基于NPE算法的工业过程故障检测特征分析
首发时间:2017-05-11
摘要:不同于传统的基于全局的维度约简算法,流形学习算法基于数据的局部几何特征实现数据降维。论文首先讨论流形学习算法的适用条件和数据的前提假设特征,并对流形学习算法在过程监控的应用进行理论分析;然后,以NPE为例,进一步深入探讨算法的特征提取思想和方法,以及T2和SPE统计量进行过程监控的有效性,阐述该统计量应用于流形学习的特点,指出利用这些统计量进行监控的特点。通过仿真实例验证上述分析的有效性。
For information in English, please click here
Analysis of Fault detection in Industrial Process Based on NPE Algorithm
Abstract:Different from the traditional global dimension reduction algorithm, the manifold learning algorithm is based on the local geometric features of data to achieve data dimensionality reduction. Firstly, the paper discusses the applicable conditions of the manifold learning algorithm and the prerequisite hypothesis of the data, and analyzes the application of manifold learning algorithm in process monitoring. Then, taking NPE as an example, this paper further discusses the effectiveness of the algorithm and the method of T2 and SPE statistics, and expounds the characteristics of the statistic applied to manifold learning, and points out that it can be monitored by using these statistics specialty. The validity of the above analysis is verified by a simulation example.
Keywords: manifold learning NPE statistic theory analysis local characteristics
论文图表:
引用
No.4732398119975914****
同行评议
共计0人参与
勘误表
基于NPE算法的工业过程故障检测特征分析
评论
全部评论