基于ANP和模糊证据理论的应答器系统风险评估方法研究

针对应答器系统风险评估过程中存在较大模糊性和主观性的问题,提出了基于网络分析法(ANP)和模糊证据理论的风险评估模型。首先,识别评估对象的风险因素,建立相应的ANP评价模型,采用专家均值置信法改进传统ANP,确定风险因素权重;然后利用模糊数学理论确定专家评语基本信度分配(BPA)函数,并运用折扣系数法修正证据理论,合成专家意见得到各风险因素的BPA函数;最后通过加权平均函数实现风险综合评判,依据最大隶属度原则确定风险等级。利用该评估模型对应答器系统功能失效进行了风险评估,验证了模型的可行性和有效性。

On the risk assessment method of balise system based on the ANP and fuzzy evidence theory

This paper is inclined to propose a risk evaluation model for the CTCS-2 balise system based on the analytic network process (ANP) and fuzzy evidence theory while introducing a comprehensive risk assessment method for the said system so as to solve the problem of faulty data absence and invalid expert qualitative evaluation in the presence of the risk assessment on the system, while considering the effects of the complexity,fuzziness and uncertainty of the system. To achieve the said purpose,we have first of all carried out a risk identification to gain the risk-involving factors of the system. Secondly, the ANP has been brought about to calculate the weights of each risk factor and convert the qualitative evaluation into the quantitative one through the fuzzy mathematical theory,whereas the evidence synthesis rules, the expert opinions should be integrated to realize the basic probability assignments (BPA) function. Thirdly,the average function can be used to evaluate and integrate the weights and BPA function in accordance with the maximum membership principle so as to determine the risk level of the balise system. In spite of this, there still exists a rather serious problem of subjectivity of the traditional ANP in building up the evaluation matrix,in case there exists serious conflict of expert opinions,for example,in case the evidence synthesis results may not be consistent with the real situations, for it can eventually lead to a false evaluation result. In that case,it would be possible to use the experts'mean assessment to improve the traditional ANP so as to reduce the errors caused by the human intervention factors. In addition,we have also proposed an improved verification theory based on the discount coefficient method in hoping to regulate the different conflicting or contradictory data and information. And,finally,a risk evaluation formula has been brought up forward for taking the function failure of the balise system as a case study sample to verify the proposed risk evaluation model. The evaluation results for the function failure we have gained prove to be the R2 (admissible) level,indicating that the process has been stated and defined in a detailed and rational way in the paper. The final evaluation results gained from the risk evaluation practice prove consistent with the actual situation with the model being rational,effective and feasible.