摘要
退化特征提取是机械健康状态监测的重要组成部分,伴随旋转机械长时间连续运转,退化特征出现性能波动甚至下降,给退化特征提取和选择造成了困难。首先利用一个特征映射算法库对振动信号提取特征,并基于Kolmogorov-Smirnov(KS)检验和Benjamini-Yekutieli过程对原始特征集进行过滤,然后利用双目标优化遗传算法(Bi-objective Optimization Genetic Algorithm, BOGA)结合支持向量机分类器(Support Vector Classifier, SVC),在有监督的环境下搜索出最佳特征子集,其中BOGA设置了SVC分类精确度和特征子集维数两个目标函数,前者进行最大化,后者进行最小化。通过在液压泵退化状态数据集上进行实验和在凯斯西楚大学轴承数据集与FRESH;CAa、ReliefF、JMIM三种方法进行对比,验证了该方法在退化状态识别上的较好性能。
Extracting degradation features is an important part of Monitoring the health status of machinery. The performance of degradation features fluctuates or even declines with the continuous operation of the rotating machinery for a long time, which makes it difficult to extract and select degradation features. First, a feature mapping Algorithm library was used to extract features from the vibration signals and the original feature set was filtered based on Kolmogorov-smirnov(KS) test and Benjamini-Yekutieli process. Then, the optimal feature subset was searched in the supervised environment by combining BOGA with SVC. The accuracy of SVC and the dimension of subset were two objective functions of BOGA, the former was maximized, the latter was minimized. The performance of proposed method was verified by the experiment on the data set of hydraulic pump degradation state and the comparison with FRESH;CAa, ReliefF and JMIM on the case western reserve university bearing data set.
作者
裴模超
张建军
李洪儒
于贺
PEI MoChao;ZHANG JianJun;LI HongRu;YU He(The Shijiazhuang Branch of Army Engineering University,Shijiazhuang 050003,China;32140 Troops of the PLA,Shijiazhuang 050003,China)
出处
《机械强度》
CAS
CSCD
北大核心
2021年第6期1280-1288,共9页
Journal of Mechanical Strength
基金
国家自然科学基金项目(51275524)资助。
关键词
旋转机械
退化状态识别
双目标优化遗传算法
支持向量机分类器
Rotating machinery
Degradation state identification
Bi-objective optimization genetic algorithm(BOGA)
Support vector classifier(SVC)
