Quantifying the impact of classification track length constraints on railway gravity hump marshalling yard performance with AnyLogic simulation

Abstract

As freight transportation demand increases worldwide, railway practitioners must carefully manage the capacity of existing facilities to ensure efficient and reliable operations. Railroad gravity hump classification (marshalling) yards, where individual railcars (wagons) are sorted into new trains to reach their destination, are an integral part of the freight rail network. Efficient operation of yard processes is critical to overall freight railway performance as individual carload shipments moving in manifest trains spend most of their transit time waiting for connections at intermediate yards, with more than half of this waiting time spent dwelling on classification bowl tracks. Previous research has developed optimal strategies to allocate bowl tracks to blocks for a given set of yard track lengths. However, these strategies make simple assumptions about the performance impact of over-length blocks due to a lack of basic analytical models to describe this relationship. To meet this need, this paper develops an original hump classification yard model using AnyLogic simulation software. A representative yard with accurate geometry and operating parameters reflecting real-world practice is constructed using AutoCAD and exported to AnyLogic. The AnyLogic discrete-event simulation model uses custom Java code to determine traffic flows and railcar movements in the yard, and output performance metrics. With complete flexibility to change track layout patterns, a series of simulation experiments quantify fundamental classification yard capacity relationships between performance metrics and the distribution of track lengths, as a function of the railcar throughput volume and size of outbound blocks created in the yard. The resulting relationships are expected to better inform railway yard operating strategies as traffic, train length, and block size increase but yard track lengths remain static.