摘要
尽早识别轻微故障,对提高生产过程设备运行的安全性具有重要意义。为实现对设备轻微故障的正确识别和及时诊断,该文提出了一种基于损失函数的支持向量机(SVM)算法。应用模糊理论的方法对支持向量机分类及最优分类面进行了解释,对可疑分类区列出了模糊隶属度的表达式。针对故障诊断等问题中误判造成的损失不同这一特点,定义了基于损失函数的模糊隶属度,并得出了修正后的最优分类面。SVM算法可以实现对设备轻微故障的准确识别,并可近似地判别故障的严重程度。文中以汽轮机减速箱轴承运行状态诊断为例,对样本数据经K-L变换后进行可视化研究,分类结果表明了该算法的可行性。
It is very important to identify the slight mulfunction earlier for improving the security of facility in the process. In order to identify the slight mulfunction of facility correctly and timely, a Support Vector Machine (SVM) algorithm based on risk function is put forward. A method based on fuzzy theory is applied to explain the classification of SVM and its optimal hyperplane. An expression of fuzzy membership on doubtful classification area is listed. Aiming at the different loss by error judge in fault diagnosis, the fuzzy membership based on the loss function is defined and optimal hyperplane is modified. Based on the algorithm, we can identify the slight mulfunction of facility exactly and the degree of faults approximately. A diagnosis example on operation of axlebox of reducer. After K-L transformation, a research on the visualization is proceeded. The results are given to prove feasibility of the algorithm.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2003年第9期198-203,共6页
Proceedings of the CSEE
