变量加权型主元分析算法及其在故障检测中的应用

发布时间：2018-07-25 10:52

【摘要】：传统主成分分析(PCA)算法旨在挖掘训练数据各变量间的相关性特征,已在数据驱动的故障检测领域得到了广泛的研究与应用。然而,传统PCA方法在建模过程中通常认为各个测量变量的重要性是一致的,因此不能有效而全面地描述出变量间相关性的差异。为此,提出一种变量加权型PCA(VWPCA)算法并将之应用于故障检测。首先,通过对训练数据进行加权处理,使处理后的数据能够充分体现出变量间相关性的差异。然后,在此基础上建立分布式的PCA故障检测模型。在线实施故障检测时,则通过贝叶斯准则将多组监测结果融合为一组概率指标。VWPCA方法通过相关性大小为各变量赋予不同的权值,从而将相关性差异考虑进了PCA的建模过程中,相应模型对训练数据特征的描述也就更全面。最后,通过在TE过程上的测试验证VWPCA方法用于故障检测的优越性。
[Abstract]:The traditional principal component analysis (PCA) algorithm is designed to excavate the correlation characteristics between the variables of the data training. It has been widely studied and applied in the field of data driven fault detection. However, the traditional PCA method usually considers that the importance of each measurement variable is consistent in the modeling process, so it can not describe the variation effectively and comprehensively. For this reason, a variable weighted PCA (VWPCA) algorithm is proposed and applied to fault detection. First, by weighting the training data, the processed data can fully reflect the difference of correlation between variables. On this basis, a distributed PCA fault detection model is established on the basis of online implementation. In the case of fault detection, the multi group monitoring results are fused into a group of probability index.VWPCA method through the Bias criterion to give different weights to each variable by the correlation size, thus the correlation difference is considered in the modeling process of PCA, and the corresponding model is more comprehensive for the description of the training data characteristics. Finally, through the TE process. The test verifies the superiority of the VWPCA method in fault detection.
【作者单位】：宁波大学信息科学与工程学院;
【基金】：国家自然科学基金项目(61503204) 浙江省自然科学基金项目(Y16F030001) 宁波市自然科学基金项目(2016A610092)~~
【分类号】：TP277

【参考文献】