Experimental dataset for optimising the freight rail operations

The freight rail systems have an essential role to play in transporting the commodities between the delivery and collection points at different locations such as farms, factories and mills. The fright transport system uses a daily schedule of train runs to meet the needs of both the harvesters and the mills (An Integrated Approach to Optimise Cane Rail Operations (M. Masoud, E. Kozan, G. Kent, Liu, Shi Qiang, 2016b) [1]). Producing an efficient daily schedule to optimise the rail operations requires integration of the main elements of harvesting, transporting and milling in the value chain of the Australian agriculture industry. The data utilised in this research involve four main tables: sidings, harvesters, sectional rail network and trains. The utilised data were collected from Australian sugar mills as a real application. Operations Research techniques such as metaheuristic and constraint programming are used to produce the optimised solutions in an analytical way.


Specifications
A near optimal scheduler for trains was produced using a real sector of Australian rail network. Data source location Queensland University of Technology, Brisbane, Australia Data accessibility Data is within this article

Value of Data
The main aim of the presented data is to develop mathematical models of the freight rail systems and help in producing effective solutions in a reasonable CPU time.
In this research, minimising the makespan is proposed as a main criterion to optimise the freight rail systems using the introduced data. The results in this research can be used to compare the performance of the proposed mathematical methods in optimising complex systems such as rail systems in many prospective studies.
The data of the produced schedules of the train runs can be used for many different types of the freight systems such as the sugarcane or coal rail systems [5]. The data describe the daily trips of each train to deliver the empty bins at different locations called sidings and collect the full bins from these sidings for delivery to the mills or the factories.

Data
Based on the feedback from our industry partners, the data utilised in this research are created in four main tables: Sidings (Table 1), Trains (Table 2), Harvesters (Table 3) and Rail Network (Table 4). In addition, three figures are presented to show the main steps of the proposed solutions: Kalamia's mill with the main original map (Fig. 1), the main steps to produce the final solution (Fig. 2), and the daily trips of each train in the system (Fig. 3).

Experimental design, materials and methods
A case study was examined to validate the constraint programming models and metaheuristic techniques. Fig. 1 shows a sector of the transport system of Townsville's mill in Queensland, Australia. Many train runs are generated where each run start at one mill and finishes at the same mill after visiting many different siding locations. The number of trains was selected to implement different runs requiring a fewer number of trains. Kalamia's mill has 58 sidings located in 9 segments but not all of them work on the same day. Approximately 14 trains can be used to construct the train trips that deliver empty bins to sidings at farms and collect full bins from farms top sidings. The data table of sectional rail network was constructed to describe the rail section length between different sidings. Constraint programming (CP) is one of solution techniques to find a near optimal scheduler for the sugarcane rail systems. The proposed mathematical model considers the siding and train capacity constraints, daily allotment constraints of each harvester, train passing constraints where each train cannot occupy more than one rail section at a time or two trains can occupy one section at a time. Constraint programming that deals with problems defined within the finite set of possible values of each variable is the main technology used for solving mathematical formulation problems  120  120  22  22  22  22  2  SELKIRK  120  120  22  22  22  22  3  BURDEKIN  120  120  22  22  22  22  4  STRATHALBYN  120  120  22  22  22  22  5  DELTA  120  100  20  18  20  20  6  AIRDMILLAN  100  80  20  18  20  20  7  CHIVERTON  100  72  20  18  20  20  8  KALAMIA  110  82  14  12  14  13.3  9  BOJACK  120  120  30  30  32  30.6  10  CARSTAIRS  110  90  28  22  30  26.6  11  NORTHCOATE  110  90  28  28  28  28  12  JARVISFIELD  120  120  34  34  34  34  13  RITA ISLAND  120  120  34  34  34  34  14  KILRIE  120  120  34  34 34 34  through the search trees. Fig. 2 shows an example of four feasible solutions to clarify the stages of obtaining these solutions using the search tree for the DFS algorithm, where each solution is shown by three subgraphs that start with discovering the nodes of the search tree to find the solution. The search tree uses coloured nodes to express the node types. For example, the red nodes are the failures, the solutions are green, the blue nodes are the explored choice points, white are the nodes  [1][2][3][4]. The use of the Gantt chart has been proven as a useful tool to validate the solutions' applicability and to evaluate the algorithms' performance through the ACTSS Schedule Checker for Kalamia Mill. As shown in Fig. 3, the different numbers of trains are indicated by using different colours to satisfy the specific allotment for each siding during a day. The rail sections have been constructed on the vertical axis while the time of each trip had been shown on the horizontal axis. The red numbers on the graph show the number of delivered empty bins and green numbers show the number of collected full bins at each siding.