摘要
针对齿轮箱故障特征重叠难以有效分离问题,提出基于局部切空间排列与多核支持向量机的齿轮箱故障诊断模型。在由振动信号时域统计指标及内禀模态分量能量构造的多元特征空间中,据局部切空间排列算法对多元特征进行非线性降维处理,得到初始低维流形结构,获取最优敏感特征向量;将该特征向量输入至多核支持向量机进行学习训练与故障辨识。局部切空间排列能克服传统降维方法的不足,多核支持向量机可实现复杂故障高精度、自动化智能诊断。通过齿轮箱故障模拟实验验证该方法的有效性。
In consideration of the overlapping of gearbox fault features and the difficulty to distinguish these features, a gearbox fault diagnosis model based on local target space alignment and multi-kernel support vector machine was proposed. In the vibration feature space constructed by time domain statistic indices and intrinsic mode energy value, the nonlinear muhi-dimensionality reduction based on local target space alignment was introduced to get the initial lowdimensional manifold feature value, then the low-dimensional feature vector which retains the fault characteristics was regarded as the input feature vector of the multi-kernel support vector machine for gearbox fault classification. Local target space alignment can overcome the shortcoming of traditional reduction method, and the multi-kernel support vector machine can realize the high-precision automatic intelligent diagnosis for gearbox. The gearbox fault diagnosis experiment shows the effectiveness of this novel model.
出处
《振动与冲击》
EI
CSCD
北大核心
2013年第5期38-42,47,共6页
Journal of Vibration and Shock
基金
国家自然科学基金项目(51275546)
重庆市自然科学杰出青年基金计划资助项目(CQ CSTC2011jjjq0006)
关键词
局部切空间排列
多核学习
支持向量机
齿轮箱
故障诊断
local target space alignment
multiple kernel learning
support vector machine
gearbox
fault diagnosis
