Improved Regularization-Based Subspace Method for Harmonic Load Identification

To address the robustness and accuracy limitations of conventional approaches in periodic dynamic load identification, an improved regularization-based subspace method for harmonic load identification is proposed. First, the state-space model is constructed using the subspace identification algorithm. Second, the adaptive adjustment index and the optimal regularization parameter are calculated to jointly achieve load identification. The method incorporates an improved Tikhonov regularization strategy into the subspace identification algorithm, innovatively introducing an adaptive adjustment index. By applying differentiated weighting to singular values, the proposed method imposes varying degrees of suppression on different components, thereby enhancing robustness against disturbances associated with small singular values while preserving the fidelity of dominant energy components. Numerical simulations on a rotor system demonstrate that the proposed method achieves a mean absolute error (MAE) of 2.32 N and a determination coefficient [Formula: see text] of 0.96 under a 5% noise level. Finally, experimental results demonstrate that the method reduces MAE by at least 30.04% under different load conditions. These validations confirm that the proposed method effectively enhances load identification accuracy while exhibiting strong robustness.