Abstract: This paper reviews the findings of research in incident duration analysis． A brief introduction of traffic incidents，incident types，and incident management systems is given first．The theoretical argument of the distribution of the incident durations is explained．Several methods used to analyse the incident durations，including variance analysis，regression，nonparametric regression， decision tree，hazard duration，and fuzzy logic，are described．The research concludes that the most research found that the distribution of the incident duration is right skewed and that standardisation of the data collection is important for the incident duration analysis． It is also found that the traditional statistic methods have difficulties in modelling the incident duration with the incomplete incident data． The future research in incident duration analysis was recommended such as artificial intelligence methods and standardisation of the data collection．

Key words: traffic incident, incident duration, incident management

摘要： 对交通事故持续时间分析的发展现状进行了综述．首先简要介绍了交通事故的定义，事故种类和交通事故管理系统，并分析了交通事故持续时间的统计分布．然后，对最近应用的几种主要的交通事故持续时间的分析方法及结论进行了概述．主要包括：方差分析，线性回归，非参数回归，决策树法，hazard duration method，模糊逻辑．主要的结论包括；大量的研究发现交通事故持续时间的分布形状是向左偏斜的钟型分布．交通事故数据采集标准化是事故分析中非常重要的环节．传统的统计分析方法处理不完整的数据有局限性．最后，本文提出了将来的发展方向，包括应用人工智能方法建立事故持续时间模型，及建立交通事故数据库标准．

关键词: 交通事故, 事故持续时间, 事故管理