Citation |

- Permanent Link:
- https://ufdc.ufl.edu/UFE0050084/00001
## Material Information- Title:
- A Software Tool for Freeway Travel Time Reliability Analysis Development and Testing
- Creator:
- Sun, Wei
- Place of Publication:
- [Gainesville, Fla.]
Florida - Publisher:
- University of Florida
- Publication Date:
- 2016
- Language:
- english
- Physical Description:
- 1 online resource (107 p.)
## Thesis/Dissertation Information- Degree:
- Master's ( M.S.)
- Degree Grantor:
- University of Florida
- Degree Disciplines:
- Civil Engineering
Civil and Coastal Engineering - Committee Chair:
- WASHBURN,SCOTT STUART
- Committee Co-Chair:
- YIN,YAFENG
- Committee Members:
- SRINIVASAN,SIVARAMAKRISHNAN
- Graduation Date:
- 4/30/2016
## Subjects- Subjects / Keywords:
- Demand analysis ( jstor )
Light ( jstor ) Rain ( jstor ) Shoulder ( jstor ) Software ( jstor ) Software development tools ( jstor ) Standard deviation ( jstor ) Travel demand ( jstor ) Travel time ( jstor ) Weather ( jstor ) Civil and Coastal Engineering -- Dissertations, Academic -- UF hcm -- reliability -- software -- traveltime - Genre:
- bibliography ( marcgt )
theses ( marcgt ) government publication (state, provincial, terriorial, dependent) ( marcgt ) born-digital ( sobekcm ) Electronic Thesis or Dissertation Civil Engineering thesis, M.S.
## Notes- Abstract:
- As traffic congestion continues to worsen, travel time reliability is receiving more attention as the appropriate performance measure for roadway facilities that regularly suffer from congestion. Motivated by its increasing importance in transportation planning and operation, various methods have been proposed to estimate and measure travel time reliability. Two recently developed TTR analysis methods are effectively the current standard, generally referred to as the "SHRP2-L08" and "HCM" methods. This document describes the development and testing of, a software tool to execute the large scale and highly iterative calculations of the SHRP2-L08 and HCM travel time reliability analysis methodologies. The software tool was developed with the C# language and the .NET Framework. A user guide for the software was also developed, which also serves as guidance for researchers to conduct the travel time reliability analysis. Through the application of the software, the generated results are also verified for consistency with the documentation of the two TTR analysis methodologies. ( en )
- General Note:
- In the series University of Florida Digital Collections.
- General Note:
- Includes vita.
- Bibliography:
- Includes bibliographical references.
- Source of Description:
- Description based on online resource; title from PDF title page.
- Source of Description:
- This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
- Thesis:
- Thesis (M.S.)--University of Florida, 2016.
- Local:
- Adviser: WASHBURN,SCOTT STUART.
- Local:
- Co-adviser: YIN,YAFENG.
- Statement of Responsibility:
- by Wei Sun.
## Record Information- Source Institution:
- UFRGP
- Rights Management:
- Copyright Sun, Wei. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
- Classification:
- LD1780 2016 ( lcc )
## UFDC Membership |

Downloads |

## This item has the following downloads: |

Full Text |

PAGE 1 A SOFTWARE TOOL FOR FREEWAY TRAVEL TIME RELIABILITY ANALYSIS : DEVELOPMENT AND TESTING By WEI SUN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUI REMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2016 PAGE 2 2016 Wei Sun PAGE 3 To my p arents PAGE 4 4 ACKNOWLEDGMENTS I am grateful to all the people that helped me along the way. I thank Dr. Washburn, my graduate instructor, for all the support and encouragement that help ed me obtain my academic achievements. I thank my parents for always being there for me and raising me into a decent man. I thank Dr. Yin, Dr. Lily and Dr. Siva for their great courses, which help ed me learn more about the transportation engineering field. I thank my peer students who have attended courses with me or worked with me, they made m y graduate life here memorable. PAGE 5 5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 8 LIST OF OBJECTS ................................ ................................ ................................ ....................... 10 LIST OF ABBREVIATIONS ................................ ................................ ................................ ........ 11 ABSTRACT ................................ ................................ ................................ ................................ ... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 13 Background ................................ ................................ ................................ ............................. 13 Problem Statement ................................ ................................ ................................ .................. 14 Research Objective and Supporting Tasks ................................ ................................ ............. 15 Document Organization ................................ ................................ ................................ .......... 15 2 OVERVIEW OF FREEWAY TRAVEL TIME RELIABILITY ANALYSIS METHODOLOGIES ................................ ................................ ................................ .............. 16 Overview ................................ ................................ ................................ ................................ . 16 Background ................................ ................................ ................................ ............................. 16 Basic Definitions ................................ ................................ ................................ .................... 17 SHRP2 L08 Methodology ................................ ................................ ................................ ...... 18 Stage 1 Base scenario generation ................................ ................................ .................... 19 Stage 2 Study period scenario generation ................................ ................................ ....... 25 Stage 3 Detailed scenario generation ................................ ................................ .............. 28 HCM Methodology ................................ ................................ ................................ ................. 29 Stage 1 Scen arios and DAFs ................................ ................................ ........................... 29 Stage 2 Weather adjustment factors ................................ ................................ ................ 30 Stage 3 Incident adjustment factors ................................ ................................ ................. 32 Stage 4 Overall scenarios ................................ ................................ ................................ 35 Performance Measures ................................ ................................ ................................ ............ 35 3 SOFTWARE TOOL DEVELOPMENT ................................ ................................ ................. 37 Brief Overview ................................ ................................ ................................ ....................... 37 Overview of Software Screens ................................ ................................ ............................... 37 4 VERIFICATION ................................ ................................ ................................ .................... 54 PAGE 6 6 Overview ................................ ................................ ................................ ................................ . 54 Input Data ................................ ................................ ................................ ............................... 54 Computational Documentation ................................ ................................ ............................... 54 Verification of Software Tool for SHRP2 L08 Methodology ................................ ................ 55 Verification of Software Tool for HCM Methodology ................................ .......................... 55 Comparison ................................ ................................ ................................ ............................. 57 5 SUMMARY ................................ ................................ ................................ ............................ 66 APPENDIX : COMPUTATIONAL DOCUMENTATION ................................ ........................... 68 LIST OF REFERENCES ................................ ................................ ................................ ............. 104 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 105 PAGE 7 7 LIST OF TABLES Table page 4 1 Detailed geometry ................................ ................................ ................................ .............. 63 4 2 Demand entry flow rate ................................ ................................ ................................ ..... 64 4 3 Overall results for SHRP2 L08 method ................................ ................................ ............. 64 4 4 Shoulder closure duration (min) statistics ................................ ................................ .......... 64 4 5 One lane closure duration (min) statistics ................................ ................................ ......... 65 4 6 Two lane closure duration (min) statistics ................................ ................................ ......... 65 4 7 Overall results HCM ................................ ................................ ................................ .......... 65 4 8 Reliability performance measure of two methods ................................ ............................. 65 PAGE 8 8 LIST OF FIGURES Figure page 3 1 Software components and relationship ................................ ................................ .............. 40 3 2 Process flow ................................ ................................ ................................ ....................... 41 3 3 SHRP2 L08 scenario generation methodology ................................ ................................ . 42 3 4 HCM scenario generation methodology ................................ ................................ ............ 43 3 5 Project properties ................................ ................................ ................................ ............... 44 3 6 Demand settings ................................ ................................ ................................ ................. 44 3 7 Weather settings ................................ ................................ ................................ ................. 45 3 8 Weather probabilities by DP for SHRP2 L08 ................................ ................................ ... 45 3 9 Weather events by month for HCM ................................ ................................ ................... 46 3 10 Weather event list for HCM ................................ ................................ ............................... 46 3 11 Weather event chart for HCM ................................ ................................ ............................ 47 3 12 Incident settings ................................ ................................ ................................ ................. 47 3 13 Incident adjustment factors form ................................ ................................ ....................... 48 3 14 Incident adjustment factors for HCM ................................ ................................ ................ 48 3 15 HCM incident event list ................................ ................................ ................................ ..... 49 3 16 HCM incident event chart ................................ ................................ ................................ .. 49 3 17 Base probabilities for SHRP2 L08 ................................ ................................ .................... 50 3 18 Study period probabilities for SHRP2 L0 8 ................................ ................................ ........ 50 3 19 Scenario event numbers for SHRP2 L08 ................................ ................................ ........... 51 3 20 TTR scenarios listing ................................ ................................ ................................ ......... 51 3 21 TTR scenarios results ................................ ................................ ................................ ......... 52 3 22 Individual scenario results viewer ................................ ................................ ..................... 52 3 23 TTR overall results ................................ ................................ ................................ ............ 53 PAGE 9 9 3 24 TTR overall results charts ................................ ................................ ................................ .. 53 4 1 Example freeway facility ................................ ................................ ................................ ... 59 4 2 TTI distribution of SHRP2 L08 method ................................ ................................ ............ 59 4 3 Incident severity distribution ................................ ................................ ............................. 60 4 4 Incident start time distribution ................................ ................................ ........................... 60 4 5 Incident location distribution ................................ ................................ ............................. 60 4 6 Shoulder closure duration distribution ................................ ................................ ............... 61 4 7 One lane closure duration distribution ................................ ................................ ............... 61 4 8 Two lane closure duration distribution ................................ ................................ .............. 61 4 9 Weather start time distribution ................................ ................................ ........................... 62 4 10 TTI distribution of HCM method ................................ ................................ ...................... 62 4 11 Difference of 95 th % TTI values between two methods ................................ .................... 63 PAGE 10 10 LIST OF OBJECTS Object page 3 1 User Guide (.pdf file 12,396 KB) ................................ ................................ ...................... 38 PAGE 11 11 LIST OF ABBREVIATIONS AP Analysis Period AADT Annual Average Daily Traffic CAF Capacity Adjustment Factor DAF Demand Adjustment Factor DP Demand Pattern FFS Free flow Speed HCM Highway Capacity Manual PTI Planning Time Index RRP Reliability Reporting Period SP Study Period SAF Free flow Speed Adjustment Factor SHRP2 Strategic Highway Research Program, Second Funding Implementation TTI Travel Time Index TTR Travel Time Reliability VMT Vehicle Miles Traveled VHT Vehicle Hours Traveled VHD Vehicle Hours of Delay PAGE 12 12 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science A SOFTWARE TOOL FOR FREEWAY TRAVEL TIME RELIABILITY ANALYSIS : DEVELOPMENT AND TESTING By Wei Sun May 2015 Chair: Scott S Washburn Major: Civil Engineering A s traffic congestion continues to worsen, t rave l time reliability is receiving more attention as the appropriate performance measure for roadway facilities that regularly suffer from congestion. Motivated by its increasing importance in transportation planning and operation, various methods have been proposed to estimate and measure travel time reliability (TTR) . Two recently developed TTR analysis methods are effectively the current standard , T his document describes the development and testing of , a software tool to execut e the large scale and highly iterative calculations of the SHRP2 L08 and HCM travel time reliability analysis methodologies . The software tool was developed with the C# language and the .NET Framework . A u ser guide for the software was also developed , which also serves as guidance for researchers to conduct the travel time reliability analysis . Through the application of the software, the generated results are also verified for consistency with the docume ntation of the two TTR analysis methodologies. PAGE 13 13 CHAPTER 1 INTRODUCTION Background When it comes to travel ing, the reliability to arrive at a certain destination safely and on time is a major concern for both the travelers and the transportation system managers . C ongestion leads to more variable travel time s for a facility , which results in travelers need ing to plan extra travel time to account for the potential unreliability of arriving at their destination on time in any given trip . The concept of t ravel time reliability is based on the analysis of a large number of trips taken over a given length of facility over an extended period of time (i.e., a travel time distribution) . Other definitions of travel time reliability in clude the consistency or dependability in t ravel times from hours of the day, days of the we e k, and months or seasons of the year. T ravel time reliability is also known as the variability between the expected travel time and the actual travel time. In general, t ravel time reliability is an indication of how often congested conditions occur , and how severe the differences between expected and actual travel time s become . T raffic congestion can be caused by various reasons, such as heavy demand relative to regular capacity , closing of road lanes due to incidents and severe weather events . Correspondingly, travel time reliability is then affected by high demand /capacity ratios , weather event s and incident event s . In recent years, the significance of travel t ime reliability in transportation system planning and operation has been recognized. The application of travel time reliability in the transportation field involves: operational planning, congestion management, demand prediction, performance evaluation, an d system optimization. PAGE 14 14 Considerable research ha s been conducted recently to analyze travel time reliability . Two recently developed TTR analysis methods are effectively the current standard: one based on the results of a Strategic Highway Research Program (SHRP research project (hereafter referred to as SHRP2 L08 L08 methodology that will be included in the forthcoming sixth edition of the Highway Capacity Manual (HCM), hereafter The SHRP2 L08 and HCM TTR analysis methods are overall similar, but the latter applies some revisions to reduce the overall