A robust data-driven system identification approach with applications to reliability analysis of DC motor systems

System disturbances have become an undeniable factor in reliability analysis. To address the issue of unknown periodic disturbances in DC motor systems, this paper proposes a robust data-driven system identification approach for the reliability analysis of DC motor systems. First, based on subspace methods, a construction scheme for an unknown periodic disturbance space is designed using modified Bernstein polynomials, where the degrees of freedom of these polynomials are determined by an information criterion. Subsequently, a data-driven closed-loop input-output (I/O) extended model incorporating controller information under disturbances is derived using the subspace method. Unknown periodic disturbances are then eliminated through orthogonal projection techniques. Finally, the relationship between past input-output and future output is established, and the system parameters for closed-loop identification are estimated via canonical correlation analysis (CCA). The feasibility and effectiveness of the proposed approach are demonstrated through a case study on a DC motor system.