Type of Document Thesis Author Wang, Jianying Author's Email Address firstname.lastname@example.org URN etd-09192010-221633 Title Accident-based Model for Estimating User Cost on Florida Bridges Degree Master of Science Department Civil and Environmental Engineering, Department of Advisory Committee
Advisor Name Title John O. Sobanjo Committee Chair Lisa K. Spainhour Committee Member Wei-chou V. Ping Committee Member Keywords
- user costs
- accident risk
- negative binomial regression
Date of Defense 2010-09-08 Availability unrestricted AbstractThe Florida Department of Transportation (FDOT) is currently in the process of implementing the AASHTO Ware Pontis Bridge Management System (BMS) for planning and programming maintenance, repairs, rehabilitation, improvements, and replacement for bridges on the state highway network. One important task of BMS is to accurately estimate the user costs to help life-cycle planning of bridge investments and realize an optimal funding and timing that will keep roads and bridges in service at minimum cost. The user costs primarily include travel time costs, vehicle operating costs, and accidents costs.
Traditionally, travel time costs and vehicle operating costs are greatly influenced by detour lengths. This thesis examines the bridges features that would increase travel time through bridges without detour, and establishes Microsoft Excel spreadsheet templates for calculating travel time costs and vehicle operating costs for 5,219 Florida highway bridges based on bridge characteristics, pavement conditions, and level of services. These templates are integrated into Florida Pontis system to help better decision-making.
At present, the accident cost employed in Pontis BMS is simply projected as a single function of the roadway width. However, this is inadequate in predicting bridge accident rates because it is widely believed that the rate strongly depends on other characteristics of bridges such as bridge length, number of lanes, and roadway conditions (Johnston et al, 1994, and Thompson et al, 1999). This thesis formulated discriminant functions and established regression models based on 2003-2007 Florida crash data at bridge sites in an effort to take the impacts of other bridge characteristics on accident costs into accounts. A discriminant function using logistic regression was established to determine whether a bridge has safety hazards. The results showed that the number of lanes, ADT, bridge length and urban arterials are key features affect the bridge safety.
To model the accident rates on bridges, three models were investigated, including linear regression model, Poisson regression model, and negative binomial regression models. Compared to linear regression and Poisson regression, the negative binomial model appears to be better in accuracy, especially for predictions within an error of one count of accident, performing at above 80% accuracy for observed counts three or less.
Filename Size Approximate Download Time (Hours:Minutes:Seconds)
28.8 Modem 56K Modem ISDN (64 Kb) ISDN (128 Kb) Higher-speed Access Wang_J_Thesis_2010.pdf 2.73 Mb 00:12:37 00:06:29 00:05:41 00:02:50 00:00:14