comput ational burden. An overview of both methods, as well as an identification of their specific differences is discussed more fully in Chapter 2. Problem Statement One thing common to both TTR analysis methods is that they are data intensive, and consequen tly computationally intensive. They both require a large number of scenarios (i.e., combinations of input conditions) to be processed with a complex facility analysis methodology (either a freeway or urban street). While the SHRP2 L08 project report and HCM do cument the respective analysis methodologies, it is not feasible to conduct the travel time analysis methods without software, given the large scale and highly iterative calculations . Furthermore, given the complexity of applying these TTR analysis methods , there are considerable demands on the software in terms of user friendliness and functional capabilities. Prior to the development of the software created for this project, the only existing software implementations of these analysis methodologies were in the form of an Excel spreadsheet, which had significant limitations in both of these areas (i.e., user friendlines s and functional capabilities). PAGE 15 15 Research Objective and Supporting Tasks The objective of th is project is to develop a software tool for exe cuting the SHRP2 L08 and HCM TTR analysis methodologies (but specific to just freeway facilities), as well as another methodology that wi ll be introduced in Chapter 3. Furthermore, this software tool must be of that is, be able to meet the user friendliness and functional requirement standards of transportation practitioners such that the application of these TTR analysis methodologies will be as intuitive and efficient as possible. The following task s were conducted to support the accomplishment of this objective: Review of freeway travel time analysis methodologies . Develop the software tool for the travel time analysis methodologies. As part of this task, obtain input on software design features from stakeholders. V erify that the analysis methodologies have been correctly implemented into the software. As part of this task, document methodology calculation processes that were not identified in the original report. Develop a user guide for how to operate t he software tool for each o f the methodologies . Document Organizati on In the remainder of the document , C hapter 2 contains an overview of freeway TTR analysis methodolo gies . Chapter 3 provides an overview of the software development and process flow, as well as the process flow of d ifferent scenario generation methodologies. Chapter 4 conducts the verification of the software tool, and provides the computational documents for both SHRP2 L08 and HCM methods in Mathcad. Chapter 5 provides a summary of this project . PAGE 16 16 CHAPTER 2 OVERVIEW OF FREEWAY TRAVEL TIME RELIABILITY ANALYSIS METHODOLOGIES Overview This chapter provides an overview of the freeway TTR analysis methodologies. First, a background of TTR analysis methodologies is provided . Second, the procedures of the SHRP2 L08 methodology , which are documented in the 5 th edition of the HCM (TRB, 2010 , are presented. Third, the procedures of the revised SHRP2 L08 method , which will be included in the forthcoming sixth edition of the HCM (TRB, 2016) , and referred to as the HCM method, are presented . Background T ravel time reliability relies on the scenario , a specific combination of inputs to the freeway facility analysis methodology, generation methods to enumerate a sufficiently complete set of scenarios that represent the variability o f demand levels, weather events, and incident events over a certain long term time period (e.g., a year) . A desirable sce nario generation method is one that produces a set of scenarios that will yield results that match the field conditions with reasonable accuracy, but also keeps the number of scenarios to a minimum because of the highly iterative and complex nature of the calculation process. The SHRP2 L08 method use s a deterministic approach to generate scenarios . In order to model every possible combin ation of demand leve ls, weather events and incidents , the deterministic approach generate s a fairly large number of scenarios. In this case , a series of assumptions are used to simp lify the combinations, such as: T he start time period for weather or incident events can only be the be ginning or middle time period of the study period . T he location of incident events can only be the first, middle or last segm ent of the studied facility . PAGE 17 17 T he occurrence of multiple incidents is substituted by a single inci dent with a long duration . Even though the application of the simplifying assumptions above may reduce the otherwise numerous number of scenarios, this deterministic approach could still lead to some limitations : The relatively large amount of scenarios to be generated , which may increase the runtime . The complexity of the scenario generation about the start times, durations and locations assignment, also possible bias may be resulted from the assumptions . The lack of consistence with the field observations for the tail o f the travel time distribution . The HCM method was proposed by the NCHRP 3 115 (HCM major update) research project team , which was meant to serve as an enhancement to the SHRP2 L08 method for freeway travel time reliability scenario generation . mainly to obtain more realistic scenarios that match the field observations , while reduc ing the number of scenarios generated. The highlight of this scenario generation me thod is that it combines the deterministic approach with a more stochastic approach (Monte Carlo methods). The method uses a deterministic way to generate a fixed number of scenarios, then applies the random assignment to each scenario generated, by pickin g random numbers based on the specific distribution. Basic Definitions Following are some basic definitions for the terms used in the travel time reliability methodologies: Analysis period (AP) or time period, is the time interval for one single applicatio n of the travel time reliability methodology, the basic facility conditions must stay unchanged, usually 15 min in duration. PAGE 18 18 S tudy period (SP) is certain time interval in a day, which consists of one or more analysis periods, usually 1h to 6h in duration. Reliability reporting period (RRP is the time over which the travel time reliability is measured , usually 1 year in duration . Demand pattern (DP) represents a group of days within the RRP that have similar demand level . Demand adjustment factors (DAF) represent the variability of demand through days in week or months in year , which are used to adjust the original demand in the base file. Capacity adjustment factors (CAF) represent the variability of capacity caused by weather or incident events , w hich are used to adjust the original capacity in the base file . Free flow speed adjustment factors (SAF) represent the variability of FFS , which are used to adjust the original FFS in the base file . Scenario is one single realization of the study period, w hich contains the relative demand level, weather and incident event information , e.g. demand pattern 1, medium rain start s from AP 5 with a duration of 30 minutes, shoulder closure on segment 3 start s from AP 2 with a duration of 45 minutes . Demand multiplier is the ratio of demand for the specified day and month to the AADT used . A n nual average daily traffic ( A A DT is the total volume of vehicle traffic for a year divided by the number of days in that year. SHRP2 L08 Methodology The SHRP2 L08 methodology uses a deterministic approach to generate scenarios that can represent conditions such as demand level, weather and incident events . First, a set of base scenarios with their initial probabilities are generated . The base probability represent s the portion of time under the specified condition during RRP. Second , the probabilities of the base scenarios are adjusted to the study period probabilities, which represent the portion of time under the specified conditions in the study period. Then, an expected travel time can be estimated for each scenario based on the relative DAF, CAF and SAF . Finally, t he calculated travel times from all PAGE 19 19 scenarios with their probabilities can be compiled into a travel time distribution , which will be used to evaluate the travel time reliability of the studied facility . An overview of the SHRP 2 L08 analysis methodology , which is described in more detail in the SHRP2 L08 working paper 1) follows. Stage 1 Base scenario generation When generating the base scenarios , the objective is to generate a set of scenarios that combines the demand levels, weather events, and incidents conditions . Step 1. Create the base fi le Base file contains the basic facility information, such as facility geometry ( segment type, number of lanes , etc. ), terrain, jam density, capacity, truck percentage, demand entry flow rates, FFS , AP, SP, RRP, etc . Step 2 . Configure demand patterns To represent the variabilities of demand in days of week or months of year, demand patterns are proposed, each demand pattern represent a certain demand condition. The number of demand pattern is the number of day groups multiplied by the number of month groups. For