Laplacian双联最小二乘支持向量机用于早期故障诊断

发布时间：2018-01-22 13:27

本文关键词： 旋转机械流形学习 Laplacian双联最小二乘支持向量机半监督学习故障诊断　出处：《振动与冲击》2017年16期 　论文类型：期刊论文

【摘要】：提出基于Laplacian双联最小二乘支持向量机(Laplacian Twin Least Squares Support Vector Machine,LapTLSSVM)半监督模式识别的新型早期故障诊断方法。用时、频域特征集广泛收集旋转机械不同早期故障的特征信息,再用提升半监督局部Fisher判别分析(Enhanced Semi-Supervised Local Fisher Discriminant Analysis,ESSLFDA)将高维时、频域特征集约简为具有更好类区分度的低维特征向量,并输入到Lap-TLSSVM中进行早期故障诊断。Lap-TLSSVM引入了包含大量无标签数据信息的流形规则实现半监督学习;其目标函数只含等式约束条件,且用共轭梯度法求解目标函数的线性方程组以加速训练过程。所提出的方法在训练样本非常稀少的情况下具有较高的诊断精度和计算效率。深沟球轴承早期故障诊断实验验证了该方法的有效性。
[Abstract]:A dual least squares support vector machine (LS-SVM) based on Laplacian is proposed. Laplacian Twin Least Squares Support Vector Machine. A new method of early fault diagnosis based on LapTLS SSVM (Semi-supervised pattern recognition). The feature set of frequency domain is widely used to collect the feature information of different early faults of rotating machinery. Then the Fisher discriminant analysis was used to promote the semi-supervision department (. Enhanced Semi-Supervised Local Fisher Discriminant Analysis. ESSLFDA reduces the feature set in high dimensional time and frequency domain to a low dimensional feature vector with better classification. And input into Lap-TLSSVM for early fault diagnosis. Lap-TLSSVM introduces manifold rule which contains a lot of untagged data to realize semi-supervised learning. The objective function contains only equality constraints. The conjugate gradient method is used to solve the linear equations of the objective function to accelerate the training process. The proposed method has high diagnostic accuracy and computational efficiency in the case of very few training samples. The effectiveness of this method is verified by diagnostic experiments.
【作者单位】：四川大学制造科学与工程学院;重庆大学机械传动国家重点实验室;四川航天技术研究院总体设计部;
【基金】：国家自然科学基金青年基金(51305283) 中国博士后科学基金第60批面上资助项目(2016M602685)
【分类号】：TH17
【正文快照】： 轴承、转子、齿轮等旋转部件是机械装备的重要组成部分,起到支撑载荷、传递运动和动力的关键作用。然而这些旋转部件一旦出现故障得不到及时诊断和处理,任由其发展、扩大,将会引起整台机械装备的重大事故以及代价高昂的长期停工。因此,及时准确地诊断出处于早期发生阶段的微弱

【相似文献】