摘要
在研究发动机各类故障诊断的基础上,结合贝叶斯网络从数据中学习的方法,提出一种能够根据实际样本数据对发动机的各类故障进行可视化诊断的方法,其充分考虑了先验知识,且能够根据实际样本数据对先验知识进行修正。以发动机WP7的故障为例,通过因果关系建立贝叶斯网络的可视化模型,结合先验知识进行参数学习和推理,实例结果表明,该模型及分析方法很好地反应了各部件或子系统的故障对于整个系统故障的影响以及各部件或子系统之间的依赖关系及依赖程度,有助于找出系统的薄弱环节和提高系统可靠性的途径。
Based on the study of various types of engine fault diagnosis,combined with Bayesian network method of learning from the data,a method that analyzed various engine fault visually according to the actual sample data was proposed.This method not only fully considers the prior knowledge,but also corrects the prior knowledge according to the actual sample data.Finally,taking the engine WP7 fault as an example,the visual Bayesian network model was modeled through the causality,parameter learning and reasoning were performed combined with priori knowledge.The result shows that the model and analysis method can well reflect the failure of the parts or subsystems for the entire system failure and the dependencies and degree of that between the parts or subsystems,which can help to find out the weak links of the system and the way to improve the reliability of the system.
出处
《计算机工程与设计》
北大核心
2015年第7期1908-1911,1916,共5页
Computer Engineering and Design
基金
国家自然科学基金委员会与中国民用航空总局联合资助基金项目(61179048)
中央高校基金项目(ZXH2012D003)
天津市支撑计划基金项目(重点)(11ZCKFGX04100)
关键词
发动机故障诊断
贝叶斯网络
可视化
参数学习
推理
engine fault diagnosis
Bayesian network
visualization
parameter learning
reasoning