Fault Diagnosis Filter for Sensor and Actuator Based on System Identification Model

Abstract

Model-based fault detection and isolation (FDI) is a method based on a mathematical model of a process, and the basic principle of the method, which is classified as a filter type, is the unknown input observer (UIO). This is the monitoring of the estimation error of the actual output. A feature of UIO is its robustness to process modeling errors, which plays an important role in improving the accuracy of FDI. However, conditions for its application are strict, and its application to the system identification model is difficult. In addition, although FDI automation is unavoidable in terms of practical application, past research has focused on improving diagnostic accuracy, and this point has been overlooked. In this study, a novel FDI filter for sensors and actuators based on a system identification model and a method for its automatic and robust application to plant-wide measurement noise are proposed. The effectiveness of the proposal was demonstrated by a simulator of the vinyl acetate monomer manufacturing process, and it was shown that plant-wide FDI can be automatically performed by dividing the large-scale process, identifying the systems, and combining the various models.