Written by: Dr. Edwin R. “Chip” Kraft, for Railway Age With a dynamic car scheduling system, railroads can improve train capacity utilization while improving service at the same time. What is a Scheduled Railroad? During the first decade of the 21st century, North American railroads universally embraced the concept of scheduled operations. Different railroads have had different interpretations of what scheduled railroading means. For example, CN has had a concept for adding unit train blocks into the carload network, but other railroads do not do this. Some railroads have tighter definitions of “on time” than do others. Regardless of exactly how individual railroads choose to implement it, the core concept behind scheduled railroading is to develop a realistic and achievable train operating plan for each day, and commit to execute that plan as closely as possible. The scheduled railroad concept does not require that every day be the same, but it does require that the Service Design and daily operations processes of a railroad be tightly integrated. This places greater demands on the Service Designers to produce realistic plans as compared to the former concept, where Service Design was less tightly coupled to daily operations. Execution issues are fed back to Service Design so the plan can be fine-tuned and made more realistic. As a result, the plan is optimized over time until it can be replicated in the field on a daily basis. Most railroads have directly applied the scheduled railroad concept to their intermodal and carload service networks. Some railroads, such as CSX, have scheduled unit trains as well. CSX’s Unit Train scheduling system is especially notable because of its flexible “booking” concept. Customers order trains and request times for delivery of empty cars. If the request cannot be fulfilled for any reason, the system suggests alternative feasible times. Then the delivery time of the train at the destination can be calculated. This actually is very similar to the way shippers work with trucking lines to schedule truck pickup and delivery appointments. It serves as a possible model for the railroad industry to follow in terms of the booking of railcar and intermodal freight, as well. What is the Problem with Demand Variability? Day to day variability in traffic challenges railroads to maintain disciplined train and yard operations. For example, if a train is twice as long today as it was yesterday, it may take twice as long to classify those cars over the hump in a yard. Longer trains on peak days may not fit in all passing sidings, resulting in dispatching delays and potentially late arrivals for other trains. For railroads, demand variability leads directly to process variability both in yards and line of road operations. This is the core issue that a flexible car routing and demand management system must be designed to address. The problem is that if railroads size their operating plans for an average day’s loading, some days there’s too much traffic leading to capacity overflows, missed connections, and congested yards—and service suffers. Other days, there aren’t enough cars to fill trains, so capacity goes to waste or planned trains get annulled. All these are direct impacts of traffic volume variability. Add to this the systemic variability associated with track maintenance curfews, signal failures, weather and even an occasional derailment and it is easy to understand the significant operating challenges that railroaders face each day. Given the negative impact that variability has on railroad operations, the literature on Statistical Process and Quality Control clearly explains why railroads should take charge of as many controllable variables as they can. As Deming’s well known “funnel experiment” shows, management’s control responses have the potential to amplify, rather than damp down process variability. The principles of scheduled railroading suggest that plan recovery is the appropriate response to disruptions. However the best way to recover schedule is not always clear. This is where a variety of Decision Support tools, including dynamic traffic management tools, may have the potential to help. Flexibility has Both Value and Risks While most railroads recognize the value of adhering to plan, some flexibility in managing day-to-day operations could be valuable for responding to variability in traffic demand. Flexibility is not a bad thing, if it leads to a better ability to accommodate customers’ needs and reduce costs. The issue is to find a way to add flexibility that focuses on plan recovery and which damps down, rather than amplifies the process variability that is inherent in railway operations. For example, scheduling everything can lead to an increase in train starts and crew expense if demand fails to materialize. By foregoing opportunities to cancel or consolidate trains “on the fly,” some railroads feel they may be missing opportunities for cost savings. But whether there’s a need for canceling or consolidating trains depends on how the operating plan is constructed in the first place: • If too much capacity is built into the plan, there will be very few capacity overflows, but lots of opportunities for canceling or consolidating trains. If a scheduled railroad is sized for peak loading, a lot of train capacity will likely go to waste every day. • If on the other hand the operating plan were sized conservatively, regular trains would always tend to run heavy, rather than run light. The operational focus shifts to the need for operating extra trains or second sections. Instead of having to determine which trains should be cancelled or consolidated, the problem becomes one of coping with capacity overflows. Running extra trains should not be the first management response. There are plenty of good reasons not to add trains on the fly, unless it is absolutely essential. To begin with, running only planned trains “meters” traffic across the system. Unless extra trains are operated, congestion can’t propagate from yard to yard. While running extra trains might solve a problem at one yard, it might simply shift the problem to the next yard. To ensure this doesn’t happen, a tactical operation planning process needs to understand yard and terminal capacity, as well as line of road capacity. Enough crews, locomotives and switching capacity must be available so a yard can classify all cars within the required time frame, without falling behind on its regular scheduled work. In congested situations, traffic metering may be needed to maintain the fluidity of the network. It is better to manage traffic flow within designated control parameters, than to accept extreme congestion and associated very high costs. What to do with those extra cars? In the past, the only way to attain operating flexibility was to change train operations, since car routing patterns were predetermined and fixed. However, a new way for gaining flexibility is by using computer