融合多源信息的液压动力单元故障诊断方法

Multi-source-information Based Fault Diagnosis Method of Hydraulic Power Unit

针对液压动力单元中元件数量繁多、联系复杂,以及难以精确进行故障诊断的问题,提出了融合多源信息的贝叶斯网络故障诊断新方法,建立了液压动力单元单一故障失效模式的贝叶斯网络拓扑结构,开发了融合多源信息的故障诊断网络。利用Noisy-OR/MAX计算了故障诊断网络的条件概率参数;针对两故障并发模式,将观测信息节点加入已建立的单一故障模式诊断网络,辅助诊断液压动力单元的两故障并发失效。实例诊断结果表明:对单一失效的诊断,观测节点对诊断结果的影响不大,正确率均接近100%;对于两失效同时发生的情况,未考虑观测信息的模型出现了漏诊和误诊的问题,正确率不到50%,而考虑观测信息的模型能够正确诊断出所有两失效并发的情况,正确率达到100%。研究结果可为水下生产系统的故障诊断奠定基础。

To address the problem of large number of components and complicated connection in the hydraulic power unit which makes it difficult to accurately diagnose the fault, a new Bayesian network fault diagnosis method based on multi-source information is proposed. A Bayesian network topology of single fault failure mode of the hydraulic power unit is established. A fault diagnosis network integrating multi-source information is developed. The conditional probability parameters of the fault diagnosis network are calculated using Noisy-OR/MAX. For the two simultaneous faults modes, the observation information nodes are added to the established single fault mode diagnosis network to assist diagnosing the two simultaneous faults of the hydraulic power unit. The case diagnosis results show that, for the diagnosis of single fault, the observation node has little effect on the diagnosis result, and the correct rate is close to 100%. For the case of two simultaneous faults, the model without the observation information has the problem of missed diagnosis and misdiagnosis, of which the correct rate is less than 50%. However, the model considering the observation information can correctly diagnose two simultaneous faults, and the correct rate reaches 100%. The study can lay the foundation for fault diagnosis of subsea production system.