基于递推规范变量分析的时变过程故障检测

由于规范变量分析(CVA)不适应过程的时变特性,容易把正常的过程改变识别为故障.因此,针对时变过程提出一种故障检测方法是十分必要的.采用指数权重滑动平均来更新过去观测矢量的协方差矩阵.递推CVA有较高的计算负荷是需要解决的关键问题.通过引入一阶干扰理论来递推更新Hankel矩阵的奇异值分解(SVD).与普通奇异值分解相比,显著降低了递推算法的计算负荷.将提出的基于一阶干扰理论的递推规范变量分析(RCVA-FOP)应用于田纳西伊斯曼化工过程中.仿真结果表明,所提出方法不仅能有效适应过程的时变特性,而且可以有效检测到两种类型的故障.

Recursive Canonical Variate Analysis for Fault Detection of Time-Varying Processes

Because CVA (canonical variate analysis) is unable to adapt the characteristics of time-varying processes， by which the normal changes of the process is easily identified as faults， it is very necessary to propose a monitoring approach for time-varying processes. The exponential weighted moving average approach was adopted to update the covariance of the past observation vectors. The most critical problem faced by recursive CVA algorithm is the high computation cost. To reduce the computation cost， the first order perturbation theory was introduced to update recursively the singular value decomposition (SVD) of the Hankel matrix. The computation cost of recursive SVD based on the first order perturbation theory is significantly less compared to the SVD. Recursive canonical variate analysis based on the first order perturbation (RCVA-FOP) was applied in the Tennessee Eastman chemical process. Simulation results indicate that the proposed method not only can effectively adapt to the normal change of time-varying processes， but also can detect two types of faults.