【基於核方法的机械故障特徵提取分类技术研究】李巍华.pdf

ADissertationSubmittedinPartialFulfillmentofthe MechanicalFaultFeatureExtractionand Classification BasedonKernelMethods Ph.D.Candidate:LI Weihua Major:Mechanical Engineering Supervisors:ProfessorYANGShuzi ProfessorSHI Tielin HuazhongUniversityofScience andTechnology Wuhan430074,P.R.China June.

华中科技大学博士学位论文通过对KFA方法分类得到的类边界进行分析,提出了基于正常域边界描述的故障检测方法,通过选择正常状态的特征样本确定正常域的边界范围,正常域边界外的范围则认为是异常状态所处的区域。此方法成功用于齿轮箱的裂纹故障检测,其检测正确率可以达到基于样本训练的KFA分类识别的正确率。最后,提出了一种基于KFA 的设备状态趋势分析方法,据此方法分析了齿轮箱由裂纹产生、扩展到崩裂断齿的劣化过程,得到了齿轮箱的趋势变化曲线,对于利用非线性方法分析设备运行状态的变化进行了初步探讨。

华中科技大学博士学位论文 mechanical faultdiagnosis.A methodfor selectingfeature samples of differentfaultsis presented in the.dissertation.The quantity of selected samples is much less than that of whole sample sets by this mean, which has quickly expediated the computation process.