Factory level preventive maintenance in Turkish air force
In this thesis, we study the Factory Level Preventive Maintenance Problem (FLPM) experienced by Turkish Air Force (TUAF). This problem is a specific case of Nonpreemptive Resource Constrained Multiple Project Scheduling with Mode Selection (NRCMPSMS); allocation of limited resources to competing activities of multiple project of different types in which the duration of an activity is determined by the mode selection and the activity flow is dependent on the type of the project. The objective is to determine the start (finish) time and the mode of each project’s each activity so that the minimal total weighted tardiness and total incurred cost are obtained. We proposed a heuristic for this problem definition which is composed of two phases and apply it to a real life problem experienced by TUAF. In the first phase, the aim is to construct an initial schedule with minimum total weighted tardiness and in the second phase, this schedule is improved in terms of total incurred cost by the mode selection exchanges. Since the activity due date information is not available but required in prioritization of the activities, we develop five FLPM specific activity due date estimation methods. We run the proposed heuristic for three different weight figures which are determined by the Analytic Hierarchy Process and the one being used by TUAF. In addition, we study the influence of the release and the due dates of the aircrafts on the objective functions. We propose a determination method for each of the release and the due dates that aims finding the tightness levels of these two parameters. The release date determination method that we propose relates the arrival rate of the aircrafts with the utilization of the bottleneck resource whereas the due date determination method that we propose relates the due dates of the aircrafts with the fraction of the number of tardy jobs in percentages. We investigate the performance of the activity due date estimation methods in terms of the objective functions and the computational effort required by the tightness levels of the release and the due date that are found by the determination methods that we propose.