Presentation on theme: "1 UMR Lock 20 through 25 Simulation Model Inland Waterway Lock/Vessel Optimization Study Upper Mississippi River Locks 20-25 Center For Transportation."— Presentation transcript:
1 UMR Lock 20 through 25 Simulation Model Inland Waterway Lock/Vessel Optimization Study Upper Mississippi River Locks Center For Transportation Studies University Of Missouri, St. Louis 15 June 2005
2 The Need for a Simulation Model Why is a simulation model needed to evaluate alternative traffic management policies on the UMR? -The seasonality of traffic demands, vessel operations, and lock operations -The interdependence of individual vessel lockage times -The scope of the management measures under evaluation and their systemic impacts
3 The Bi-modal Distribution of Lockage Times at UMR Locks for
4 The Distribution of the Wait For Lock Service at UMR Locks for
5 The Seasonality Of System Use Total Lockages by Month at UMR Locks for
6 Seasonality Of System Use (Continued) The Number of Tows Using the System
7 Seasonality of the Wait For Lockage Time Distributions
8 Seasonality of Vessel Lockage Time Distributions
9 Seasonality of Non-Stop Pool Travel Time Distributions
10 Seasonality of Total Queue Sizes Locks 20 Through
11 Trend in Seasonality of Total Queue Sizes Locks 20 Through
12 The Simulation Model A discrete event simulation model of the segment of the UMR composed on Locks 20 through 25 and connecting pools is constructed using Micro Analysis and Designs Micro Saint Sharp. Micro Saint Sharp is a widely used, commercially available software package designed to build discrete event simulation models that facilitates model building and animation. Any user with a Micro Saint Sharp license may use and alter the simulation model. Simulation results may be analyzed in Micro Saint, any statistical package, and most spreadsheet applications.
13 The Simulation Model Vessels (large tows, small tows, and recreation craft) enter the system at one of ten entry points following seasonally estimated, independent inter- arrival time distributions. Vessels complete a lockage after system entry and then make a seasonally adjusted decision to: (1) continue to the next sequential lock in their direction of travel; (2) stop; or (3) re-configure their flotilla. If vessels stop or re-configure their flotilla, they are terminated in the appropriate pool after completing their lockage and then later regenerated in the pool in which they terminated. All recreation craft are terminated after a single lockage.
14 The Simulation Model Vessel lockage times depend on the vessel configuration, the direction of travel, the month of occurrence, and the state of the lock when the lockage occurs. Pool transit times depend on the vessel configuration, the direction of travel, and the month of occurrence. Periods of lock closure are modeled as independent occurrences with independent durations.
15 The Simulation Model Monthly and annual measures of system output and performance such as the categorized tow-miles produced, categorized utilized tow hours, categorized lockage times and utilizations, categorized lock delay times, and categorized pool transit times are recorded. The performance measures are analyzed using both Micro Saints built in analytical tools and SPSS.
16 Simulation Model Schematic Diagram Tow Traffic
17 Simulation Model Schematic Diagram Recreation Vessel Traffic
18 Simulation Model Detail Lockages There are 360 classes of lockages (lognormal distributions) at each lock characterized by: -Direction of vessel travel (upbound, downbound); -Class of vessel (multi-cut tow, single cut tow, jackknife, knockouts, and recreation traffic); -Lockage type (fly, turnback, exchange); and -Month of occurrence. Locks are periodically made not available to service vessels (exponential distributions).
19 Simulation Model Detail Vessel Traffic Seasonally adjusted independent entrances of four different types of tows at ten separate system locations (exponential distributions) Seasonally adjusted transition probabilities for directing each class of tow movement through the system Seasonally adjusted independent lock-specific recreation vessel arrivals (exponential distributions) Seasonally adjusted and directionally specific travel times for four separate tow classes through the lock pools (lognormal distributions)
20 Comparison of 100 Runs of the Simulation Model with the Omni Data
21 Comparison of 100 Runs of the Simulation Model with the Omni Data
22 Comparison of 100 Runs of the Simulation Model with the Omni Data
23 Results of 100 Simulations with Existing Traffic Management N Minimum (hours) Maximum (hours) Mean (hours) Std. Deviation (hours) Wait Time - All Vessels All Locks 10031, , , , Total Observable Tow Time , , , , Tow Time Large Tows , , , , Tow Time Small Tows 10057, , , , Tow Wait Lock , , , Tow Wait Lock , , , Tow Wait Lock , , , , Tow Wait Lock , , , , Tow Wait Lock , , , , Valid N (listwise)100
24 Results of 100 Simulations with an Example of a Locally Optimal Queue Re- sequencing Policy (Fastest First) N Minimum (hours) Maximum (hours) Mean (hours) Std. Deviation (hours) Wait Time - All Vessels All Locks 10029, , , , Total Observable Tow Time , , , , Tow Time Large Tows , , , , Tow Time Small Tows 10052, , , , Tow Wait Lock , , , Tow Wait Lock , , , Tow Wait Lock , , , , Tow Wait Lock , , , , Tow Wait Lock , , , , Valid N (listwise)100
25 Changes Resulting from a Locally Optimal Queue Re-sequencing Policy (Fastest First) N Minimum (hours) Maximum (hours) Mean (hours) Std. Deviation (hours) Wait Time - All Vessels All Locks , , , Total Observable Tow Time , , , , Tow Time Large Tows 100 3, , , Tow Time Small Tows , , , Tow Wait Lock Tow Wait Lock Tow Wait Lock , Tow Wait Lock , Tow Wait Lock , , Valid N (listwise)100
26 Vessel Re-sequencing Discussion Mean annual reduction of approximately 3,600 total tow hours required to complete the same set of vessel itineraries. This reduction represents approximately a 2% decrease in equipment time needed to complete the same set of movements through these five locks. Some vessels win and other vessels lose. System performance variability is also reduced.