Modeling and monitoring of multimode transition process based on reconstruction

In this paper, a new nonlinear reconstruction algorithm is proposed for handling the multimode problem in the industrial processes. The contributions are as follows: (1) in different batch cycles, the transition patterns may follow different trajectories and reveal different characteristics. After the operation patterns change, the two neighboring modes still share some common correlations; (2) nonlinear reconstruction model is built based on traditional characteristics; and (3) a new fault direction matrix is designed to modify the reconstruction equation. The corresponding confidence regions are constructed according to their models respectively. The effectiveness of the proposed method has been demonstrated via simulated examples.