基于随机邻域嵌入的无监督复杂工况识别

Unsupervised Complex Condition Recognition Based on Stochastic Neighborhood Embedding

现代工业生产设备通常结构复杂并交替运行于不同工况,基于监测数据进行准确的工况识别是对系统进行健康监测的基础,但系统的监测数据通常维度较高、数据量较大。针对设备复杂工况的识别问题,提出了一种基于随机邻域嵌入的无监督工况识别方法。采用随机邻域嵌入算法,能够保留数据的局部和全局结构特性;计算了高维和低维空间中数据点的概率相似性,可实现设备高维监测数据的降维和无监督聚类,在不建立系统模型的基础上达成准确识别系统工况的目的。结果表明:该方法可有效实现高维监测数据的复杂工况识别,是一种有效的无监督聚类学习方法。

Modern industrial production equipment usually has a complex structure and runs alternately in different working conditions.Accurate working conditions identification based on monitoring data is the basis of health monitoring of the system,but the monitoring data of the system usually has a high dimension and a large data volume.To identify the complex equipment operating conditions,an unsupervised operating condition identification method based on stochastic neighborhood embedding is proposed.The stochastic neighborhood embedding algorithm can simultaneously preserve the local and global structural characteristics of the data,and also calculate the probability similarity of data points in high-dimensional and low-dimensional space to achieve the dimensionality reduction and unsupervised clustering of equipment high-dimensional monitoring data and to accurately identify the system operating conditions without establishing a system model.The results show that the proposed method can effectively identify complex operating conditions from high-dimensional monitoring data,which is an effective unsupervised clustering learning method.