Burlington Northern Santa Fe Corporation

1
Burlington Northern Santa Fe Corporation
The Future of Capacity Planning INFORMS Annual
Meeting
November 13, 2005
2
BNSF Service Design PerformanceCapacity
Management Multiple Levels

• Strategic capacity planning
• Service plan capacity planning
• Tactical capacity planning
• Capacity management challenges

3
BNSF Service Design PerformanceCapacity
Management

Planning Process Attributes Level Strategic
Capital Plan Horizon 90 Days to 20 Years
4
BNSF Service Design PerformanceStrategic
Capacity Planning - Process

• Maintain subdivision historical train counts/mix
  info
• Stay abreast of train routing changes
• Entertain requests from field Operations /
  Engineering
• Convert Marketing forecast plans into train
  forecasts
• Update 5 year plan rolling project list
• Identify alternative routes available (i.e.
  Raton, Newton)
• Place critical projects on top of funding list

5
BNSF Service Design PerformanceStrategic
Capacity Planning Volume/Demand Determination
Forecast Development
The Crew Forecast starts with Loaded Unit Miles
(LUM) calculated from Marketings O-D level
Forecast. Units are routed across BNSFs rail
network to determine loaded unit miles by
forecast group.
TUM / UPT Train Miles Train Miles / TMCS
Crew StartsTxD
6
BNSF Service Design PerformanceStrategic
Capacity Planning Volume/Demand Determination
Crew Forecast Key Inputs/Outputs

• Inputs
• Marketings origin/destination forecast of unit
  volume
• Service Designs productivity forecast
• Train size
• Routing
• Monthly historical relationships
• Empty unit mile to loaded unit mile ratio
• Forecast group to train group traffic mix (i.e.
  which commodities on which train types)
• Train group to division traffic mix (i.e. which
  train groups move on which divisions)
• Outputs
• Crew starts (by year and month, train group, crew
  district) for Capital Planning
• Crew starts (by month, by train group, by
  division) for Finance
• Crew starts (by month, by train group, by crew
  district) for Labor Relations
• Train productivity forecast (by month, by train
  group, by division)
• Measured by train size and unit miles/crew start

7
BNSF Service Design PerformanceStrategic
Capacity Planning Factors to Consider

• Number and length of trains
• Horsepower of trains
• Peak train counts (day of week and time of day)
• Terrain (grades and curves)
• Variance in train speeds and dispatching
  priorities
• Track speed, signal systems
• Terminal infrastructure
• Required service levels
• Maintenance of way windows

8
BNSF Service Design PerformanceStrategic
Capacity Planning Methods/Tools

• Historical (run time, train length, HPT) trend
  analysis
• Dispatch system replays
• Mainframe extracts of rail infrastructure
• List of dispatching priorities by territory
• Train performance calculators (fuel burns,
  unimpeded times)
• Train dispatching simulators (RTC)

9
BNSF Service Design PerformanceStrategic
Capacity Planning Example Territory

• Heavy intermodal
• High speed
• Rapid growth

10
BNSF Service Design PerformanceStrategic
Capacity Planning Example Territory
Southern transcon average daily train counts 3Q
2002 to 2004
11
BNSF Service Design PerformanceNetwork
Capacity Planning
Southern transcon constraint areas addressed past
10 years

• CTC (signal work) CA, AZ, NM, IL
• Double track TX panhandle
• Double track western OK
• Double track NM, west TX
• Terminal work Chicago, Kansas City, Clovis,
  Barstow, San Bernardino, Los Angeles

12
BNSF Service Design PerformanceStrategic
Capacity Planning Example Territory
Southern transcon 2004 train counts vs. line
capacity
13
BNSF Service Design PerformanceStrategic
Capacity Planning Example Territory
Southern transcon forecasted 2009 train counts
vs. line capacity
14
BNSF Service Design PerformanceStrategic
Capacity Planning Example Territory

Simulation of Design Alternatives (RTC dispatch
simulation)
15
BNSF Service Design PerformanceStrategic
Capacity Planning Example Territory
Southern transcon constraint areas addressed next
5 - 8 years

• Double track remaining single track segments NM
• Finish double track western OK
• Upgrade Mulvane Newton Ellinor route for
  effective double track route southern KS
• Triple track San Bernardino Summit (Cajon Pass)
• Triple track through Amarillo, TX terminal area
• Terminal work Chicago, Kansas City, Belen,
  Needles, San Bernardino, Los Angeles

16
BNSF Service Design PerformanceCapacity
Management

Planning Process Attributes Level Base Service
Plan Horizon 14 to 120 Days
17
BNSF Service Design PerformanceOptimizing the
Base Service Plan Major Drivers

• Several types of factors affect the design that
  is implemented
• Customer service requirements and behavior
• Rail network capability
• Connecting railroad interactions
• ROI considerations

18
BNSF Service Design PerformanceOptimizing the
Base Service Plan Customer Requirements

• Customer service requirements and operational
  capability drive the basic structure of our
  service plan
• Service requirements transit and switch window
• Customer facility capacity
• Demand variability
• Day of week, seasonal, mix-induced
• Demand forecast information

19
BNSF Service Design PerformanceOptimizing the
Base Service Plan Rail Network Capability

• BNSFs asset and resource base further dictate
  what type of service plan is implemented
• Route and terminal capacity
• Locomotive power
• Crew levels and availability
• Route limitations
• Siding length, bridge weight capacity, seasonal
  restrictions, etc.
• Equipment capacity
• Mix of products sharing the same resource base

