Modeling Rail Capacity





Rail Capacity Definition
Stringline Diagrams
Parametric Models
Train Performance Calculators
Train Dispatching Simulation
AASHTO SCORT
Sept 22, 2010
David Hunt of Oliver Wyman
Suggested Definition of Rail Capacity
The maximum number of trains that
can be moved between two locations
in a day without exceeding a
predefined level of service.
Elements in Determining Rail Capacity
Line Capacity
Yard
Capacity
Line Capacity: number of tracks; type
and spacing of control system; number,
spacing, and length of sidings; mix of
train types; operating and maintenance
plans
Yard Capacity: total acreage; number of
tracks; container storage slots
Crew
Capacity
Equipment
Capacity
Crew Capacity: available crew starts;
yard crews; maintenance crews
Equipment Capacity: locomotives;
railcars; containers/trailers
This presentation will focus on line capacity
Some of the Factors Determining Line Capacity

Physical Track Layout




Operating Plan



Schedules
Type of service
Train makeup



Number of tracks
Type of signals
Number and spacing of sidings
Number and horsepower of locomotives
Train length and weight
Geography



Mountains
Tunnels
Bridges
Maximum Versus Effective Capacity

Transportation firms can never utilize a facility 100% of
the time
 Maintenance
 Weather
 Peaking of traffic volumes
 Disruptions and recoverability
 Normal variability in operational conditions

Industry practices call for standards to maintain fluidity
of operations and avoid major issues at chokepoints


Useable (effective) capacity is 70% to 80% of the maximum
(theoretical) capacity
Utilizing the capacity buffer between effective and maximum
capacity results in deferred maintenance, reduced ability to react
to variability with increasing recovery time, significant reduction
in reliability
Stringlines



Graphical depiction of a timetable
Provides a visual representation of trains
scheduled to operate on a corridor
Typically show stations (space) along the y-axis
and time along the x-axis


Also referred to as time-space and time-distance
diagrams
Have many uses:
Identification of schedule conflicts (meets/passes)
 Identification of slots for new service
 Scheduling track maintenance
 Resource planning (crews, locomotives)

Example Stringline
Transnet Freight Railroad: Witbank-Komatipoort
Stringlines are a Primary Analytical Tool Used Worldwide
Example from Kazakhstan Temir Zholy
Green & Blue = Passenger, Red = Freight, Brown = Locals
Software Products Allow:
•Selecting timeframes,
corridors, and trains
•Building or adjusting
schedules by adding and
dragging strings
Levels of Effort in Modeling Rail Capacity
Planning
Models
Initial Estimation



“Back of the envelope”
methods
Expert with basic
knowledge of number
of tracks, type of
signals, and special
conditions (e.g.
mountainous terrain)
Useful for quick
assessment of a single
corridor or facility



Variety of methods,
requires more data
than back of the
envelope but less than
a full simulation
Parametric (statistical
based) models stem
from 1975 FRA work
and the CN model
(Krueger, 1999)
“Paper” simulations
(now evolved to
spreadsheets) are also
used for capacity
estimation
Simulation




Uses a commercial rail
simulation product
(RTC, RAILS,
FastTrack)
Requires precise
network layout
(tracks, sidings,
interlockings, signals,
etc.)
Requires knowledge
of operating plan:
trains that will be run
and schedules
Initial setup is
expensive
The AAR Approach to Modeling Rail Capacity
National Rail Freight Infrastructure Capacity and Investment Study





