The Fall 1998 INFORMS
National Meeting was held in Seattle, October 25-28, 1998. For
more information, contact Bruce Patty, the cluster chair, at 415-491-1770,
fax 415-491-1251, or by e-mail at bruce@mcgi.com.
13:00 Network Modeling Needs in the Current Business Environment
Guest Speaker: R. Mark Schmidt, Assistant Vice President Strategic
Studies, Burlington Northern Santa Fe
14:15 Overview of Objectives, Architecture, and Technology of the
BNSF Network Simulation
- Todd McClain, BNSF
14:30 Simulation Model Development at Union Pacific
- Myron Lewellen, Union Pacific
14:45 Break
15:00 Comparison on Discrete Event Vs Discrete Time Approaches to
Network Simulation
- Warren Powell, Princeton University
15:15 Overview of Requirements
- Ajith Wijeratne
15:30 Group Discussion
16:45 Panel Discussion Summary
- Ajith Wijeratne
In recent years, railroads have debated the extent to which schedule
adherence affects asset utilization. A simulation model was used to examine
network performance under three operating strategies: schedule adherence,
flexible short-run scheduling, flexible operations. No strategy was best
in terms of all the major measures of cost, capacity,
and service. Schedule adherence provides better service, but requires more
assets than flexible operating strategy. For best results, the different
operating strategies need to be applied to different market segments.
MultiRail is an advanced railway operations planning and control system
used by service designers. Add-on's to MultiRail are being developed to
provide visibility into the downstream operations of the railway. This
allows the service designers to understand the impact of the operating
plan on the crew and capital assets. We discuss how MultiRail incorporates
locomotive, crew, and track asset management in its planning tools.
Railroads have traditionally addressed network-wide policy questions
by analyzing each major asset individually. A high level simulation model
of the railroad including all of the major assets reveals the interaction
between the assets. The model allows better strategic planning and quantification
of network decision policies.
Working for a major Canadian railroad, we developed an approach for
determining the number of locomotives that would be required under
different operating schedules in a planned operating environment. The model
generates key network results like locomotive utilization by class of power,
light moves, and expected train delay. Various parameters in the model
can be adjusted to reflect actual operational practices of locomotive management,
e.g., rates of required servicing at terminals, and minimum allowable inventory
levels at terminals.
This is the largest activity conducted by US railroads. We model the
Orin Line, an area producing one third of domestic coal. We tackle current
congestion problems on two levels: analytically for issues of capacity,
pricing, and cost allocation; and with discrete simulation for dispatching
and variability management.
We propose a B&C approach for assigning locomotives to trains. The branching decisions define facets of a restricted constraint set polyhedron. The approach improved the best known solution for an almost 2000-leg weekly problem faced by Canadian National, generating potential savings of more than $30,000,000 per year.
Models of the relationship between transit time, slot frequency, and
inter-siding spacing will be explored. In certain situations the slot frequency
can 'resonate' with the track layout to give smaller train delay.
Most models of rail capacity deal with line capacity; a few deal with
yard capacity; almost none deal with complex terminals. We will describe
TERMCAP, a spreadsheet model that is designed to estimate the capacity
of a terminal area where a half dozen mainlines converge upon a number
of classification yards, intermodal yards, and industrial support facilities.
Railroad mainline capacity is necessarily analyzed at train level. Yard capacity is usually analyzed separately in terms of blocks and cars. Identifying root causes of bottlenecks from a network perspective requires a common ground between mainline and yards. We present a simulation approach to this problem.