Quantitative Performance Based Framework for Guardrail Maintenance Investment Decisions
Li, Ning, Civil Engineering - School of Engineering and Applied Science, University of Virginia
Park, Byungkyu, Department of Civil Engineering, University of Virginia
Guardrail system is an important roadside safety feature to reduce crash impacts for run-off-roads vehicles. In current practices of guardrail asset maintenance, investment decisions are rarely made based on the quantitative performance measures such as the safety benefits of guardrail. Existing literature is very limited when it comes to a methodology to quantify safety benefits of traffic assets down to the asset level.
This dissertation research establishes a quantitative performance based framework for guardrail maintenance investment decisions by integrating advanced spatial analysis, crash prediction model, crash severity model, and engineering economics model. A proof-of-concept case study, which quantifies the annual safety benefit of each guardrail run for the Interstate Highway Systems of Richmond, Virginia, has been successfully implemented as a base for a data driven and risk based guardrail maintenance investment strategy.
A core part of the framework is the development of a Roadway Departure (RD) crash prediction model. The Highway Safety Manual (HSM) is considered as the current best practice in crash modeling but it does not address the spatial heterogeneity issue. This dissertation innovatively applies a spatial modeling technique for road segments and develops Geographically Weighted Poisson Regression models (GWPR). The GWPR model not only significantly improves model performance over the HSM model as indicated by all goodness of fit measures, but also effectively addresses spatial heterogeneity and over dispersion issues.
Although guardrail systems are designed to reduce crash impacts, the actual effects of in-service guardrails on reducing crash severity have rarely been validated and quantified. This dissertation also uses real-world guardrail and crash data to assess the effectiveness of guardrail systems in reducing fatal and severe injury crashes. A roadway departure crash severity model is developed using a binary logit model and statistical proportion tests were conducted to compare roadway departure crash severity with and without guardrail hits. Both methods suggest that hitting guardrail could reduce the probability of fatal and severe injury crashes by about 45-50% and that the reduction is statistically significant.
Lastly, an economic model integrates crash frequency and severity results to quantify the annual safety economic benefits of each guardrail asset. The results of the economic model enable a data driven and risk based approach to optimize maintenance investment decisions of guardrail assets and to maximize the safety benefits with limited funding resources. With minor modification, the framework could be adapted and applied to other transportation safety assets such as lightings or crash cushions.
PHD (Doctor of Philosophy)
Guardrail, Roadway Departure Crash, Maintenance Investment, Crash Prediction Model, Geographically Weighted Regression
All rights reserved (no additional license for public reuse)