基于并行算法的证据理论合成器的复合故障诊断  

Fault Diagnosis for Complex Fault Based on the Parallel Algorithm of Evidence Theory Synthesizer

作　　者：汪醒鹏[1] 屈波[1] 文亚南[1] 

机构地区：[1]河海大学,江苏南京211100

出　　处：《流体机械》2014年第4期32-36,共5页Fluid Machinery

摘　　要：针对旋转机械复合故障的不确定性和模糊性,在蚁群神经网络的基础上,引入并行机制改进算法。利用多线程技术增大蚁群的搜索区域,同时采用编码映射匹配法则(EMM)提高匹配效率,缩短蚁群寻路时间,加快算法收敛速率,并对BP神经网络进行优化,结合概率转化(BPA)辅助决策。计算结果表明,合成器对复合故障识别率高,与人类决策一致,对其他模拟进化算法有借鉴意义。Considering the uncertainty and fuzziness of compound fault of rotating machinery, ant colony neural network algorithm is modified by the introduction of parallel reasoning algorithm in this article. The region of search of the hive is enhanced by multithreading technology, and the application of EMM improves the matching efficiency to a certain degree. Meanwhile, the time of finding the route is obviously shortened, thus accelerating the rate of convergence. Moreover, BP neural network is also optimized. As for aid decision making,BPA is applied. As the result shows, synthesizer learns pretty fast and combined failures cart easily be detected, which seems to be in line with human decision-making. Consequently, the modification provides a reference for other evolntionary algorithm simulations.

关 键 词：复合故障 蚁群算法 EMM 多线程 BPA 

分 类 号：TH13[机械工程—机械制造及自动化] TP273.3[自动化与计算机技术—检测技术与自动化装置]