A Data-Driven Attack Detection Approach for DC Servo Motor Systems Based on Mixed Optimization Strategy

This article is concerned with the data-driven attack detection problem for cyber-physical systems with the actuator attacks and measurement noise. In most of existing data-driven detection methods, H
_∞
index is used to characterize the sensitivity performance. It is well-known that compared with the H
_∞
index, H_ index can significantly improve the diagnostic performance. However, the detection system design based on the H_/H
_∞
mixed optimization technique has not been solved within the data-driven framework. In this article, a residual generator is constructed from the available input-output (I/O) data. H
_∞
and H_ indices are defined from the viewpoint of time-domain to characterize the robustness of residual generator against measurement noise and sensitivity to attack signals, respectively. In particular, a novel weighting system, which is expressed as an I/O model, is designed to transform the H_ performance into an H
_∞
constraint, and the detection system design problem based on H_/H
_∞
mixed optimization technique is finally formulated into a constraint-type optimization one, which can be solved by the classical Lagrange multiplier method. Also, the proposed detection method is applied to a networked dc servo motor system to verify its advantages and effectiveness.