Abstract: In seismic analysis of non-structural components (NSCs), decoupled analysis has become a prevailing approach. Among the available strategies, the use of response-spectrum-matched excitation records as input offers notable advantages, including reduced computational cost and lower dispersion in analysis results, thereby substantially improving the efficiency of NSC numerical analysis. To this end, this study proposes an excitation generation method for seismic response analysis of NSCs. By integrating Hilbert-Huang transformation, genetic algorithms, and single-objective optimization techniques, the proposed method generates excitation time histories that precisely match target spectra. A composite error function model and a dual-objective dynamic control mechanism are introduced to significantly enhance the convergence speed and spectral matching accuracy over the full frequency range. For numerical validation, the tri-axial acceleration response spectra with a 5% damping ratio of a typical cylindrical shell structure are adopted as the target spectra. Results demonstrate that the proposed method can accurately match the target spectra with both regular and irregular spectral characteristics while effectively preserving the non-stationary nature of generated ground motions. The proposed method provides reliable seismic input for refined performance assessment of NSCs in complex-shaped long-span structures, and also offers methodological reference for the development of excitation generation approaches optimized for response spectrum matching involving complex spectral shapes.