A novel process monitoring method based on improved DTW-MKECA

Aiming at the features that modem industrial processes always have some characteristics of complexity and non-linearity, an integration of improved dynamic time warping (DTW) and multi-way kernel entropy component analysis (MKECA) is proposed in this paper. DTW-MKECA applies DTW method at first to process unequal-length data, which ensures that the original data distribution can be preserved. Then it employs MKECA method to make data from the input space map to high dimensional feature space and analysis these data. MKECA ensures that little information of the original data is lost during the dimensionality reduction, meanwhile, a monitoring model can be established. The proposed process monitoring method combines the advantages of DTW and MKECA. The simulation results of monitoring between modeling data and online data of reheating furnace demonstrate the availability of proposed method in this paper.