20
BNSF Service Design PerformanceOptimizing the
Base Service Plan Connecting Roads

• The service plan must also accommodate the
  interdependencies of connecting railroads
• Coordinating interchange schedules and blocking
• Shared facilities terminal and line
• Haulage arrangements
• Reciprocal switching
• Interchange performance variability
• Timing and volume

21
BNSF Service Design PerformanceOptimizing the
Base Service Plan
Test Alternative Designs for Service/Cost Impact
Monitor Plan Fitness and Execution
Assess Factors Impacting the Service Plan
Implement Most Effective Plan

• Customer requirements
• Rail network capability
• Connecting railroads
• Cost considerations

• Rapid testing cycles
• Route alternatives
• Blocking/train options
• Balancing resources
• Velocity/consistency/ cost trade-offs

• Design system tied to operating mgmt system
• Tightly coordinate w/field
• Communicate! Communicate! Communicate!

• Automated monitoring
• Train, terminal, line performance
• Network demand
• Connection integration

22
BNSF Service Design PerformanceService Plan
Capacity Planning Enhanced Service Design (ESD)

• Starting assumption is that the mix of traffic on
  the network will vary continuously and the design
  toolset must be adaptive. ESD workstation
  capabilities
• Provide for a quick design cycle
• Utilize real-world traffic data to proactively
  test design impact
• Line segment balance
• Crew and locomotive consumption
• Terminal volumes
• Shipment velocity
• Incorporate optimization technology to balance
  service and economic drivers

23
BNSF Service Design PerformanceESD Proactive
Testing of Hypothetical Service Plans

• Developed the ability to rapidly prototype the
  impact of design changes on service and asset
  utilization.
• Using our network simulator (developed jointly
  with NuTech Solutions), Design Managers quickly
  test various solutions before implementing the
  most effective one.

24
BNSF Service Design PerformanceESD
Implementation / Functionality Plan

• Current functionality
• Design Scenarios are created on the mainframe
• Scenarios are run using BNSF server-resident
  network simulator
• Typical turn time for simulation is about 20
  minutes versus 4-6 hours on mainframe
• Scenario diagnostics are ported to spreadsheets
  for Design Manager review
• Scenario creation is the primary bottleneck
• Development targets for 2006
• Develop the capability to create pending yard
  blocks in Service Scheduling
• Build report-generator on mid-tier, including
  drill-down interactive capability (potential
  Corporate Dashboard tie-in)
• Adapt TTP scenario creation front end for use in
  ESD
• Bridge to port ESD output to Marketing
  Communication process

25
BNSF Service Design PerformanceCapacity
Management

Planning Process Attributes Level Tactical
Operating Plan Horizon Zero to 96 Hours
26
BNSF Service Design Performance Tactical
Capacity Planning (TTP)

• Easy access to operations/performance data
  requiring multiple queries in current environment
• Automatically identify exceptions while they can
  still be managed
• Utilize live data to proactively test possible
  tactical changes
• Line segment balance
• Crew and locomotive consumption
• Terminal volumes
• Shipment velocity
• Shipment performance impact
• Communicate scenario with stakeholders Command
  and Control protocol ensures local optimization
  doesnt sub-optimize the network
• Make it so capability once decision is made to
  proceed

27
Tactical Traffic Planner (TTP) Screenshot
Terminal Dashboard
28
Tactical Traffic Planner (TTP) Screenshot with
Drill Down
29
(No Transcript)
30
Tactical Traffic Planner (TTP) Screenshot Plan
Modification
31
BNSF Service Design PerformanceCapacity
Management

Capacity Management Challenges
32
Service Design PerformanceDrivers of Service
Variability
Sources of inconsistency occupy a spectrum of
controllability
Non-Controllable Weather Grade Crossing
Incidents Trespasser Incidents Customs Shipment
Inspections (USDA) Fuel / Other Resource Costs
How can consistency be maintained as buffer
capacity is consumed by base traffic growth?
33
BNSF Service Design PerformanceDemand
Variability Impacts
Daily Variation in Volumes for the H-GALTUL Based
on Shipment Entry into Network

• The base Transportation Service Plan (TSP)
  assumes a certain amount of predictability in
  traffic flows. If the same mix of traffic was
  offered by all shippers and connections week
  after week, life would be simple.
• But volumes and lanes vary continually. So, our
  base plan must be tactically modified on an
  hourly basis to support tendered volumes while
  intelligently managing our cost structure.
• Customers and railroads will both benefit when we
  find ways to smooth and predict traffic flows in
  advance. Can demand be demand be managed across
  multiple networks?

34
BNSF Service Design PerformanceMarket and
Terrain Sub-optimize Assets
Day of week volume variability, unbalanced market
flows and network characteristics lead to
imbalances in crew and locomotive resources.
Whenever possible, we attempt to balance
resource flows via design modification. However,
service requirements, capacity constraints,
terrain and other factors impose practical limits.
35
BNSF Service Design PerformanceHi Line Sub
Scheduled Operation
36
BNSF Service Design PerformanceHi Line Sub
Actual Operation
37
Service Design PerformanceOR / Decision
Support System Challenges

• Many sources of performance variability are
  exogenous
• Episodic events are the rule not the exception
• Volume growth is absorbing buffer capacity,
  exacerbating propagation effects at increasing
  rates
• Work load variability today cannot be reliably
  forecasted or controlled
• Fog of War capacity and demand is not known
  with certainty at some distance of space and/or
  time
• In this world, can you accurately model rail
  network operations, let alone optimize them?