(541g) Optimal Railyard Management for Site Operations in the Chemical Industry: Connecting Production to Outbound Trains

Bulk rail, both tank and hopper, is a common mode of transportation in the chemical industry. At a chemical plant, once the railcars are loaded with products, they are temporarily placed in a staging yard before being assembled into an outbound train and shipped to the customers. A typical single-ended staging yard consists of a single track that leads to the yard, which ultimately diverges to multiple tracks via several layers of switches. The distribution of railcars to tracks in a staging yard may depend on multiple factors, which may lead to situations where railcars containing a single product are scattered across the yard.

Given a list of orders to satisfy, railcars that contain ordered products are selected and assembled to several outbound trains, depending on customer locations. During the assembly process, the selected railcars are isolated from each track through a series of pull and push locomotive operations and ultimately marshaled into the outbound trains. Such selection and assembly processes are done manually, heavily relying on rules of thumb and established policies. Hence, an opportunity exists for such decisions to be made in a more systematic way resulting in an optimal set of car moves, which will lead to significant time and cost savings (e.g., fuel and labor).

Marshaling of isolated railcars to outbound trains, referred to as railroad classification, is a well-studied area in rail logistics since 1950’s. Several classification policies have been proposed, and formulas to estimate the time to complete a classification task have been developed (Beckmann et al. 1956; Daganzo et al. 1983). More recently, Dirnberger and Barkan (2007) improved the performance of such policies through bottleneck management methods. A thorough review on the theoretical aspects of railroad classification can be found in Boysen et al. (2012). Nonetheless, the aforementioned studies are limited in a couple of aspects. First, the scope of study does not cover the process of isolating the railcars from the staging yard which is equally as important in our case as marshaling them afterwards. Second, the policies do not necessarily guarantee the optimal sequence of marshaling operations.

In this work, we developed two mixed-integer linear programming (MIP) models to address the two problems respectively: (1) the selection and assignment of railcars to orders, and (2) the isolation and marshaling of selected railcars to outbound trains. The two models are solved in an iterative manner that allows us to simultaneously optimize the decisions of both problems. In the first model, railcars that are closer to the entrance of the yard are preferred to be selected, since such selections lead to reduced effort in the marshaling stage. Also, we propose the concept of railcar blocks (i.e., sets of consecutive railcars that contain the same product) to consider the fact that isolating the whole block requires nearly no additional effort when compared to isolating any single car within the block. In the second model, we optimize (i) the sequence in which the selected cars are isolated, (ii) which track each isolated car should be stationed before marshaling, and (iii) the sequence of marshaling operations. We allow for a wide range of locomotive operations, such as pulling/pushing of railcars and connection/disconnection of railcar blocks. We consider the layout and dimension of the yard (including tracks in between switches) to explicitly calculate the total distance the locomotive has to travel in order to complete the isolation and marshaling tasks.

Through case studies, we have shown that our method significantly reduces the time and cost of locomotive operations when compared to traditional methods. In addition, our method reduces the number of connections and disconnections between railcars required, an aspect that is favorable from the perspective of yard crews’ safety.

Reference

Beckmann M.J. et al., Studies in the economics of transportation. London, Oxford University Press, 1956

Daganzo C. et al., Railroad classification yard throughput: The case of multistage triangular sorting. Transportation Research-Part A. 1983, 2, 96-106

Dirnberger J.; Barkan C.P., Lean railroading for improving railroad classification terminal performance: Bottleneck management methods. Transportation Research Record. 2007, 1, 52-61

Boysen N. et al., Shunting yard operations: Theoretical aspects and applications. European Journal of Operational Research. 2012, 220, 1-14