State space identification of closed loop systems

Within the joint input/output identification scheme, stochastic realization is used to solve the spectral factorization problem, thus providing a time-domain closed-loop identification method. First, a stochastic realization of the joint closed-loop system is obtained. Through state space partitioning, nonminimal realizations of the subprocesses can be calculated. Balancing reduction then provides minimal realizations of the subprocesses in the loop. Simulation examples show good results with the proposed method.<>