example, demand patterns usually consist three weekday groups (Monday and Friday, Tuesday to Thursda y and weekends ) and four month groups (seasons), which will give us twelve demand patterns. Each scenario contains o ne demand pattern, demand level of the demand pattern is determined by the demand multipliers. The demand multipliers represent the variability of demand in days of week or months of year. The demand multiplier for a demand pattern is the weighted demand multipliers based on the day groups and month groups of t he demand pattern. The base file demand multiplier is the ratio of base file demand to AADT. PAGE 20 20 DAF of the demand pattern (scenario) equals to the demand multiplier of that scenario divided by the base file demand multiplier . rL 2 1 is the demand adjustment factor associated with scenario s analysis period tp and segment seg is the demand multiplier associated with scenario s is the demand multiplier associated with the base file for analysis period tp The p robability of a demand pattern is the duration (min) of the demand pattern in the RRP di vided by the total RRP duration. is defined as the probabilit y of demand pattern N , and is computed from the following: rL 2 2 Step 3 . Configure weather data Usually there are eleven weather categories: medium rain, heavy rain, light snow, light medium snow, medium heavy snow, heavy snow, severe cold, low visibility, very low visibility, minimal visibility, non severe weather (Normal). Th e weather data should be collected and classified into the weather categories above. The probabilities of these weather types are stated by month. Weather data for each weather type include probability, duration, and adjustment factors. For data rich envir onments, analysts may estimate the probabilities of weather types from the following: rL 2 3 PAGE 21 21 If analysts do not have access to the detailed local weather data to estimate the weather probabilities, they can use the 10 year average weather probabilities of the nearby metropolitan areas in HCM . The SHRP2 L08 assumes that a weather event occurs either at the start of the study period or in the middle of the study period with equal probability. In this case, considering 11 weather types and 2 possible start times, there will be 22 weather scenarios . Adjustment factors are assigned to the analysis periods from the weather start time to the end time based on the start time period and weather duration, and weather affects all segments of the facility. Since wea ther types are mutually exclusive, if two or more weather types are generated at the same time period, the weather event is assigned to t he weather type with the greatest capacity reduction effect. Step 4 . Configure incident data The incident data must be collected and classified into one of the following incident categories : no incident, shoulder closure, one lane closure, two lane closur e, three lane closure, and four or more lane closure. For data rich environments, the time based probability of inciden t type i in month j is estimated by the following : rL 2 4 If analysts do not have access for local incident data to directly estimate t he incident probabilities, they can use local incident rates or crash rates to obtain the inc ident probabilities. The incident probability for incident type i of month j can be calculated from: rL rF rF rH rH 2 5 is the mean duration of incident type i PAGE 22 22 is the incident distributio n probability of incident type i is the expected incident frequency of month j The distribution of incident severities can be specified by the analyst, default distribution is as following : rL rL rL rL rL rL i is the incident severity type, 1 = shoulder closure, 2 = one lane closure, 3 = two lane closure, 4 = th ree lane closure, 5 = four or more lane closure. The expected frequency per study period in month j can be calculated using : rL rH 2 6 is the vehicle miles traveled (VMT) in the base file is the incident rate per 100 million VMT f or month j is the w eighted average demand multi plier of the all days in month j relative to base demand multiplier If incident rates are not available directly, analyst can convert the crash rates into incident rates using the following equation : rL rH 2 7 is the crash rate per 100 million VMT f or month j is the local incident to crash ratio If crash rates are not available directly, analyst can use the HERS model to ca lculate crash rates based on the following equ ation : PAGE 23 23 rL rF rH rE rH rF rH rH rH 2 8 is the lane width, usually 12 ft ACR is the average crash rate, can be calculated from the following equation: rL rH 2 9 is the two way hourly capacity is the average number of lanes for all segments of the facility, can be calculated using the following equation: rL rH f8 f8 f, f8 f8 f, 2 10 n is the number of segments of the facility is the number of lanes for segment i is the length of segment i The method assumes that incident can only happen at three locations of the facility (first, middle or last basic segment of the facility) and two start times (beginning or middle of the study period), so the maximum number of incident scenarios should be : 2 start times 3 (incident durations 3 (incident loc ations 5 incident severities + 1 no incident ) = 91 incident scenarios . Adjustment factors of the incident will be assigned to the relative time periods and location s . Step 5. Overall base scenarios If we consider all the possible scenarios , a maximum number of scenarios of : 12 (demand patterns 22 weather scenarios ) 91 incide nt scenarios = 24,000 scenarios will be generated . These scenarios are called base scenarios. The basic assumption of base scenario is that contributing factors su ch as demand patterns, weather events and incident events are PAGE 24 24 independent. Thus, the probability of a base scenario is the product of the probability of all contributing factors using the following equation: rL rH rH 2 11 However, the probabilities of weather or incident are given on a monthly basis, the probabilities of demand level are by demand pattern. or incident probability must first be aggregated across the demand pattern and then used in Equation 2 1 to calculate the base scenario probability. The aggregation of weather and incident probabilities across demand patterns are based on the following equations . In the equations below, i refers to a demand pattern, j refers to a weather type, k refers to an incident type, m refers to a month. rL rH 2 12 is the weather probability of weather type j and month m is the number of days of demand pattern i and month m in RRP rL rH 2 13 is the incident probability of incident type k and month m Equation 2 11 can be rewritten in the form of Equation 2 14 : rL rL rL rL rH rH 2 14 In this case, there could be some base scenarios with very low probabilities, a threshold is set to remove these scenarios . Base scenario with a probability lower than the threshold is removed and its probability is assigned to the rest of the scenarios proportionally. User can specify the value of the threshold, which is defaulted as 0.1%. The threshold can reduce the total number of scenarios generated. However, a large value of thr eshold is not PAGE 25 25 recommended, since that would result in a s ignificant loss of scenarios, which will affect the accuracy of travel time distribution. Stage 2 Study period scenario generation The base scenarios describe the time the facility will be under the specified condition during the RRP . However, they need to represent the time the facility will be under certain condition during the study period . So the weather or incident event duration s are considered to adjust the base scenario probabilities in to the study period scenario probability . The computational procedures to convert probabilities of base scenarios into the study period probabilities are as following (on a demand pattern basis : For the selected demand pattern, the base scenario probabilities ar e c lassified into four categories: Category 1, demand only Category 2, demand and weather Category 3, demand and incident Category 4, demand, weather and incident Note that the sum of probabilities of all scenarios in this demand pattern remains the same after the following adjustment procedures : 1) Compare the weather and incident events durations Since the SHRP2 L08 method uses 15 minutes as the analysis period, weather and incident durations are rounded into the nearest 15 minutes for the following calculations. Calculate the minimum durations of each weather and incident combinations using Equation 2 15 , which is the time that both weather and incident events occur in category 4 scenarios . rL rk ro 2 15 is the duration of weather type i PAGE 26 26 is the duration of incident type j Calculate the difference between weather and incident dur ation of each combination using the following equation : rL rF rk ro 2 16 2) Adjust the category 4 scenario probabilities For weather event i and incident event j , the study period probability can be calculated from the following : rL rH f 2 17 is the base scenario probability of weather event i and incident event j The sum of all adjusted category 4 probabilities should be less than the sum of the base scenario probabilities of the demand pattern using Equation 2 14 . rO 2 18 is the number of weather types for the selected demand pattern is the number of incident types for the selected demand pattern Should the constraint in Equation 2 18 not be met , certain weather or incident events with high probab ilities in category 4 need to o ccur more than once , and the proba bility of each occurrence is equal. 