Abstract: To enhance the efficiency of dynamic testing, a method to reconstruct dynamic responses of cable dome structures considering step excitations was developed, with a corresponding strategy to screen high-precision unmeasured points suggested. Dynamic responses were approximated by linearly combining several contributing modes, and the deployment of measured points was optimized by using the effective independent method. Before reconstructing dynamic responses, the low-pass filtering (FIR), the variational mode decomposition (VMD) and the singular value decomposition (SVD) were jointly utilized to treat measured signals, so that the unfavorable influences of high-order modes, of measurement noise and, of low-order non-contributing modes on the reconstruction precision could be eliminated as far as possible. After the low-pass filtering of measured signals, the best number of decomposition layers of VMD was determined according to the dispersion degree of center frequencies of intrinsic mode functions (IMFs). The VMD was subsequently conducted to the signals, and the effectively triggered modes were selected as per the smoothness of frequency curves of IMFs. These triggered modes were further treated by the SVD and finally superposed to obtain the preprocessed signals of which the signal-to-noise ratios were significantly improved. The preprocessed signals were used to reconstruct dynamic responses, the error mechanism of dynamic response reconstruction was carried out, and a screening strategy based on the amplitudes of reconstructed signals was suggested to obtain high-precision unmeasured points. A Geiger cable dome structure with a span of 100 m was taken as the example and numerically analyzed. The analysis results show that: the method proposed is suitable to a real structure with random parameter deviations, is able to acquire high-precision unmeasured points, and can provide more usable testing information for subsequent following works like modal identification, modal updating, damage detection, and etc.