基于隐半马尔可夫模型的动态多故障诊断方法研究  

Research on Dynamic Multiple Fault Diagnosis Model(DMFD) Based on Hidden semi-Markov Model

作　　者：葛鹏岳[1] 黄考利[1] 刘晓芹[1,2] 连光耀[1] 

机构地区：[1]军械工程学院,河北石家庄050003 [2]陆军航空兵学院,北京101123

出　　处：《计算机测量与控制》2010年第6期1280-1282,1286,共4页Computer Measurement &Control

摘　　要：针对现有故障诊断方法无法适应复杂装备多故障诊断要求的技术问题,提出了基于隐半马尔可夫模型(HSMM)的动态多故障模型,并给出动态多故障诊断问题的形式化描述;在隐半马尔可夫公式推导过程中对推导方法进行了改进,使得公式推导过程大大简化;通过将目标函数转化为若干个独立的子问题,并分别用基于二进制的粒子群算法进行求解,有效的实现了动态多故障诊断;该方法同现有方法相比,在不影响检测率和隔离率的前提下,计算简单,运行时间短,优化效果明显。The static multiple fault models in existence can not fit the need of the more and more complicated equipments as the development of electron technique. So, this paper put forward a dynamic multiple fault model which is based on Hidden semi-- Markov Model （HSMM） and give the formalization description of dynamic multiple fault diagnosis; this paper improve the way of getting the formulation and make the process simpler; this paper decomposed original dynamic multiple fault problems into several independent sub--problems and uses the Binary Particle Swarm Optimization （BPSO） to solve this problems. Compared with the methods in existence this method are simpler and costs less time without affect the FDR and FIR.

关 键 词：动态多故障诊断 隐半马尔可夫 粒子群 

分 类 号：TP206[自动化与计算机技术—检测技术与自动化装置]