| 基于LLE算法和SVM的旋转机械故障诊断 |
| |
| 引用本文： | 苏盈盈,马飞,刘兴华,熊德杨. 基于LLE算法和SVM的旋转机械故障诊断[J]. 重庆电力高等专科学校学报, 2013, 0(6): 63-66 |
| |
| 作者姓名： | 苏盈盈 马飞 刘兴华 熊德杨 |
| |
| 作者单位： | [1]重庆科技学院电气与信息工程学院,重庆401331 [2]重庆大学自动化学院,重庆400044 [3]重庆电力高等专科学校,重庆400053 |
| |
| 基金项目： | 重庆市自然科学基金项目（项目编号este2012jjA40026）;重庆科技学院校内重点科研基金项目（项目编号CK2011Z01） |
| |
| 摘 要： | 利用LLE（Locally Linear Embedding）算法对众多的观测变量进行降维，再利用支持向量分类器SVM（Support Vector Machine）方法对降维后的变量数据集进行故障诊断。通过算例仿真表明，旋转机械故障的23维变量因素可降到14维，同时得到的诊断结果中，训练集的正确率为94．8％，测试集的正确率为100％。结果表明基于LLE算法和SVM的旋转机械故障诊断的模型精度有效。其既降低了模型的复杂度，又不影响故障诊断模型的精度。
|
| 关 键 词： | 局部线性嵌入 SVM 旋转机械 故障诊断 |
A Study on the Dimension Reduction of the Fault Diagnosis Model for Rotating Machinery Based on LLE |
| |
| SU Ying-ying,MA Fei,LIU Xing-hua,XIONG De-yang. A Study on the Dimension Reduction of the Fault Diagnosis Model for Rotating Machinery Based on LLE[J]. Journal of Chongqing Electric Power College, 2013, 0(6): 63-66 |
| |
| Authors: | SU Ying-ying MA Fei LIU Xing-hua XIONG De-yang |
| |
| Affiliation: | 1. School of Electrical and Information Engineering of Chongqing University of Science and Technology, Chongqing 401331, China ; 2. School of Automation of Chongqing University, Chongqing 400044, China ; 3. Chongqing Electric Power College, Chongqing 400053, China) |
| |
| Abstract: | There are numerous observation variables of the fault diagnosis model for the rotating mechanical system, which can lead to the problem of high dimension. This essay presents the application of LLE （ Locally Liner Embed- ding） in the dimension reduction of the variables. Then, SVM （Support Vector Machine）is used for subsequent fault diagnosis of the data sets. The results of the research show that the primary 23 dimension can be reduced to 14 di- mension. Meanwhile, the results of the diagnosis show that the accuracies of the training set and the test set are 94. 8% and 100% respectively. In conclusion ,the model is effective and accurate in the dimension reduction. |
| |
| Keywords: | LLE SVM rotating machinery fault diagnosis dimension reduction |
| 本文献已被 维普 等数据库收录！ |
