CCR based Key Performance Indicator Monitoring Method for Industrial Processes

In this paper, a canonical correlation regression (CCR) based process monitoring method is proposed for detecting a class of ker performance indicator (KPI)-related and KPI-unrelated faults. The proposed method aims at estimating the null space of the modelled process of interest, and then the corresponding test statistics are constructed, based on the estimation, to offer meaningful monitoring performance. In order to demonstrate the effectiveness of the proposed approach, a comparison study with least squares based and partial least squares based method is evaluated with respect to three performance indices, namely false alarm rate, fault detection rate and expected detection delay. Case study on one simulation example shows the effectiveness and applicability of the CCR-based fault detection method.