对PNET中代表失效模式识别方法的改进

Improvement of the method of representative failure pattern recognition in PNET method

采用概率网络估算技术(probabilistic network evaluation technique,PNET)计算结构系统可靠度时,需要对代表模式进行识别,通常的做法是选择适当的相关系数界限值,然后对各失效模式与代表模式的关系进行判断,但是这种"一刀切"的方式忽略了事物固有的模糊属性。文章通过引入代表模式识别过程中的模糊性,采用λ-截集法对识别方法进行改进。通过算例验证得知,采用文中改进方法计算得到的结构可靠度更加精确,更加符合工程实际情况。采用不同的隶属函数形式和置信度值会影响计算得到的结果,因此应根据实际情况选择适当的隶属函数形式和置信度值。

As the probabilistic network evaluation technique(PNET)is used to calculate structural system reliability,it is necessary to identify representative failure patterns,and the usual approach is to select appropriate value of correlation coefficient index,then judge the relationship between the failure modes and the representative modes.This“broad brush”approach ignores the inherent attribute of fuzziness.By introducing the fuzziness of the representative pattern recognition process,the recognition method is improved by usingλ-cut set method.Example shows that the structural system reliability result is more accurate and more in line with the actual engineering situation.The different forms of membership function and confidence value will influence the calculation results,so the appropriate form of membership function and confidence value should be selected according to the actual situation.