Direct Multivariate Intrinsic Time-Scale Decomposition for Oscillation Monitoring

The detection of plant-wide oscillation in an industrial process is of great significance. Recently, indirect multivariate intrinsic time-scale decomposition (MITD) (IMITD) has been pioneered for the adaptive processing of multi-loop data, which is restricted by the problem of projection sensitivity. To solve the challenge, the direct MITD (DMITD) algorithm is proposed and featured by the following contributions: 1) Three novel concepts including the multivariate extremum, multivariate baseline-node, and baseline-operator are defined for the purpose of developing DMITD; 2) Compared with IMITD, the implementation of DMITD is more robust to the selection of projection directions; and 3) DMITD outperforms traditional techniques in capturing both the regularity and evolution of the plant-wide oscillation from noisy signals in the nonlinear and nonstationary process. The proposed method is demonstrated by simulation as well as one industrial case.