times is clarified by statistical control methods and the similarity of the schedules created by the estimated processing times is examined with the schedules created with actual processing times. The effectiveness of The Genetic Timetable Gap Filling Algorithm is evaluated by comparing the solutions obtained by mathematical model and the algorithm.