基于时空近邻标准化和局部离群因子的复杂过程故障检测

针对复杂过程数据的非线性、动态性和中心漂移等特征,提出了基于时空近邻标准化和局部离群因子的故障检测方法(TSNS–LOF).首先使用训练样本在时空两个方向上的近邻集来标准化训练样本;然后在标准样本集上计算样本的局部离群因子,并确定其上分位点作为检测控制限,进行在线故障检测.时空近邻标准化解决了复杂过程数据的非线性、动态性和中心漂移的问题;局部离群因子通过度量样本的相似度实现了故障样本和正常样本的分离.将TSNS–LOF应用于田纳西–伊斯曼过程(TE)过程进行故障检测实验,结果表明相对于主元分析、动态主元分析、k近邻、局部离群因子等方法, TSNS–LOF对故障预警更加及时且具有更高的故障检测率.理论分析和仿真实验说明TSNS–LOF方法适用于具有动态性或多模态特性或两者兼具的过程故障检测,能够更好地保障生产过程的安全性和产品的高质量.

Time-space neighborhood standardization-local outlier factor based fault detection for complex process

A fault detection method based on time-space nearest neighborhood standardization and local outlier factor (TSNS-LOF) was proposed to deal with the problem of nonlinear, dynamic and mean drift of complex process data. Firstly, use the time-space nearest neighborhood set to standardize the training sample; then the local outlier factor of standard sample was calculated on standard sample set, and used as a detection index to detect faults. The time-space nearest neighborhood standardization overcomes the difficulties of the dynamics and mean drift. The local outlier factor measured the similarity of samples, and the fault samples and the normal samples were separated. The fault detection experiment of TE process was carried out. The experimental results show that TSNS-LOF is timelier for the early fault warning, and have higher detection rate than PCA, DPCA, FD-kNN and FD-LOF methods. The theoretical analysis and simulation experiments show that the TSNS-LOF method is suitable for fault detection of dynamics or multiple or both operating faults and ensures the safety of the production process and high quality of products.