Subways, as one of the critical structures in urban design for solving traffic problems, require continuous monitoring during operation. Therefore, it is imperative to study the scattering of seismic waves caused by the presence of these underground structures. System identification is one of the most efficient methods for investigating structural behavior during operation and identifying potential problems that may be encountered for an in-service underground structure. In this study, through extending an advanced system identification method called Frequency Response Function Matrix and Singular Value Decomposition (FRFM-SVD) and wavelet transform methods, modal parameters of a soil-subway system were identified. Further, using a random vibration method based on coherence functions, non-uniform accelerations under site condition were generated and the seismic behavior of soil-subway model under uniform and non-uniform seismic excitations were studied. The obtained results indicated that the FRFM-SVD method developed in this study shows a reasonable accuracy for identifying the seismic behavior of underground structures. The maximum amplification of the responses will occur in the proximity of the subway to the amount of 1.75 under non-uniform excitation, which is almost 20% greater than that from uniform excitation. Thus, to increase the accuracy of the estimation of ground surface accelerations, the effect of the presence of underground structures should be accompanied by considering non-uniform excitation in the analysis.
