Montaj hattı dengeleme ve yetkinlik bazlı operatör atama problemi üzerine sezgisel algoritmalar

In this study, we focus on assembly line balancing and worker assignment problem, where multiple production lines are executed by a shared pool of workers with varying skills and having absenteeism issues. We developed two mathematical models for the real life situation at Vestel Electronics, which is the second largest TV manufacturer in Europe. Nonlinearity in one of the developed mathematical models is resolved through McCormick relaxations and both models provide optimal results for small instances in short time. However, it is seen that the models are intractable for the real life cases. In Vestel Electronics, worker-line and task-worker assignments should be determined by considering skill of workers. Due to the absenteeism problem, line balancing should be resolved at the beginning of each shift, which brings up a rigid limit on the solution generation time. To overcome this complexity, three heuristic approaches are developed. In first heuristic, master problem is decomposed into two sub-problems and constructing a multi-pass hierarchical solution framework. In Heuristic 2, solution of first heuristic is aimed to be improved. Once second sub-problem in Heuristic 1 is solved, tasks are reassigned to stations with the given worker assignment and algorithms are run iteratively until objective value no more improves. In Heuristic 3, workers are first assigned to lines in a balanced way in terms of task abilities of the workers with an algorithm. Then, master problem is solved with this input. Experimental results reveal that our decomposition approaches provide good enough solutions within the allowed time; encouraging Vestel Electronics' production team to implement the proposed approach as an online decision support system. Keywords: Multi Product Parallel Assembly Line Balancing Problem, Worker Assignment, Worker Skill and Absenteeism, Mixed Integer Nonlinear Programming, McCormick Relaxation, Decomposition Methods.