Abstract: In this study, we focus on the records with permanent displacement and (or) velocity pulse and compared the impact of different process procedures on structural responses. We firstly classified the strong ground motion into four categories and picked typical records from different earthquake events, e.g. 2018 Taiwan Hualien earthquake, 2014 Ludian earthquake in Yunnan, and the 2011 Tokyo earthquake in Japan. These records were processed using different filtering and baseline correction process methods. The corresponding SDOF/MDOF inelastic demand and EDPs of RC frames were computed and compared. For the records with pulse-like velocity and (or) permanent displacement, the Butterworth filtering method will significantly distort the pulse portion and erase the permanent displacement, which will further influence structural responses regardless of the structural fundamental periods. For this kind near-fault ground motion, it is recommended to use the baseline trend line correction processing method before structural time-history analysis. For ground motions without pulse-like velocity and permanent displacement, the process procedure have a negligible effect on structural responses. In order to eliminate the effect of causal filtering on the time-history’s phase spectrum, we propose a batch-processing procedure based on a causal Butterworth filtering method.