3) Calculate residual probabilities for category 2 and 3 probabilities For weather and incident events that have different durations, the effect of the longer event will be modeled through the residual probabilities, which can be calculated using Equation 2 19 and 2 20 . PAGE 27 27 is t he residual probability for weather type i in category 2, can be calculated using Equation 2 19 . For a scenario of weather type i and incident type j , if the duration of weather type i is longer than incident j , rL , else, rL . rL rH rH 2 19 is t he residual probability for incident type j in category 3 , can be calculated using Equation 2 20 . Fo r a scenario of weather type i and incident type j , if the duration of incident type j is longer than weather type i , rL , else, rL . rL rH rH 2 20 Since the probabilities in category 4 may not only represent the category 4 base scenarios, but a portion of category 2 or 3 also. So the calculated residual probabilities should be taken out from the initial weather or incident probabilities. Also, t he re sidual probabilities should be lower than the category 2 and 3 initial base ; if not, more than one event need s to be applied to solve the problem. And we n eed to restart from Stage 2 procedure 3) . 4) Calculate the study period probabilities for categor y 2 and 3 The residual probabilities should be taken out from the initial weather and incident base scenario probabilities, using Equation 2 21 and 2 22 , respectively: rL rF 2 21 rL rF 2 22 The remainder probabilities for weather and incident are adjusted using Equation 2 23 and 2 24 , respectively, to get the study period probabilities for category 2 and 3: rL rH rk ro 2 23 PAGE 28 28 rL rH 2 24 The sum of category 2, 3, and 4 study period probabilities should be lower than the sum of the base scenario probabilities of the selected demand pattern, if not, then some of the events in category 2 and 3 should occur more than once to solve the problem. 5) Calculate the study period probabilities for category 1 The study period probabilities for category 1 is the sum of base scenario probabilities of the selected demand pattern minus the sum of study period probabilities for category 2, 3 , and 4. Stage 3 Detailed scenario generation Scenarios with probabilities and adjustment factors are assigned with detailed information such as start time, duration, and location for weather or incident. Each possible combination has equal probability to occur. For weather events, there are two possible start times : beginning of SP, middle of the SP. For incident events, there are two possible start times : beginning of SP, and middle of the SP . The durati on of an incident has three possibilities : 25 th percentile incident duration, 50 th percentile incident duration and 75 th percentile incident duration . An incident has three possible locations: first basic segment, middle basic segment, and last basic segme nt . Considering all the combinations of the detailed scenarios, the maximum number of scenarios is: N = 12 (DPs) + 12 (DPs) 10 (weather types) 2 ( weather start times) + 12 (DPs) 5 ( incident types) 2 (incident start times) 3 (incident durations) 3 (locations) + 12 (DPs) 10 (weather types) 5 (incident types) 2 (incident start times) 3 (incident durations) 3 (locations 2 (weather start times) = 22,932 . PAGE 29 29 With the detailed scenarios generated, analyst can use the HCM freeway facilities method to estimate the travel time for each scenario, and then all the travel times calculated can be compiled into a travel time distribution to evaluate the reliability of the facility . HCM M ethod ology The HCM methodology (HCM 6 th Edition, TRB 2016) introduces some stochastic elements to the general procedure introduced through the SHRP2 L08 project. First , the number of scenarios with their probabilities are calculated, and the demand adjustment factors are c alculated and assigned with the generated scenarios. Then, weather and incident events are generated and randomly assigned to the scenario. Specific procedures for the HCM methodology (2) are explained as following : Stage 1 Scenarios and DAFs In this stage , the number of scenarios are generated based on demand pattern s, and scenario probabilities are also calculated based on number of days associated with the scenario. Unlike the SHRP2 L08 method, the number of scenarios and their probabilities are fixed an d will not change in the following calculations. Step 1. Create the base file Base file contains the basic facility information, such as facility geometry (such as segment type, number of lanes), terrain, jam density, capacity, demand entry flow rates, truck percentage, FFS, AP, SP, RRP, etc. Step 2 . Determine the number of demand pattern s For HCM method, by default, the number of demand pattern is: 5 (weekdays) 12 (months) = 60 (demand pattern s). Step 3 . Demand pattern scenario sets and the total numb er of scenarios PAGE 30 30 The scenario sets number could be 4 or 5, since e ach demand pattern usually consists of 4 or 5 calendar days, the default scenario sets number is 4. Total number of scenarios: 4 (scenario sets) 60 (demand pattern s) = 240. These 240 scenar ios are believed to represent the demand variability throughout the studied RRP. Step 4 . Calculate the DAF of the demand pattern using Equation 2 1 . Step 5 . Calculate scenario probabilities The probability of a scenario is the number of days for the demand pattern associated with the scenario divided by the product of number of days in the RRP and number of scenario sets , it can be calculated from the following: rL rH f8 f2 25 is the number of days in demand pattern k is the sum of number of days for all demand pattern s or the number of days in RRP Stage 2 Weather adjustment factors In this stage, the calculations are on a monthly basis. Firstly, the expected weather event frequencies for a selected month are calculated. Secondly, the generated weather events are randomly assigned to the scenarios in the current month. Then, the start times are randomly assigned to the weather events . Step 6 . Group scenarios b y month Since the assigning of weather events to scenarios is on a monthly basis, the scenarios should be grouped by month. Step 7 . Expected frequency of weather events by month PAGE 31 31 The expected weather event frequency of weather type i in month m can be calculated using Equation 2 26 . rL rH rH 2 26 is the tim e wise weather event probabili ty of weather type i in month m is the duration of study period in hours is the n umber of s cenarios associated with month m is the expected duration of the weather type i rounded to the nearest 15 minutes and expres sed in hours Step 8 . Update the list of weather events for the current month Firstly, f or the weather events generated in Step 7, associ ate them with their durations, SAFs and C AFs. Secondly, randomly assign the scenarios of the current month to the weather event list generated based on the scenario probabilities. Thirdly, randomly assign the start times (from the time periods in the study period) to the weather event list. Step 9 . Check for temporal overlap with other weather events In one scenario, t empor al overlap between weather events is not all owed , Equation 2 27 is the constraint for assigning weather start times , if the constraint is met, th e weather events are not overlapping, if not, the weather events are overlapping , and a new start time should be assigned to the current weather event . rF rP 2 27 is the weather event start time assigned with a smaller time period is the weather event start time assigned with a larger time period is the duration in time periods of the smaller time period weather event PAGE 32 32 When assigning scenario number and sta rt time to a weather event, check the scenario number, start time and weather duration of all the former weather events. If t here is overlap, u ndo the last weather event assignment ; else move to S tep 10 . Do this until all weather events in the curr ent month are associated with scenarios and start times . Stage 3 Incident adjustment factors In this stage, the calculations are on a monthly basis. Firstly , the expected incident frequencies of each month are calculated. Secondly, the incident frequencies fo r all scenarios in the current month are calculated. Thirdly, incident durations, incident start times and incident locations are randomly assigned to the scenarios generated. Step 10 . Expected frequency of incident events by month Same as the SHRP2 L08 me thod, t he expected frequency for each month can be calculated based on Equation 2 6 to Equation 2 10 . Step 11 . Generate a set of incident frequencies for all scenarios in the current month The number of incidents in a study period follows the Poisson distribution: rL 2 28 is the expected incident frequency for month m from Equation 2 6 The number of scenarios that are assigned k incidents for month m can be calculated using the following: rL rH 2 29 Doing S tep 10 and 11 for all the month s and we can