Building a Rail Yard Digital Twin with AnyLogic Simulation – SimWell Builds a Rail Yard Digital Twin for Serious Gaming

The Need For a Rail Yard Digital Twin

A rail yard is a complex mix of railroad tracks and switches designed for the storing, sorting, loading, and unloading of rail cars. 

Operational efficiency in the rail yard directly affects the economic benefits of rail freight transportation companies while indirectly impacting numerous stakeholders in the supply chain line including supplier, client, etc.

A rail yard digital twin helps organizations to improve operational efficiency. 

The Challenge With Optimization

Optimization is often used in rail yards, but these models fail to capture the complexity of the yard. 

What makes the rail yard too complex for optimization?

The rail yard is dynamic, meaning it changes over time. Optimization does not consider time, so simulation is best suited to capture the dynamic aspects of the yard. 

The rail yard is stochastic, meaning there are a lot of variability in the system.

The rail yard is constrained by space.

Simulation captures the complexity of the rail yard including its variability and space components changing over time .

Dynamic Simulation with AnyLogic

Discrete event simulation, however, caters for these factors
by creating a rail yard digital twin which will allows the planning
and leadership teams to  try out and efficiently plan operations ahead of time. 

When a rail yard digital twin is used in planning, it improves current operations and ensures the rail yard can handle future challenges.

You should never have to worry about choosing the wrong option when making a decision inside your business.  We’ve developed that philosophy through hundreds of projects at SimWell. 

Here’s how we’ve built rail yard digital twin models with AnyLogic in the rail industry.

railyard digital twin built in AnyLogic

Rail Yard Digital Twin

This rail yard digital twin model helped our customer to plan their rail operations.  The ultimate objective is to develop a machine learning model to quickly make the best decisions in the yard.  To achieve that we are taking a multi-phase approach. 

First, we build a rail yard digital twin in AnyLogic Simulation.  The model is then used for serious gaming that allows planners to dynamically control the virtual rail yard to improve decision making in the real world. 

If you want to learn how to build models yourself, you can access our Complete Guide To Simulation that walks you through the step-by-step process to build a simulation model.

Second, we embed the heuristic decision logic used by planners to evaluate logic and compare scenarios.  

Finally, we create a machine learning tool by training an AI algorithm to find the best case scenario. 

Phase 1 - Serious Gaming

In the Serious Gaming phase, we develop a model to help planners with analysis through their rail yard digital twin.  

With a Serious Gaming model, planners have manual control over train movement on all tracks in the rail yard digital twin. They can move rail cars in commodity bulks or as single cars, dynamically couple and decouple a train, and switch/place/process commodities.

As show on the figure below, the interactive Serious Gaming model built by SimWell allows users to test these changes anywhere on the tracks.

Railyard Serious Gaming

The outputs of the Serious Gaming model can include any KPI important to the rail yard operation.  In this case, KPI’s include logging car info, location, processing start and end time.  KPI’s also include utilization of resources. For example, the figure below shows a workload chart highlighting busy/idle resource states over time. 

The serious gaming model is a low-risk, fast paced tool that can be used to train yard managers.

Simulation KPIs

Phase II - Heuristic

If a rail yard already has logic that operators use to process rail cars, we can build and imbed that logic in the model.

This will allow the rail yard planners to run scenarios and automate analysis.

Planners can test various scenarios and compare the performance of each decision, virtually, without interrupting yard operations.  They can also compare and improve heuristic using the simulation model.

Perform multiple runs with different scenarios requiring limited to no user Intervention

Phase III - Machine Learning

Once a large enough sample space of inputs and outputs has been produced from running the model, this data can then be used to train an AI Algorithm. 

A trained algorithm  can help us converge faster to a “best case” scenario that we can then validate using the simulation model.

SimWell is leading research to use machine learning in this manner.

Contact Us

SimWell is a consulting and software distribution company that specailizes in simulation, data science, and optimization. If you operate a rail yard and need help with a rail yard digital twin for serious gaming you can click here to schedule a call.

A video of the rail yard digital twin simulation model built with AnyLogic has been recorded.

Schedule A Call

Scroll to top