摘要
针对工业过程复杂的数据分布特性,本文提出了一种基于局部近邻标准化(LNS)的概率主成分分析(PPCA)故障检测方法(LNSPPCA)来解决由于过程数据的多模态特性和不确定性所引起的故障检测效果不理想问题。首先,通过LNS解决数据多模态问题,使标准化后数据尽可能的服从单一高斯分布,然后,使用PPCA方法从概率的角度对数据进行分析,能够考虑到数据的随机性,从而更真实的描述数据,提取更加全面有价值的信息,有效的在复杂的数据分布过程中对故障进行检测。因此,LNSPPCA方法可以有效提高多模态过程复杂数据分布的工业过程故障检测能力。利用数值例子和TE过程进行应用实验,并将测试结果与主成分分析法(PCA)、PPCA方法进行对比,验证了LNSPPCA方法的有效性。
Aiming at the complex data distribution characteristics of industrial processes, this paper proposes a probabilistic principal component analysis fault detection method based on local neighborhood standardization(LNSPPCA) to solve the problem of unsatisfactory fault detection effect caused by multi-modal characteristics and uncertainty of process data. Firstly, LNS is used to solve the data multi-modal problem, so that the standardized data obey a single Gaussian distribution as much as possible. Then, the PPCA method is used to analyze the data from the perspective of probability, which can take into account the randomness of the data, so as to describe the data more realistically, extract more comprehensive and valuable information, and effectively detect faults in the complex data distribution process. Therefore, the LNSPPCA method can effectively improve the industrial process fault detection capability in multi-modal process complex data distribution. Numerical examples and TE process were used to conduct application experiments, and the test results are compared with those of principal component analysis(PCA) and PPCA methods, which verifies the effectiveness of the LNSPPCA method.
作者
李元
张昊展
唐晓初
Li Yuan;Zhang Haozhan;Tang Xiaochu(School of Information Engineering,Shenyang University of Chemical Technology,Shenyang 110142,China;School of Automation,Shenyang Aerospace University,Shenyang 110136,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2021年第2期75-85,共11页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(61673279)资助项目
关键词
多模态
全信息
局部近邻标准化
概率主成分分析
故障检测
multimodal
full information
local neighborhood standardization
probabilistic principal component analysis
fault detection