have all incident events generated throughout the months. Step 12 . Randomly assign each generated incident event to a scenario in current month PAGE 33 33 Randomly assign the scenarios of the current month to the incident event list generated based on the scenario probabilities. Step 13 . Generate incident severities for each incident event Num ber of incidents with severity i : rL rH 2 30 is the total number of incident s generated Step 14 . Randomly assign an incident severity to each incident Random ly assign incident severity to the number of incidents generated based on the distribution of incident seve rit ies from Step 13. Step 15 . Generate incident durations by incident severity The duration for each incident severity type follows a truncated lognormal distribution. For each incident type, a set of duration bins can be determined, usually with the bin i nterval of 15 min utes , then truncate the first and last bin interval depending on the range of the incident duration. For example, the duration of shoulder closure is from 8.7 to 52.5 min, the set of bin values can be 15, 30, 45, 60, the bin intervals are: (8.7, 22.5], (22.5, 37.5], (37.5, 52.5]. The probability of each bin can be calculated from Equation 2 31 , and then normalize the probabilities to make the total to be 1. rL rF f. f. rP 2 31 d is the set of incident durations (bin values) in 15 minutes incident duration range is the parameter conve rted by the mean incident duration m is the parameter converted by the stand ard deviation of incident duration v PAGE 34 34 Since the incident duration sample is non logarithmized, the mean incident duration and standard deviation need to be converted to and , respectively , using the following equations : rL f. 2 32 rL rE f. 2 33 Step 16 . Randomly assign in cident durations by severity Random assign incident durations to the incident events based on the probabilities from Step 15. Step 17 . Generate the distribution of incident start time and location The distribution of incident start time s will coincide with the distribution of facility VMT across the analysis periods . Also, the distribution of incident locations will be tied to the distribution of study period VMT across segments . Distribution of the incident location: rL 2 34 is the VMT on a specific segment is the VMT on the whole facility Distribution of the incident start time: rL 2 35 is the VMT of the assigned analysis period for the incident start time is the VMT across all the analysis periods of the study period Step 18 . Generate incident start times a nd locations for all incidents Number of incidents assigned a location (segment) x : PAGE 35 35 rL rH 2 36 Number of incid ents assigned a starting time (analysis p eriod) y : rL rH 2 37 Step 19 . Random assign incident start time and location From the list of events, select an incident whose start time and location have not been assigned, and randoml y assign a start time and location based on the probabilities and numbers calculated from Step 17 and 18 . Step 20 . Check spat ial or temporal overlap with other incident events Spat ial or temporal overlap of incidents for the same scenario is not allowed: When assigning a location and start time to a n incident event, check the scena rio number, location and start time of all the former incident events. If t h ere is overlap , un do the last start time and locat ion assignment ; if there is no overlap , move to the next assignment. Stage 4 Overall scenarios Now that all scenarios have the demand, weather and incident information assigned, these scenarios are believed to represent the demand, weather and incident variabilities. Same as the SHRP2 L08 method, analyst can use the HCM freeway facilities method to estimate the travel time fo r each scenario, and then all the travel times calculated can be compiled into a travel time distribution to evaluate the reliability of the facility. Performance Measures To evaluate the results and conduct the travel time reliability analysis of the fac ility, proper performance measures are needed. The following performance measures are commonly used in the TTR analysis : Travel time index (TTI), the ratio of the actual travel time on a facility to the theoretical travel time wh en traveling at free flow speed. PAGE 36 36 P lanning time index (PTI), the 95 th percentile travel time index (TTI) (95 th percentile travel time divided by the free flow travel time) . Misery index, the average of the highest five percent of travel times divided by the free flow travel time . Fa ilure/ On time measures, such as the percent of trips complet ed exceed/within a defined travel time threshold. Semi standard deviation, the standard deviation of travel time pegged to free flow travel time rather than the mean travel time . Reliability rating, the percent of trips serviced at or below a threshold travel time index (1.3 for freeways) . PAGE 37 37 CHAPTER 3 SOFTWARE TOOL DEVELOPMENT Brief Overview The TTR software was built on the .NET Framework using the C# language. The TTR software program developed in thi s project is designed to utilize the Freeway Facility software module (of the HCM C ALC softw are suite) for setting up base facility network file and to perform the HCM freeway facilities analysis metho dology calculations on each generated scenario (i.e., set of input conditions) from the TTR software program . The HCM CALC: Freeway Facility software module was developed by Dr. Scott Washburn prior to the start of this project . The basic procedure of TTR analysis via the soft ware tool is : First, a base file that includes the basic facility information (facility geometry, demand entry flow rates, etc.) is generated through HCM CALC: Freeway Facility software module . Second, additional data (demand, weather and incident variabil ity information) are specified through the TTR user interface ( UI . Third, the TTR software tool generate s scenarios based on both the base file and the user specified settings of demand, weather and incident inputs . Then, both the base file information an d scenario information are passed in to the HCM CALC: Freeway Facility software module for the core HCM freeway facilities analysis procedure calculation. Finally, in the TTR software tool , the scenario results obtained from the HCM CALC: Freeway Facility software module are aggregated into travel time distribution and reliability MOEs are calculated. Figure 3 2 shows the basic process flow of the TTR analysis . Overview of Software Screens The f ollowing is an overview of the software screens . The guiding design principle for the software was to set u p separate screens for each unique component of inputs (base file PAGE 38 38 properties, demand inputs, weather inputs, etc.) and results, to . cription, analyst information, and base file. There are three scenario generation methodologies in this software . In addition to the SHRP2 L08 and HCM methods, the Unrestricted method is also included in this software tool. U nlike the other two methods, it allows users to specify any combination of demand, weather, incident, and work zone settings; thus, a very large number of input scenarios can be generated. While a detailed overview of the Unrestricted scenario generation methodology is beyond the sco pe of this project , guidance on its use is provided i n the user guide . Obje ct 3 1. User Guide (.pdf file 12,396 KB The screen is where the user specif ies the input settings for the generation of scenarios, it includes the demand patterns , weather events and incident events screens. month and day groups for the demand patterns , specify the demand multipliers, ratio of base file demand to AADT, and number of scenario sets per demand pattern. s specify the time based weather probabilities for each month, wea ther durations and relative adjustment factors for each weather type. For SHRP2 L08 method, the results of this module are the weighted weather probabilities for each demand pattern, Figure 3 8 . For HCM method, the results of this module are the expected weather events for each month, as shown in Figure 3 9 and Figure 3 1 0 . Charts of weather events type, weather events by month and weather start times are also provided, Figure 3 11 . PAGE 39 39 incident rate or crash rate or HERS model input and incident duration input . Incident adjustment factors can be specified in the form shown in Figure 3 13 . Results of this module for SHRP2 L08 method are the incident probabilities by demand pat tern, as shown in Figure 3 12 . Results for HCM method are the incident events by month, as shown in Figure 3 14 and Figure 3 15 . Charts of incident events are also provided, as shown in Figure 3 16 . User can check the distribution of the generated incident severities, start times, locations and durations. Basic statistics for the charts of incident durations are also provided, so that u sers can check if the mean and standard deviation from the chart could For SHRP2 L08 method, there is a n screen, which combines the demand pattern, weather and incident probabilities and convert the base probabilities to the study period probabilities. The u ser can set the probability threshold to remove scenarios with probabilities lower than the threshold. A summary table is provided to show the information of the generated scenarios with their probabilities. T