Institute of Transportation, MOTC

       Complex network theory (CNT)is one of big data analytics methods which can be used to explore unsolved problems in previous studies. It has been found recently that researchers have just begun to apply CNT to railway accident and incident studies.

       This study aims at establishing a complex network model to explore factors contributing to TRA’s (Taiwan Railway Administration) accidents and incidents. In addition, the text analysis and correlation analysis are also used to find out the key factors of accidents and incidents. In order to discover which key factors can greatly reduce the occurrence probability of accidents and incidents, cascading failure theory is used to establish a safety propagation model in this study.

       Empirical results point out that the key factors of accidents and incidents are abnormal train power, abnormal train brake releasing, personal irregularities and signal failures. The key factors of vehicle failures are abnormal train power, abnormal train brake releasing, automatic train protection (ATP) system failure and unclosed VCB. To avoid from personal irregularities can reduce the occurrence probability of accidents and incidents by 47%. To avoid from abnormal train brake releasing can reduce the occurrence probability of vehicle failure by 40%.

       This study has also used visualization toolbox to show the network model and the key factors in graphs which can more easily overview the factor correlations and the importance of the key factors.