No abstract

<p style='text-indent:20px;'>A railway network is an indispensable part of the public transportation system in many major cities around the world. In order to provide a safe and reliable service, a fleet of passenger trains must undergo regular maintenance. These maintenance operations are lengthy procedures, which are planned for one year or a longer period. The planning specifies the dates of trains' arrival at the maintenance center and should take into account the uncertain duration of maintenance operations, the periods of validity of the previous maintenance, the desired number of trains in service, and the capacity of the maintenance center. The paper presents a nonlinear programming formulation of the considered problem and several optimization procedures which were compared by computational experiments using real world data. The results of these experiments indicate that the presented approach is capable to be used in real world planning process.</p>

No abstract

This paper investigates the overhaul maintenance scheduling problem in which the maintenance duration is uncertain at the time of planning. This problem involves specifying the dates of trains' arrival at the maintenance centre while taking into consideration the due windows, the desired number of trains in service, and the capacity of the maintenance centre. The cycle time of each type of trains is random with a known probability distribution. The objective is to minimise a weighted sum of two components: (i) the deviation of the assigned arrival dates from the due windows and (ii) the penalty for violating the resources' constraints. A combined genetic algorithm with sample average approximation solution approach is developed to solve this problem. The solution approach consists of a genetic algorithm for global search and an exact method to determine the arrival dates of train-sets when a sequence of train-sets is known. The results with data provided by one of the leading Australian maintenance center show that the proposed method can produce good solution within acceptable computation time.

No abstract