screen displays the list of scenarios generated . Click the and results of each scenario will show up i . Time period results for an i ndividual scenario can also be viewed, Figure 3 22 . Overall results and charts are provided for analysis, as shown in Figure 3 23 and Figure 3 24 . PAGE 40 40 Figure 3 1 . Software components and relationship PAGE 41 41 Figure 3 2 . Process flow PAGE 42 42 Figure 3 3 . SHRP2 L08 scenario generation methodology PAGE 43 43 Figure 3 4 . HCM scenario generation methodology PAGE 44 44 Figure 3 5 . Project properties Figure 3 6 . Demand settings PAGE 45 45 Figure 3 7 . Weather settings Figure 3 8 . Weather probabilities by DP for SHRP2 L08 PAGE 46 46 Figure 3 9 . Weather events by month for HCM Figure 3 1 0 . Weather event list for HCM PAGE 47 47 Figure 3 11 . Weather event chart for HCM Figure 3 12 . Incident settings PAGE 48 48 Figure 3 13 . Incident adjustment factors form Figure 3 14 . Incident adjustment factors for HCM PAGE 49 49 Figure 3 15 . HCM incident event list Figure 3 16 . HCM incident event chart PAGE 50 50 Figure 3 17 . Base probabilities for SHRP2 L08 Figure 3 18 . Study period probabilities for SHRP2 L08 PAGE 51 51 Figure 3 19 . Scenario event numbe rs for SHRP2 L08 Figure 3 20 . TTR scenarios listing PAGE 52 52 Figure 3 21 . TTR scenarios results Figure 3 22 . Individual scenario results viewer PAGE 53 53 Figure 3 23 . TTR overall results Figure 3 24 . TTR overall results charts PAGE 54 54 CHAPTER 4 VERIFICATION Overview This chapter documents the efforts performed to verify the accuracy of the implementation of the SHRP2 L08 and HCM TTR analysis methodologies into the TTR software tool. An example freeway facility is created for TTR analysis . Computational documentation for the two re liability scenario generation methods are provided. Calculated results from both the computational documentation and the sof tware tool are used to verify the accuracy of the software tool. Input Data The studied freeway facility contains 11 segments, detai led geometry of the facility are given in Table 4 1 and Table 4 2 . Other fact s of the example includes: Study period is from 4 to 7 p.m., with a duration of 3 hours Analysis period is 15 min RRP are all weekdays in the calendar year Mainline segments FFS = 60 mi/h, ramp segments FFS = 40 mi/h Acceleration and deceleration lane length = 500 ft Jam density is 190 pc/mi/ln Capacity is 2,300 pc/h/ln Short length of the weaving segment is 1640 ft Total ramp density is 1.0 ramp/mi Terrain is level for all segments Percent of truck on the facility is 5% Incident rate is 1050 incidents per 100 million VMT Computational Documentation This chapter provides computation al documentation via Mathcad (Appendix A) regarding the step by step calculations in the SHRP2 L08 and HCM methods, as wells as the overall results calculation. The computational documentation serve as a verification of the software tool, PAGE 55 55 in the documentation, besides the calculation results from the functions, there will be screen shots from the softw are to ol for comparison . Verification of Software Tool for SHRP2 L08 Methodology Following are the results from the TTR software tool for the TTR analysis using SHRP2 L08 method on the example facility specified above. The SHRP2 L08 method is a d eterm inist ic approach , and for this case a total of 604 scenarios will be generated. The overall results from the software considering all the time period results in each scenario are in Table 4 3 . Verification of Software Tool for HCM Methodology The f ollowing are the results from the TTR software conducting the TTR analysis using the HCM method on the example facility specified previously . The HCM method involves the generation of random numbers in weather event generation and incident event generation part , the software has random number seeds settings for the HCM method , same random number seeds setting will output the same results. To verify the weather event and in cident event generation , the average weather events information (weather start times) and incident events information (incident severities, incident start times, incident locations and incident durations) are collected fro m 100 sets of random seeds runs. F or all the incident verification figures, the blue bar is the frequency of the incidents generated from the software, the orange line is the probability based on the user input. The blue The incident severity distribution specified in this example is: 0.75, 0.20, 0.05, 0, 0 for shoulder closure, one lane closure, two lane closure, three lane closur e and four or more lane closure , respectively . From Figure 4 3 we can see that the generated incident severity distribution matches with the input. The incident start time distribution is based on the VMT by time period, in this example, the probability of VMT in each of the 12 time period s is: 0.068, 0.08, 0.09, 0.10, 0.11, 0.12, 0.10, PAGE 56 56 0.08, 0.07, 0.06, 0.05, and 0.05. From Figure 4 4 we can see that the generated i ncident start time distribution matches with the user input. The incident location distribution is based on the VMT by segment, in this example, the probability of VMT in each of the 11 segment is: 0.16, 0.05, 0.07, 0.05, 0.16, 0.09, 0.16, 0.04, 0.01 , 0.04 and 0.17. From Figure 4 5 we can see that the generated incident location distributi on matches with the user input. The shoulder closure duration distribution follows a truncated lognormal distribution based on the mean duration, duration range and standard deviation of the duration for shoulder closure. The shoulder closure duration distribution probabilities based on the user input shoul d be: 0.239, 0.483, 0.238 and 0.041 for 15, 30, 45 and 60 min duration, respectively. Figure 4 6 shows that the frequencies of the shoulder closure durations generated match with the shoulder closure duration distribution based on the user input. Statistics from Table 4 4 shows that the durations gen erated from the software match with the user input , the truncation of the distribution due to duration range may cause certain loss of the standard deviation . The one lane closure duration distribution follows a truncat ed lognormal distribution based on the mean duration, duration range and standard deviation of the duration for one lane closure. The one lane closure duration distribution probabilities based on the user input should be: 0.188, 0.511, 0.259 and 0.042 for 15, 30, 45 and 60 min duration, respectively. Figure 4 7 shows that the frequencies of the one lane closure durations generated match with the one lane closure durati on distribution based on the user input. Statistics from Table 4 5 shows that the durations generated from the software match with the user input, the truncation of t he distribution due to duration range may cause certain loss of the standard deviation. PAGE 57 57 The two lane closure duration distribution follows a truncated lognormal distribution based on the mean duration, duration range and standard deviation of the duration for two lane closure. The two lane closure duration distribution probabilities based on the user input should be: 0.121, 0.496 and 0.383 for 30, 45 and 60 min duration, respectively. Figure 4 8 shows that the frequencies of the two lane closure durations generated match with the two lane closure duration distribution based on the user input. Statistics from Table 4 6 shows that the durations generated from the software match with the user input, the truncation of the distribution due to duration range may cause certain loss of the standard deviation. For each weather event, the start time is randomly assigned from the 12 time periods in the study periods, so the weather start time distribution should be uniformly distributed, as shown in Figure 4 9 . The overall results for the HCM method are from the average of 20 sets of random seeds runs, as shown in Table 4 7 . Comparison A detai led comparison of the differences between the results produced by the two methods is beyond the scope of the project. However, a brief summary will be provided in this section. From the output results in Chapter 4, the following reliability performance mea sures for SHRP2 L08 and HCM methods can be obtained. From Table 4 8 we can see that, for the example facility, the mean TTI of SHRP2 L08 method is greater than HCM method . The 50 th % TTI of the two methods are the same, a s the percentile goes up, the relative percentile TTI for HCM method tend to be smaller than the SHRP2 L08 method. This means that the HCM method has less extreme TTI values than SHRP2 L08 PAGE 58 58 method, which is further backed up by the fact that the misery index and semi standard deviation of HCM m ethod are both smaller than SHRP2 L08 method. The reason for the difference of extreme TTI values between the SHRP2 L08 method and HCM method s lies in the assignment of weather and incident events, which is also the major difference between the two scenari o generation methods. For SHRP2 L08 method, a weather event can only start at either the beginning or middle of the study period, and an incident event can only start at either the beginning or middle of the study period, only on the first, middle or last basic segment. For the HCM method, however, the start times of weather and incident events are randomly assigned