The Canadian Pacific Railway Locomotive Resource Planning Model
Presentation to INFORMS
Seattle
October 26, 1998
A & L ASSOCIATES
Overview of Presentation
• Project background
• Model design
• Methodology and implementation
• Status
• Conclusions
The Canadian PacificService Design Vision
“Canadian Pacific Railway will have and use a structured methodology and state of the art
tools for designing, validating, executing, and refining a committed operating plan providing for consistently reliable, competitive service at
a low cost.”
MultiRail Implementation at Canadian Pacific
• Objective: “The MultiRail Project will provide CPR with a service design package to improve [the] operating plan as represented currently in MTP, CYards, and SMS.”
• Major tasks:– Development of service group blocking and train design
– Creation of unified operating design structure
– Validation of model and resource planning, including crews and locomotives
– Plan publication
CPR Requirements from the Locomotive Resource Planning Model
• Locomotives required, by type of locomotive and class of service
• Inventory of locomotives by type, location, and time
• Deadheading requirements
• Calibration and manual adjustment capabilities
• Summary statistics
Possible Approaches to Locomotive Resource Planning
(1) Aggregate planning– Requirements developed from HP-miles and HP-hours required
during modeling cycle
(2) Network scheduling– Requirements developed from definition of locomotive cycles
required to power all trains
(3) Combined approach– Requirements developed from locomotive inventories at key
terminals and balancing power in network by deadheading
Input Requirements for the A & L Locomotive Resource Planning Model
• Train schedules and expected tonnages
• Physical network characteristics
• Rules governing assignment of locomotives to trains
• Locomotive processing requirements and capabilities
• Rules governing balancing of locomotive supply and demand
Overview of the A & L Model
A & L AssociatesModel
(written in MS Access)
User
Adjustm
entsby C
PR
Number and type of required locomotives
Running inventories of locomotives at terminals and on the road
Summary statistics
Segment and schedule informationHP/ton requirements
Locomotive fleet characteristicsStandard locomotive consistsPower pool locationsOperational parameters
MultiRail Inputsfrom CP
Model Capabilities• Size of problem:
– 1,000 + trains over seven-day cycle– 2,400 + terminals– 1,000 + unit locomotive fleet comprised of 16
categories
• Time required:– Less than 10 minutes on Pentium class PC– Terminal hierarchy would be defined for each new
network
Model Capabilities (Continued)
• Scenario Options– Define ideal consists for trains
– Define “cycled” power
• Default Scenario Options– Locomotive categories and fleet assignments
– Default consist assignments
– Terminal network data
– Locomotive management inputs
– Substitution priorities
Outputs of A & L Model• Fleet requirements: locomotive units of different classes
required to operate plan
• Terminal activity: inventories throughout modeling cycle
• Fleet performance statistics: utilization, horsepower-hours, gross-ton miles, and productivity
Modeling Assumptions Regarding Pooling of Locomotives
• Units that come off inbound trains are serviced and then become part of a power pool.
• Outbound trains take units from the power pool as needed without regard to inbound trains.
• In cases of shortages of units, additional power must be sent to the terminal or certain trains must be under-powered, delayed, or cancelled.
• In cases of surpluses of units, fleet managers must decide whether units should be sent to terminals where they can be used more effectively.
Modeling Methodology (1)
• Default consists based on train category and terminal location are assigned for each train route segment where a locomotive consist is indicated in MultiRail.
• The number of units assigned is adjusted to reflect the minimum power requirements and manual overrides entered by the model user.
• Locomotive lineups are created for each terminal.
Modeling Methodology (2)
• Deadheading requirements are determined by aggregating weekly supply and demand of locomotives at terminals to identify surpluses and deficits.
• Locomotive demand is balanced over the CP network in a pre-arranged set of steps in which smaller and more peripheral terminals reconcile surpluses and deficits with their “source” terminals.
Calculating Network Locomotive Demand
• Classification of terminals into four levels:– Hump Yards
– Regional Yards
– IMS and Local Yards
– Other Terminals (e.g., customers)
• Categorization contained in MultiRail data on terminals
• Higher order yards are successively balanced with lower order yards associated with them based on proximity
Calculating Network Locomotive Demand Simplified Diagram of Stage 1—
Balancing Demand from Customer Yards into Local Yards
Winnipeg
St. Paul Yd
BensenvilleClearing Yard
Toronto Yard
Thunder Bay
Milwaukee
Moose Jaw
South Edmonton
Coquitlam Alyth Yard
Lethbridge
Sutherland
Blue Island
St. Luc
Selkirk
Lambton
Golden
Stinson Yd
Quebec
Kenwood Yd
Ottawa
Calculating Network Locomotive Demand Simplified Diagram of Stage 2—
Consolidation of Demand into CPR’s Hump and Regional Yards
Winnipeg
St. Paul Yd
BensenvilleClearing Yard
Toronto Yard
Thunder Bay
Milwaukee
Moose Jaw
South Edmonton
Coquitlam Alyth Yard
Lethbridge
Sutherland
Blue Island
St. Luc
Selkirk
Lambton
67
Calculating Network Locomotive Demand Simplified Diagram of Stage 3—
Consolidation of Demand into the Six CP Hump Yards
Winnipeg
St. Paul Yd
Alyth Yard
BensenvilleClearing Yard
Toronto Yard
Calculating Network Locomotive Demand Simplified Diagram of Final Stage—
Balancing Demand Among Calgary, Winnipeg, and St. Paul Hump Yards
Winnipeg
St. Paul Yd
Alyth Yard
Modeling Methodology (3)
• Taking into account the required light moves, terminal and road inventories are created for the modeling period.
• Initial inventories are adjusted to avoid excess deficits, taking into account servicing and maintenance requirements.
• The number of units required under an operating plan, as well as other summary statistics, can be obtained by summing over the terminal and road inventories.
Model Architecture
MMTRN
MMTRNSCH
MultiRail Input Tables Intermediate Tables Output Tables & Reports
Queries and Visual Basic Code
Queries and Visual Basic Code
Terminal locomotive transactions
Terminal locomotive transactions
Consist assignments
Terminal locomotivesurpluses & deficits
Terminal locomotive inventories
Required fleet size andsummary statistics
MMNDESUM
MMTRNRTV
Current Status
• Model has been completed and tested with inputs from CP Rail– Basic functionality has been demonstrated– MultiRail functions as effective source of
inputs
• Calibration and validation requires comparison to actual performance
Conclusions
• Use of knowledge of the network and of the problem made it possible to simplify the analysis into a one-pass process
• Several parameters facilitate calibration, but calibration still requires actual performance data that may not be readily available
• Speed and capabilities of personal computers with standard software now permit effective handling of “real world” problems