Improved Data-Driven SKRs Based Fault Detection for Closed-Loop Systems with Deterministic Disturbance

摘要

In various practical industrial systems, control and monitoring performances are always more or less influenced by deterministic disturbance. To improve the fault detection performance with deterministic disturbance under closed-loop conditions, two improved data-driven stable kernel representations (SKRs) based on CSIMPCA are proposed in this paper. The first one is dedicated to decouple deterministic disturbance from closed-loop systems, which implies that only the noise information will be retained in the final residual signals. The second one will emphasize the compromise between deterministic disturbance and faults whose goal is to make the identified SKR more sensitive to system faults rather than deterministic disturbance. The proposed methods are tested on a randomly generated 4-order MIMO discrete-time LTI system. Two improved data-driven SKRs are demonstrated to achieve a superior fault detection result.