based on the proportion of VMT of each time period in the study period, and the locations of the incident events are randomly assigned based on the proportion of VMT of each segment of the facility. In this case, the limitations of start times and locations for weather and incident events of the SHRP2 L08 method may result in severe conditions for cert ain time period and segment, which lead to lo ng travel times. In this case, the more weather and incident events generated, the more likely severe conditions for certain time period s and segment s would happen for the SHRP2 L08 method , which will lead to the increase of difference between the extreme travel time values of the two methods . To demonstrate this, a set of travel time reliability runs have been conducted to the same example facility through the software by increasing the incident rate. For the HCM method, increase of incident rate will lead to the increase of incident frequencies generated. For SHRP2 L08 method, increase of incident rate will increase the probabilities of scenarios that involve incidents. For the example problem, the incident rate for each month is 1050 per million VMT, a se t of runs are conducted by increasing the incident rate by 50 per million VMT at a time. PAGE 59 59 Figure 4 11 shows that as the incident rate increases, the difference of the 95 th % TTI between the two methods also increases, which means that the effect of extreme travel times will be more significan t for SHRP2 L08 method as the number of incidents generated increases . Figure 4 1 . Example freeway facility Source : Highway Capacity Manual 2010 Note : ONR = on ramp segment, OFR = off ramp segment Figure 4 2 . TTI distribution of SHRP2 L08 method PAGE 60 60 Figure 4 3 . Incident severity distribution Note : 1 = shoulder closure, 2 = one lane closure, 3 = two lane closure, 4 = three lane closure, 5 = four or more lane closure Figure 4 4 . Incident start time distribution Figure 4 5 . Incident location distribution PAGE 61 61 Figure 4 6 . Shoulder closure duration distribution Figure 4 7 . One lane closure duration distribution Figure 4 8 . Two lane closure duration distribution PAGE 62 62 Figure 4 9 . Weather start time distribution Figure 4 10 . TTI distribution of HCM m ethod PAGE 63 63 Figure 4 11 . Difference of 95 th % TTI values between two methods Table 4 1 . Detailed geometry Following are the bas e demand entry flow rate for each ana lysis period in the study period : Segment Type Length (ft) Number of Lanes FFS (mi/h) 1 Basic 5280 3 60 2 On Ramp 1500 3 60 3 Basic 2280 3 60 4 Off Ramp 1500 3 60 5 Basic 5280 3 60 6 Weaving 2640 4 60 7 Basic 5280 3 60 8 On Ramp 1140 3 60 9 Ramp Overlap 360 3 60 10 Off Ramp 1140 3 60 11 Basic 5280 3 60 PAGE 64 64 Table 4 2 . Demand entry flow rate Analysis Period Demand Flow Rate ONR 1 ONR 2 ONR 3 OFR 1 OFR 2 OFR 3 1 3,095 270 270 270 180 270 180 2 3,595 360 360 360 270 360 270 3 4,175 360 450 450 270 360 270 4 4,505 450 540 450 270 360 270 5 4,955 540 720 540 360 360 270 6 5,225 630 810 630 270 360 450 7 4,685 360 360 450 270 360 270 8 3,785 180 270 270 270 180 180 9 3,305 180 270 270 270 180 180 10 2,805 180 270 270 270 180 180 11 2,455 180 180 180 270 180 180 12 2,405 180 180 180 180 180 180 Table 4 3 . Overall results for SHRP2 L08 method TT (min) TTI VHT VHD Speed Avg. (mi/h) Density (pc/mi/ln) Mean 7.07 1.18 110.82 13.48 55.86 24.49 Min 6.1 1.02 54.04 1.19 1.81 11.98 Max 212.84 35.47 746.45 723.94 59.03 161.58 50% 6.20 1.03 99.85 3.41 57.96 22.13 80% 6.37 1.06 134.97 8.43 58.51 29.91 95% 7.89 1.31 183.12 39.69 58.66 40.12 Standard Deviation 7.21 1.20 58.33 48.19 7.20 12.64 Misery Index 4.88 Semi standard Deviation 8.12 Table 4 4 . Shoulder closure duration (min) statistics Statistics User Input Generated Shoulder Closure Durations Mean 32.0 31.1 Standard deviation 15.0 11.8 Range 8.7 58.0 15 60 PAGE 65 65 Table 4 5 . One lane closure duration (min) statistics Statistics User Input Generated One lane Closure Durations Mean 34.0 32.5 Standard deviation 14.0 12.0 Range 16.0 58.2 15 60 Table 4 6 . Two lane closure duration (min) statistics Statistics User Input Generated Two lane Closure Durations Mean 53.0 49.6 Standard deviation 14.0 10.5 Range 30.5 66.9 30 60 Table 4 7 . Overall results HCM TT (min) TTI VHT VHD Speed Avg. (mi/h) Density (pc/mi/ln) Mean 6.85 1.14 108.08 9.85 56.35 23.92 Min 6.13 1.02 50.92 1.18 2.29 11.28 Max 168.72 28.12 620.18 591.79 58.62 134.25 50% 6.20 1.03 102.18 3.77 57.89 22.65 80% 6.29 1.05 133.79 6.45 58.51 29.64 95% 7.29 1.22 169.19 27.06 58.58 37.33 Standard Deviation 5.47 0.91 43.66 32.52 6.36 9.53 Misery Index 3.05 Semi standard Deviation 6.39 Table 4 8 . Reliability performance measure of two methods Reliability Performance Measure SHRP2 L08 Scenarios HCM Scenarios Mean TTI 1.18 1.14 50 th % TTI 1.03 1.03 80 th % TTI 1.06 1.05 95 th % TTI (PTI) 1.31 1.22 Misery Index 4.88 3.05 Semi standard Deviation 8.12 6.39 PAGE 66 66 CHAPTER 5 SUMMARY In this project, a software tool was developed that implements the large scale and highly iterative calculation processes of the SHRP2 L08 and HCM TTR analysis methodologies, specific to freeway facilities. After studying the SHRP2 L08 and HCM methodologies , they were implemented into a commercial grade software tool using the .NET Framework and C# programming language. The software tool focuses on the scenario generation for the two methods, and utilizes the HCM CALC: Freeway Facility software module to create the facility base file and perform the HCM freeway facility analysis methodology calculations on the scenarios generated from the TTR software tool . During the development process, besides addressing the functional requirement s , user friendliness also had to be taken into consideration due to the volume and complexity of the required input s , so that users can conduct the reliability methodologies effic iently . Also, feedback on the softwar e design features were obtained from stakeholders during this process. To test if the TTR methodologies had been correctly implemented in the software, a verification process was conducted. From the demand pattern group ings, weather events generation, incident events generation to the adjustment factors calculation and overall results calculation, each major step of the two TTR methodologies were included in the verification. To better present the verification calculatio ns and show the completeness of the verification process, computational documentation was developed in Mathcad, which can both show the functions and calculate the results. The Mathcad documents present the full travel time reliability methods and correspo nding calculations based on an example problem run with the software tool . T he results of each major step in the computational documents were compared to the corresponding results in the software tool . The verification process demonstrated that the two TT R analysis PAGE 67 67 methodologies were correctly implemented in the software tool. Additionally , a brief comparison was made between the relative results generated by the two TTR analysis methodologies , which showed that the SHRP2 L08 method tends to generate more extreme travel times than the HCM method when the number of incident events increased. This is likely because the SHRP2 L08 method limit s the start times and locations of the events, which may result in extreme travel times within certain time period s and segments. A u ser guide w as also developed to help guide the transportation practitioner on how to operate the software tool to perform the TTR analysis methodologies. PAGE 68 68 APPENDI X COMPUTATIONA L D OCUMENTATION PAGE 106 106 LIST OF REFERENCES 1. Kittelson, W., & Vandehey, M. (2013). Incorporation of Travel Time Reliability into the HCM (No. SHRP 2 Reliability Project L08). 2. Aghdashi, S., Hajbabaie , A., Schroeder, B. J., Trask, J. L., & Rouphail, N. M. (2015). Generating Scenarios of Freeway Reliability Analysis: Hybrid Approach. Transportation Research Record: Journal of the Transportation Research Board , (2483), 148 159. 3. Highway Capacity Manual, 5 th Edition . Transportation Research Board of the National Academies , Washington, D.C., 2010. 4. Highway Capacity Manual, 6 th Edition . Transportation Research Board of the National Academies , Washington, D.C., 201 6 . 5. Margiotta, R., McLeod, D., & Scorsone, T. (2015). Travel Time Reliability as a Service Measure for Freeways within Extensive Freeway Networks. In Transportation Research Board 94th Annual Meeting (No. 15 4381). 6. McLeod, D. S., Elefteriadou, L., & Jin, L. (2012). Travel Time Reliability as a Per formance Measure: Applying Florida's Predictive Model to an Entire Freeway System. Institute of Transportation Engineers. ITE Journal , 82 (11), 43. PAGE 107 107 BIOGRAPHICAL SKETCH Wei Sun was born in 1991 in the Shandong province, China . An o nly child in the family, he grew up in Weifang city. He graduated from Weifang Number One High School and got admitt ed to South China University of Technology in 2010. Wei earned his B.E. in civil engineering , specializing in transportation in 2014 and the n came to the University of Florida to study in the graduate program for transportation engineeri ng. After obtaining his m aster s degree, Wei will continue to pursue his graduate studies in the doctoral program at the University of Florida. |