Application of a discrete jump-detection algorithm to a continuous system

Abstract

The effectiveness of a discrete jump detection and state estimation algorithm applied to a continuous system is investigated. A three-state decision rule (yes, no, maybe) is used to detect parameter jumps. For the maybe region, a random walk model of the system is invoked. The continuous system modeling program was used to simulate a continuous system and obtain sampled input-output sequences. A sudden change in the original system produced a change in one of the parameters of its autoregressive moving average model, which was detected by the algorithm, and the new value was estimated. The problems of sensitivity and false alarm are discussed.<>