Constructing service networks in liner shipping

Seaborne shipping is the most important mode of transport in international trade. After years of continuous growth, world seaborne trade volumes fell in 2009. However, signs of recovery can be observed, so seaborne trade is expected to recover. Current service networks should probably be reconsidered, because, for example, the financial crisis has caused an enormous rise in the oil price. In this thesis a model is proposed that can be used to construct new liner shipping service networks. This service network has to consist of a set of routes. For each route, the ship type and sailing speed used to serve the route have to be determined together with the cargo allocation on the route. First, a linear programming model is formulated that determines the optimal cargo allocation of a given set of routes. The performance of this model is compared to the performance of some heuristics. The optimal model proposed in this thesis performs on average more than 20% better than the heuristics. To determine the best service network, a genetic algorithm based approach is provided in this thesis. However, because the linear programming model that determines the cargo allocation has to be solved repeatedly, this approach will become too time consuming when the size of the problem is not decreased. Therefore, an aggregation method is described in this thesis that divides the ports in a few clusters. In this way, the method can be performed in a reasonable amount of time. However, the results have to be disaggregated after the linear programming problem is solved. Therefore, also a disaggregation method is proposed together with some improvement steps. The data used in this thesis are mainly based on the Asia-Europe tradelane of Maersk during spring 2010. Therefore, this method is used as a reference network in this thesis. Furthermore, an upper bound on the profit that can be obtained is calculated. The profit of the best network found in this thesis is 43% away from the upper bound. However, it is an improvement of about 40% compared to the reference network.