Nonlinear system monitoring with piecewise performed principal component analysis

An innovated method of piecewise performed principal analysis (PCA) is proposed to monitor a nonlinear system in this article. By analyzing the highest concerned variable and critical influence variables distribution through the collected data, we separate the variable space into smaller sections and perform PCA for each section to derive the PCA model. The process is monitored by applying the new data to the model for the section in which the data locates. To demonstrate the performance, the method is applied for the monitoring of single steam pipe in a steam pipe network of an iron& steel plants. The simulation result shows the proposed method is effective and sensitive to the faults.