Abstract: In order to accurately identify the modal parameters of structures from noisy signals monitored by sensors, a combination algorithm of random decrement technique (RDT) and the improved variational mode decomposition (IVMD) was presented. The simulation results indicate that the IVMD has more advantages in theory and solves the shortcomings of the traditional VMD algorithm. The proposed algorithm used RDT to obtain the free decay response (FDR) with noise directly from the original monitoring signal, without denoising the original signal first, which improves the analysis efficiency. The denoised FDR was acquired by IVMD. The structural modal frequencies and damping ratios were obtained based on the denoised FDR. The RDT-IVMD algorithm is applied to analyze the displacement of a super-high-rise building and the acceleration of a suspension bridge, respectively. The results show that RDT-IVMD can efficiently and precisely detect frequencies and damping ratios of two large-scale structures, and the reliability and superiority of the method are proved by comparative analysis. The results can offer a basis for structural monitoring and diagnostic research.