TY - JOUR

T1 - Freeway Incident Duration Estimation and Temporal Instability Analysis Using Correlated Random Parameter Hazard-Based Model with Heterogeneity in Means and Variances

AU - Li, G.

AU - Xie, X.

AU - Yao, Z.

AU - Du, S.

AU - Jiang, C.

AU - Cai, J.

AU - Wu, L.

N1 - Export Date: 24 July 2025; Cited By: 0

PY - 2025/7/10

Y1 - 2025/7/10

N2 - Excessive freeway incident duration may increase the risk of secondary incidents or lead to other uncertainties. Investigation of the influence of various factors on freeway incident duration is crucial for reducing such risks and uncertainties. To mine the correlation and heterogeneity among influencing variables as well as the temporal stability, the correlated random parameters accelerated failure time (AFT) model with heterogeneity in means and variances (CRPHMV) was proposed to analyze freeway incident duration data. Given the incident duration data collected in China, the proposed model was compared with the other two state-of-the-art models. Estimation results showed that the CRPHMV AFT model can better fit the incident duration data and provide insights into the influence mechanism of random parameters with heterogeneity in means. Five variables (scraping, heavy vehicle, multivehicle, rainy, and lane 1 closure) exhibit heterogeneous effects. We also discovered that the mean heterogeneity of collision with fixed objects versus heavy vehicle indicators, and phone alarm versus lane 1 closure indicators, result in longer incident durations, while the mean heterogeneity of Keqiao West-Zhangjiafan versus rainy indicators is associated with reduced incident durations. Collision with fixed objects and nighttime significantly affect the variance of lane 1 closure's random parameter. Interaction effects between the unobserved heterogeneity of lane 1 closure and multivehicle indicators, as well as lane 1 closure and scraping indicators, were found to significantly impact incident duration. The results of a likelihood ratio test for global and local temporal instability showed that estimated coefficients of some variables were not stable among the 3 years of data used in model estimation. The findings of this study provide valuable information for reducing the duration of freeway incidents and formulating effective management policies in China. Additionally, they contribute to the knowledge of incident duration and methodology.

AB - Excessive freeway incident duration may increase the risk of secondary incidents or lead to other uncertainties. Investigation of the influence of various factors on freeway incident duration is crucial for reducing such risks and uncertainties. To mine the correlation and heterogeneity among influencing variables as well as the temporal stability, the correlated random parameters accelerated failure time (AFT) model with heterogeneity in means and variances (CRPHMV) was proposed to analyze freeway incident duration data. Given the incident duration data collected in China, the proposed model was compared with the other two state-of-the-art models. Estimation results showed that the CRPHMV AFT model can better fit the incident duration data and provide insights into the influence mechanism of random parameters with heterogeneity in means. Five variables (scraping, heavy vehicle, multivehicle, rainy, and lane 1 closure) exhibit heterogeneous effects. We also discovered that the mean heterogeneity of collision with fixed objects versus heavy vehicle indicators, and phone alarm versus lane 1 closure indicators, result in longer incident durations, while the mean heterogeneity of Keqiao West-Zhangjiafan versus rainy indicators is associated with reduced incident durations. Collision with fixed objects and nighttime significantly affect the variance of lane 1 closure's random parameter. Interaction effects between the unobserved heterogeneity of lane 1 closure and multivehicle indicators, as well as lane 1 closure and scraping indicators, were found to significantly impact incident duration. The results of a likelihood ratio test for global and local temporal instability showed that estimated coefficients of some variables were not stable among the 3 years of data used in model estimation. The findings of this study provide valuable information for reducing the duration of freeway incidents and formulating effective management policies in China. Additionally, they contribute to the knowledge of incident duration and methodology.

KW - Correlated random parameter

KW - Freeway

KW - Heterogeneity in means

KW - Incident duration

KW - Unobserved heterogeneity

U2 - 10.1061/AJRUA6.RUENG-1444

DO - 10.1061/AJRUA6.RUENG-1444

M3 - Journal article

VL - 11

JO - ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering

JF - ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering

IS - 3

M1 - 04025050

ER -