An Efficient Hierarchical Identification Method With Kernel-Based SVM for Equivalent Systems of Aircrafts

The identification of low order equivalent system (LOES) models from measured test data is critical to assessing system qualities and design of control law. This paper focuses on the situations with finite data records and the time-response method is employed. In such a case, although least squares (LS) algorithm is efficient for the identification of linear models, it suffers low accuracy in ill-conditioned scenarios. To address this issue, a hierarchical identification method based on the LS algorithm is designed and the optimal solution is attained while maintaining the low computational complexity and small memory requirements. In particular, an LS support vector machine (SVM)-based identification scheme is developed, where the system parameters are estimated in a reproducing kernel Hilbert space. The theoretical analysis and pertaining simulation results demonstrate the effectiveness of the proposed method.