摘要
提出了一种基于人工免疫系统的在线故障诊断方法,实现人体内抗体库中保留有效记忆抗体及故障类型的实时准确识别的能力。基于人体免疫系统中T细胞和B细胞的理论,将抗体库分为第一抗体库和第二抗体库,第二抗体库的生成依赖于保留有记忆效应的第一抗体库的监测范围。该方法根据平均信息熵和欧氏距离理论,有效地设计了第一抗体集的覆盖范围,保证了探测的效果,并以此为基础生成整个故障诊断的抗体集,以更加有效地实现故障的诊断。提出的故障诊断系统应用于碳纤维生产过程中牵伸系统,结合牵伸环节实时采集的数据信息,得出了较准确的故障识别结论。
As a novel natural computation method, online fault diagnosis system（OFDS） inspired by artificial immune systems for realizing the real-time fault type identifying, memorizing, even self-responding to normal conditions after diagnosis in further development, was defined as fault detection system on dynamic stretching process in carbon fiber production line. OFDS proposed some improved operators, such as distance calculation for antibody selecting and rapidly antigen identified, in which the Average Information Entropy theory and the Euclidean Distance were used to design the first antibody population in order to ensure the detection space cover and to further enhance the effectiveness of entire antibody population. It was also applied to stretching model to demonstrate the effectiveness of fault diagnosis and corresponding ability of antibody memorizing. The results show that five fault types can almost be identified accurately.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2014年第4期830-835,共6页
Journal of System Simulation
基金
国家自然科学基金重点项目(61134009)
长江学者和创新团队发展计划(IRT1220)
上海领军人才专项资金
上海市科学技术委员会重点基础研究项目(13JC1407500
11JC1400200)
上海市教育委员会科研创新项目(14ZZ067)
中央高校基本科研业务费专项资金资助(2232012A3-04)
关键词
人工免疫系统
故障诊断
平均信息熵
抗体记忆
碳纤维生产
牵伸过程
artificial immune system
fault diagnosis
average information entropy
antibody memorizing
carbon fiber production line
stretching process
