基于故障树的复杂装备模糊贝叶斯网络推理故障诊断

复杂装备的小批量、个性化定制属性, 注定了其生命周期过程中存在着相对较多的不确定性, 故障隐患必不可免, 故障诊断尤为重要。因此，提出基于故障树的复杂装备模糊贝叶斯网络推理故障诊断模型。首先, 通过分析复杂装备的结构组成, 建立复杂装备的故障树模型。其次, 利用故障树转化法, 构建基于故障树的贝叶斯网络拓扑结构。然后, 针对复杂装备结构数据缺乏和专家打分的不确定性, 通过模糊集合论方法确定条件概率等参数。最后, 进行案例研究, 利用模糊贝叶斯网络推理中的因果推理和诊断推理, 诊断出案例中的故障(潜在故障)节点, 证明了所提方法的有效性。研究成果不仅解决了贝叶斯网络中利用搜索函数构建最优网络不符合实际的问题, 也通过模糊集合论解决了复杂装备数据缺乏和专家打分不确定性的不足。所提模型不仅适应于过程诊断中故障的确定, 同时也适用于事前诊断中潜在风险的识别, 而且对于故障(或潜在故障)节点的改善效果还能起到检测评估的作用。

Fuzzy Bayesian network inference fault diagnosis of complex equipment based on fault tree

The small-lot and customized attributes of complex equipment determines that there are relatively many uncertainties in the process of its life cycle,so potential fault hazard of complex equipment is inevitable,and the fault diagnosis of complex equipment is particularly important.Therefore,a fuzzy Bayesian network inference model for complex equipment fault diagnosis based on fault tree is proposed.First,the fault tree model for complex equipment is established by analyzing its structural composition.Secondly,the fault tree transformation method is used to construct the Bayesian network topology structure based on fault tree.Then,in view of lacking of structural data of complex equipment and the uncertainty of expert scoring,the fuzzy set theory is used to determine the parameters such as conditional probability.Finally,a case study is carried out,and the fault nodes(potential faults)of the cases are diagnosed by using causal inference and diagnostic inference in fuzzy Bayesian network inference,which is proved effective.The research results not only solves the problem that it is not practical to construct the optimal network by using search function in Bayesian network,but also solves the problem of data deficiencies of complex equipment and uncertainty of expert scoring by using fuzzy set theory.The proposed model is suitable not only for determining fault in process diagnosis,but also recognizing potential risk in advance diagnosis,and also play a role in the detection and evaluation of the improvement effect of faults(or potential faults)nodes.