algorithms to dynamically route rail cars to fill train capacity that has already been scheduled. In fact, algorithmic approaches to car scheduling were first proposed in 1970 for the Missouri Pacific Railroad, but the computer hardware and software at the time wasn’t up to the challenge. As a result, Missouri Pacific implemented fixed table-based blocking which remains standard to this day. Some railroads such as Norfolk Southern have replaced fixed table-based systems with upgraded software that uses algorithms to determine how best, on average, to route the cars. However routings are still determined independently of dynamic operating conditions, since the software that railroads are using today was only designed to replicate existing routing patterns, not to enhance them. With modern computing capabilities however, dynamic algorithmic systems have become so commonplace that we now take them for granted. For example, modern GPS systems take real-time traffic conditions into account when recommending the best way to go from point A to point B. In a congested network the best route isn’t always the same every day. Similarly, a dynamic car scheduling system could help fill train capacity and keep cars moving towards their destinations. It would do this by employing the same kinds of real-time revenue management algorithms that airlines use to control their bookings and ticket sales. The immediate benefit of such a system would be to improve train capacity utilization and help railroads match supply to demand, but it would also lead to a noticeable improvement in both service and car utilization. Railroads can and should be deploying this kind of technology for optimal railcar routing and scheduling. How a Dynamic Car Scheduling System Could Work If a preferred train is over capacity, but other trains moving in the same general direction still have room, a dynamic car scheduling system could change the car blocking to allow the extra cars to ride on those trains instead. It is true that not all car-rerouting alternatives make sense; so the option of holding a car until the next day’s train always remains available. A Dynamic Car Scheduling system could decide what to do based on the projected time and costs of each available alternative, as well as delivery time constraints associated with each rail car. These determinations would be made automatically by the computer system in a manner that is transparent to yard operations. For example, all yard operating and train makeup constraints would be respected by the system. Based on simulation testing that has been performed to date, it would appear that rerouted cars are often able to move in a different block/same day, rather than being held for the next day’s train. Advancing cars rather than holding them back improves the efficiency of both line-of-road and terminal yard operations. In conjunction with dynamic car scheduling, a tactical train planner would focus on where to add trains, rather than on cancelling already scheduled trains. Dynamic car scheduling uses a cost-based algorithm to route cars, so it can suggest to the tactical train planner (through its “shadow pricing” mechanism) where more capacity would likely provide the greatest economic benefit. In concept, dynamic car scheduling can find a set of trip plans that best utilize capacity that has already been scheduled. It adjusts traffic to fit the trains, rather than trying to adjust trains to fit the traffic. A dynamic car scheduling system would recognize capacity constraints in trains and yards, model yard connections in a realistic way, and find economical alternative routings to get cars to their destinations. It certainly makes sense to fill up existing train capacity first, before asking for additional capacity to be added to the system. A Dynamic Car Scheduling system would do this. As a result, such a system would sharply reduce the need for tinkering with the day-to-day train operating plan, since most of the day-to-day traffic variations can likely be absorbed within the train capacity that has already been scheduled. Stabilizing Demand and Optimizing Supply Chains Dynamic car scheduling along with tactical train planning could provide an effective approach to managing cars that are already moving on the railroad. However, over a longer term, railroads need to work together and with customers to improve demand forecasting for at least a 10-14 day planning horizon. Railroads can’t predict the number of cars that will arrive from connections, or effectively plan their own locomotive and crew assignments if their interline partners are still running irregular operations. Unless customers are willing to pay a substantial rate premium (as airline business travelers do) they can’t expect railroads to be able to provide reliable services if they can’t forecast their own needs, or control their own operations. Ultimately, the need to reduce demand variability (or at least improve demand predictability) can only be addressed by a collaborative effort for improving the way the whole supply chain operates. Dynamic car scheduling does not contradict scheduled railroading, but rather enhances it. It can improve train load factors without sacrificing operating discipline, since the technology can reduce the daily need for making tactical changes to the train plan. This will help railroads in the short term to improve operating efficiency while at the same time improving service levels. In the future, with additional improvements to yard process control systems as well as the development of a reliable demand forecasting capability, dynamic car scheduling would lay the groundwork for new “time definite” carload service offerings. Such services would help railroads replace lost oil and coal revenues because it would enable carload service to effectively compete with truck for certain industrial commodities. In the meantime, since dynamic car scheduling focuses on managing the cars that are already moving on the railroad, it can offer railroads an opportunity for immediate efficiency gains, as well as to lay the groundwork for a future quantum leap in the ability to manage the service delivery process. Dr. Edwin “Chip” Kraft, Director of Operations, TEMS (Transportation Economics & Management Systems, Inc.) has more than 25 years of experience in the railroad industry. Dr. Kraft has worked on both freight and passenger planning projects in prior positions at ConRail, CSX, Union Pacific, and Amtrak. He is a recognized expert in passenger and freight rail operations, simulation and project evaluation. Dr. Kraft is the author of numerous papers and articles on rail operations and line capacity modeling, and holds two U.S. patents for improved rail yard designs. His current project management responsibilities include the Midwest Regional Rail Initiative, Cleveland Hub Study, VIA Rail infrastructure and operations systems project and freight railroad capacity studies. Continue reading...
