Fault Detection and Identification Based on the Neighborhood Standardized Local Outlier Factor Method
详细信息 查看全文
- 作者：Hehe Ma ; Yi Hu ; Hongbo Shi
- 刊名：Industrial & Engineering Chemistry Research
- 出版年：2013
- 出版时间：February 13, 2013
- 年：2013
- 卷：52
- 期：6
- 页码：2389-2402
- 全文大小：766K
- 年卷期：v.52,no.6(February 13, 2013)
- ISSN：1520-5045
文摘
Complex chemical processes often have multiple operating modes to meet changes in production conditions. At the same time, the within-mode process data usually follow a complex combination of Gaussian and non-Gaussian distributions. The multimodality and the within-mode distribution uncertainty in multimode operating data make conventional multivariate statistical process monitoring (MSPM) methods unsuitable for practical complex processes. In this work, a novel method called neighborhood standardized local outlier factor (NSLOF) method is proposed. The local outlier factor of each sample, which means the degree of being an outlier, is used as a monitoring statistic. A new normalized Euclidean distance based on the local neighborhood standardization strategy is employed during the calculation of the monitoring index. Then, a contribution-based fault identification method is developed. Instead of building multiple monitoring models for complex chemical processes with different operating conditions, the proposed NSLOF method builds only one global model to monitor a multimode process without needing a priori process knowledge. Finally, the validity and effectiveness of the NSLOF approach are illustrated through a numerical example and the Tennessee Eastman process.