Analysis of U.S. freight-train derailment severity using zero-truncated negative binomial regression and quantile...

Area of Science:

• Transportation Safety
• Statistical Modeling
• Railroad Engineering

Background:

• Freight train derailments are the most frequent U.S. freight-train accidents, causing significant infrastructure damage, service disruptions, and potential harm.
• Understanding and preventing derailments is a high priority for the rail industry and government due to their severe consequences.
• Despite low probability, the impact of train derailments necessitates better understanding of factors influencing their severity.

Purpose of the Study:

• To develop statistical models for estimating train derailment severity.
• To provide a comprehensive understanding of derailment severity distribution.
• To offer insights for developing cost-efficient train safety policies.

Main Methods:

• Developed a zero-truncated negative binomial (ZTNB) regression model to estimate the conditional mean of derailment severity.
• Developed a quantile regression (QR) model to estimate derailment severity at different quantiles.
• Utilized both models to analyze the complete distribution of train derailment severity.

Main Results:

• The ZTNB model estimates the average severity of train derailments.
• The QR model provides insights into severity at various points in the distribution, not just the mean.
• Combined models offer a more complete picture of derailment severity under different conditions and causes.

Conclusions:

• The developed regression models enhance the understanding of train derailment severity.
• Results can inform risk assessments for various operational conditions and accident causes.
• This research supports the development of effective and economical train safety strategies.