Flexible flow shop scheduling problem under the effects of learning and deterioration

In modern production systems efficiency is not only limited to the effective use of machines, but also the dynamic nature of human factors directly impacts production performance. Behavioral elements such as experience gained by employees over time (learning effect) and skill loss (deterioration effect) are taken into account to enhance the accuracy of scheduling studies. This study addresses a flexible flow shop scheduling problem that jointly considers learning effect and deterioration effect, an area with limited real life applications in the literature. Firstly a literature review was conducted based on the needs of the company where the study implemented. A mixed integer linear programming based mathematical model was developed incorporating both learning and deterioration effects with worker based sequence dependent setup times. The mathematical model was designed in accordance with the company's requirements and demands aims to the produce outputs that include both machine and worker assignments for the jobs. The model was tested on different datasets that varied in complexity. The tests were conducted on a flexible flow shop system with varying number of stages, four different types and rates of learning and deterioration effects, and a changing number of jobs. The test results show that the model can find a solution very quickly for simpler problems which have fewer stages and jobs with only learning effects. However, for more complex problems, with more stages, more jobs and both learning and deterioration effects, finding the best solution takes a long time or is sometimes not even possible in a reasonable time limit. In order to generate practical solutions for large scale and complex problems, such as the real life problem addressed in this study, a heuristic method based on the weighted earliest due date rule was proposed and applied to this problem. The proposed heuristic method effectively determines job sequences and worker assignments by taking job priorities and due dates into account. Thereby reasonable total weighted tardiness values can be achieved within reasonable computation times.