Research on a real-time change detection method for feedback systems using the fast recursive algorithm of subspace model identification

点击翻译按钮获取中文摘要

The goal of this research is to develop on-line change detection methods which are useful for real-time surveillance and early warning of dynamical changes, e.g., machine faults, of a system connected into a feedback loop. In this research, it has been clarified that the asymptotic properties and the optimality of a closed-loop subspace model identification which is used for surveillance of a system. Several experiments on real-life systems, such as a cart system, a cart-inverted pendulum system and an altitude control system of a radio-controlled helicopter, have been performed to show usefulness and effectiveness of proposed methods.