82 / 2025-04-05 20:33:33

Nonstationary Process Monitoring Based on Analytic Stationary Subspace Analysis

Nonstationary,process monitoring,principal angle,analytic stationary subspace analysis

Stationary subspace analysis (SSA) is an effective tool for nonstationary process monitoring, by dividing process data into stationary and nonstationary components. However, only stationary components are used for process monitoring in most of SSA-based methods. Nonstationary components are difficult to monitor because anomalies can be easily masked by trend variations. In this paper, a novel nonstationary process monitoring method is proposed based on the analytic SSA (ASSA) method. Different from conventional methods, nonstationary components are monitored by constructing a principal angle-based statistic. The core idea is that the corresponding subspace of nonstationary components is invariant in the fault-free case, and anomalies can be monitored by comparing the variation between offline and online nonstationary subspaces. The effectiveness of the proposed method is validated on the Tennessee Eastman benchmark process.