Risk Assessment Of Serious Multi-fatality Crashes And Their Corresponding Countermeasures

In recent years,the overall road safety situation in our country keeps stable while road accidents have been slightly deceased in both frequency and fatal rate.However,significant crashes with injuries and deaths still happens from time to time,which cast a negative impact on economy,social stability and peoples’ living.The Serious Multi-fatality Crash(SMC)is defined as a motor-vehicle crash that occurs on a road and results in more than 10 deaths(including 10 deaths).Up till now,there are a number of published researches in regard with traffic accident,whereas few researches involved SMCs.Such crashes have characteristics of randomness,high fatal rate,and grave consequence.Therefore,it’s necessary to build up risk evaluation mechanism which includes measures and controls regarding with the potential risk evaluated to reduce the likelihood and consequency of SMCs.This dissertation focuses on the SMCs in China with application of theories and methods in safety engineering,traffic engineering,risk management,statistics and computer science,aims to give suggestions to key problems such as risk factor,influence mechanism,development tendency,risk level,countermeasures and so on.Firstly,risk identification was implemented to identify the statistical factors and risk motivators of the SMCs.Based on crash data from 2000 to 2012,the method of descriptive statistics was adopted to analyze tendency,temporal and spatial distributions,and classifications of SMCs;whereas comparative analysis was used to compare the similarity and difference in distributions of risk factors between SMCs and general road accidents with consideration of driver,vehicle,road and environment,respectively.Secondly,risk analysis was completed to quantify the effects of contributing factors of the SMCs.Based on the crash data from 2000 to 2012,the structure of the Bayesian Network related to the SMCs was built given specific model assumptions,expert’s subjective judgments,mutual information and conditional independence to analyze the influence mechanism of the SMCs.In addition,the EM algorithm was applied to calculate their posterior probabilities of the Bayesian Network to identify the influence degree of various factors in terms of the likelihood and the consequence of the SMCs.Thirdly,risk prediction was employed to estimate the number of the SMCs and the fatalities in the next few years.Based on the crash data from 1996 to 2012,time series analysis and grey prediction theory were adopted to develop two individual prediction models.The prediction accuracy of these two models was compared.Moreover,four combined prediction models were established to compare their prediction accuracy in further.Results indicate that the combined prediction model with optimal weights minimizing the sum of squared error is the most appropriate for risk prediction of the SMCs.Thus,the number of such crashes and their fatalities from 2013 to 2017 were estimated by using the abovementioned model.In addition,risk evaluation was conducted to determine the levels of risk associated with the SMCs.Based on the crash data from 1990 to 2014,the rank order cluster analysis was applied to determine classification criterions of the number of crashes and fatalities.Then,the risk picture was constructed to evaluate the risk of the SMCs from 1990 to 2017 and was validated with a reference of the SMCs from 2013 to 2014.Results show that the levels of crash risk from 1994 to 1998 are high and the levels of crash risk from 2013 to 2017 are low.Furthermore,the actual risks and the predictive risks related to the SMCs from 2013 to 2014 are consistent.Finally,risk control was implemented to reduce the likelihood and consequence of the SMCs.Based on the results of risk identification and analysis,relevant countermeasures for each risk factor were proposed to compose the solution database used for risk control.Accordingly,the specific scheme of risk control was made to lower the level of risk.If the level of risk was still too high for the selected scheme,a new scheme of risk control should be chosen from the solution database.Moreover,the solution database used for risk control needs to be updated and improved regularly.