Estimating order delivery times and fleet capacity in freight rail networks: part I - simulation modelling

We consider an unscheduled rail network with stochastic demand for freight movement between stipulated origins and destinations. Freight orders arrive over time and each order is fulfilled as soon as a locomotive and a rake become available. After transporting an order to its destination, the locomotive and associated rake become available to move new shipments and the dispatcher must determine a suitable resource management policy at this juncture, i.e., whether to hold resources at the terminating junction, or deadhead them to be used at an alternate origin. The dispatcher must also determine if deadheading is to be undertaken reactively, after receiving a request from another station, or proactively, in order to pre-position rolling stock in anticipation of future demand. The amount of deadheading required also depends upon the system resource levels, e.g., rake and locomotive fleet capacities. The determination of the best fleet size and associated deadheading policy to guarantee a given service level (e.g., expected customer order flow time) is a complex operational problem. In this paper, we describe a simulation modelling approach to this problem. Our approach is illustrated using real data from the Indian Railway System, one of the largest freight carriers in the world.