Directional KPLS for Fault Detection

In this paper, a directional kernel partial least squares (DKPLS) monitoring method is proposed. The contributions are as follows: (1) DKPLS algorithm is proposed to extract the output-relevant variables. A more direct relationship between the input and output variables is established based on the proposed method. (2) based on the proposed DKPLS algorithm, a process monitoring method is proposed. Faults are monitored by the proposed DKPLS monitoring method more accurately than KPLS monitoring method. The DKPLS monitoring method is used to monitor electro-fused magnesia furnace. The experiment results show the effectiveness of the proposed DKPLS monitoring method. Key Words: Directional Kernel Partial Least Squares (DKPLS); Output-relevant Variables; Residual Subspace; Process Monitoring.