Requested by the National
Surface Transportation Policy
and Revenue Study
Commission
Commissioned by the AAR
Prepared by Cambridge
Systematics, Inc.
Purpose was to estimate the rail
freight infrastructure
improvements and investments
needed to meet the U.S. DOT’s
projected demand for rail
freight transportation in 2035
Used STB Waybill data, empty
car estimates, and ORNL
network attributes
AAR Study Recommended Levels of Service for Rail
LOS Grade
Description
A
B
Below Capacity
C
Volume/Capacity
Ratio
Low to moderate train
flows with capacity to
accommodate
maintenance and recover
from incidents
0.0 to 0.2
0.2 to 0.4
0.4 to 0.7
Near Capacity
Heavy train flow with
moderate capacity to
accommodate
maintenance and recover
from incidents
0.7 to 0.8
E
At Capacity
Very heavy train flow with
very limited capacity to
accommodate
maintenance and recover
from incidents
0.8 to 1.0
F
Above Capacity
Unstable flows; service
break-down conditions
D
Source: AAR “National Rail Freight Infrastructure Capacity and Investment Study”, September 2007.
> 1.00
Forecasted Growth of Freight Trains Per Day
2035 – Based on US DOT Freight Analysis Framework
Source: AAR “National Rail Freight Infrastructure Capacity and Investment Study”, September 2007.
Track Attributes Were used to Determine Capacity
Type of Control System
Number of Tracks
CTC=Blue, ABS=Green, Manual=Red
Two or More Tracks=Blue, Single Track=Tan
Source: Oak Ridge National Labs rail network. Raw data not verified for accuracy.
Capacity Tables Used for AAR Study
Practical Max
w/Multiple
Trains Types
Practical Max
w/Single Train
Type
16
20
1 ABS
18
25
2 No signal
28
35
1 CTC
30
48
2 ABS
53
80
2
CTC
75
100
3
CTC
133
163
Tracks & Control
1
No signal
Source: AAR “National Rail Freight Infrastructure Capacity and Investment Study”, September 2007.
Future Volumes Compared to Current Capacity
45% of Network at Level of Service E or F
Source: AAR “National Rail Freight Infrastructure Capacity and Investment Study”, September 2007.
Parametric Modeling

One form of a parametric model:
C = β0 x (1 + β1 N)α1 x (1 + β2 L)α2 x (1 + β3 S)α3 x (1 + β4 M)α4 x …


Where:
 C = maximum capacity in trains/day
 N = number of tracks
 L = type of control system (categorical variable)
 S = average spacing between sidings
 M = mix of train types
 αi, βj = coefficients
 … other parameters may be considered
Calibrate model using:
 Capacity information for selected lines obtained from the
simulation studies
Reporting Parametric Model Results Using AAR V/C

Establish the level of service for each rail line:
R=V/C

Where:

R = volume to capacity ratio, from which the level of
service (LOS) is determined

V = volume in trains/day
C = maximum capacity in trains/day


Identify potential chokepoints:


Lines with a LOS of “D”, “E”, or “F”, using the AAR
capacity scale
Other special considerations (e.g. tunnels, bridges)
Limitations of Parametric Models

The Canadian National parametric model is the
best known example
It was designed for single track corridors
 Does not handle complex track configurations




FRA has sponsored research into improving
parametric models for more complex track layouts
Parametric models are best used for high-level
national or regional modeling to identify potential
problem areas
Detailed capacity analysis is done with simulation
What is Computer Simulation?
•
Set of simplifying mathematical
assumptions that attempt to duplicate actual
train operations
•
The simulation should allow a comparison
of the current actual train operation with
alternative assumptions about:
 Changes in the number of trains
 Changes in the types of trains
 Changes in the schedules of trains
 Changes to the physical plant
Train Performance Calculator (TPC)
The Physics of Railroading

Objectives
Determine Minimum Run Time (unobstructed) on a
specific route for specific train characteristics
 Develop run time information for use as input to
dispatching simulation models
 Estimate fuel consumption


Uses



Provides key component analysis for schedule
preparation
Determines effect of line speed changes on run time
Determines effect of train consist changes
 Locomotive (tractive effort characteristics, etc.)
 Cars (weight/braking characteristics, etc.)
Train Performance Calculator

Inputs




Track profiles: Grades and curves
Speed limits
Typical train and locomotive consists
Outputs


Unhindered speeds and times
Fuel consumption
Train Performance Calculator
Graphical Output
Speed Limit
Actual Speed
e
nc
a
t
is
sD "
V
e
ine
Tim ringl
t
"S
Why do we use Train Dispatching Simulation Tools?

