基于神经网络与改进证据理论融合的故障诊断方法  

Fault Diagnosis Method Based on the Fusion of Neural Network and Improved Evidence Theory

作　　者：刘保杰 杨清文[1] 吴翔 房施东[1] 

机构地区：[1]陆军军官学院五系,安徽合肥230031

出　　处：《火炮发射与控制学报》2017年第4期92-97,共6页Journal of Gun Launch & Control

基　　金：军内科研项目(wj2015cj020001)

摘　　要：为提高液压驱动的火箭炮随动系统故障诊断准确性,将双BP神经网络和D-S证据理论相结合的数据融合方法引入到故障诊断中,运用并行的两个BP神经网络对液压驱动的火箭炮随动系统进行局部诊断,再用D-S证据理论对局部诊断的结果进行全局融合,克服了单一BP网络诊断的缺陷,使证据理论的基本可信度分配不再完全依赖专家进行的主观化赋值,从而实现了赋值的客观化,为解决冲突证据无法判别的问题,引入了信任系数的概念,修正融合结果,减少了故障分类识别的不确定性,提高了诊断系统的可靠性。通过液压系统实例,验证了该方法的可信度和可行性。In order to improve the fault diagnosis accuracy of hydraulic driven rocket launcher servo system （HDRLSS） , the data fusion method based on BP neural network and D-S evidence theory was introduced into the fault diagnosis, which used two parallel BP neural networks to perform local diagno- sis, and then used D-S evidence theory to fuse the results of local diagnosis comprehensively. The method overcomes the defects of the single BP network diagnosis, reduces the uncertainty of the fault classification and improves the reliability of the diagnosis system. Through making the basic reliability distribution of the evidence theory not completely depend on the expert subjective valuations, the im- personal valuations were realized. To resolve the conflict evidence problem of not being able to be dis- tinguished, the concept of confidence coefficient was introduced to correct the fusion result, which was able to reduce the uncertainty of the fault classification and improve the reliability of the diagnosis system. The reliability and feasibility of the method were verified by an example of HDRLSS.

关 键 词：流体传动与控制 多传感器信息融合 故障诊断 D—S证据理论 液压系统 

分 类 号：TJ393[兵器科学与技术—火炮、自动武器与弹药工程]