TPC’s only provide the expected operational
characteristics for a free-running train
without regard to:


Meets and passes with other trains
Capacity for schedule adherence
 Train priorities
 Train density and characteristic mix
 Physical plant
 Main track work
Train Dispatching Simulation Software

Main objective: a mathematical
approximation of operational results for a
given set of variables based on a reasonable
dispatching algorithm
 Logic attempts to realistically replicate decisions by a
“good” dispatcher (not an optimal one)

Other objectives:


Reliable, repeatable results
Ease of use
 Minimize or automate input data
 User-friendly input procedures
 Comprehensive user interface
 Externally usable results
Event-Based Simulation Dispatching Attributes

Dispatcher logic

Follows rules - preferences - dynamic priorities
 Might lower the priority for high priority train
running late
 Might raise the priority for low priority train if crew
close to exceeding hours


Variable - limited outlook
Anti-lock-up logic
All moves tested before implementation
 When move is rejected, next most preferable move
is selected

Dispatching Logic
Across each train’s look-ahead, conflicts are
identified
 If there are any, all “reasonable” resolution options
are identified
 Each option for resolution is “costed” using user
defined delay costs. These costs dynamical change
as the simulation runs.
 Penalties are added for less preferred moves such as
changing tracks, entering siding, etc.
 Options are sorted by “cost”
 Each option is submitted to the anti-lock-up logic
until one is accepted
 The necessary moves of that option are implemented
in the simulation

Train Dispatching Simulation
Types of Inputs


Infrastructure
 Plant
 Signals
Traffic
 Operating characteristics
 Which trains are running
 Train priorities
 Train size and power
 Route (including reverse moves if applicable)
 Time of operation
 Schedule based and non-schedule based
Special Routing Events

Replicate track maintenance
 Remove track/control point from service


Apply/remove temporary speed restrictions
Replicate train work at a location
 Train delay(s)
 Train characteristics


Train connections
Replicate passenger operation
 Schedule train(s) departure times
 Specify a route or track
Train Dispatching Simulation
Types of Outputs





Stringline (time-distance plot)
User configurable reports
TPC profiles
Track occupancy charts
Animation of simulation
Types of Reports

Individual train ”logs"

Scheduled & unscheduled delays
 In total
 By train type
 By location

Statistical analyses



Distribution of trip times
Locomotive-miles
Distribution of delays
 Used to determine if plant is balanced


Operating costs
Timetables
Animation of Results
Source: www.berkeleysimulation.com
Animation of Results – Yard with Industrial Leads
Source: US DOT Rail Capacity Workshop, 2002.
Track Occupancy Chart
Source: www.berkeleysimulation.com
The Iterative Planning Process
Using Computer Modeling

Develop “base case” traffic
 Track configuration, signals & other physical attributes
 Operating plan
Develop alternative scenarios (changes to physical
plant or operating plan)
 Compare alternative scenario to base, or to other
alternative

The Iterative Planning Process
Using Computer Modeling - 2

Inject track maintenance and operating failures at
critical points, and determine if plant still works.
 If not, refine plant some more and re-run
 First without perturbations
 Then with perturbations

Select alternatives that meet operating objectives
 If none, refine alternatives & re-run
 Balanced plant is achieved when delays are evenly
distributed
Limitations of Simulation Models






Data and time intensive
Must validate to actual
Yard operations are modeled separately
(hump operations, intermodal lifts, etc.)
Resource constraints (crews, locomotives,
etc..) are largely ignored
Models do not look beyond study area to
the rest of the network
Even detailed simulation requires
simplifying assumptions