This paper presents a novel practical multiscale analysis (MSA) method for simulating local damages in large reinforced concrete (RC) structures during seismic events. Traditional techniques like the finite element (FE) method often struggle to balance computational costs with detailed simulation. The proposed method facilitates sequential local refinement from macroscopic to mesoscopic scales of RC structures, focusing strategically on critical areas prone to damage and cracking. Key features include: (1) 'Correction forces' enable integration and collaboration among models of different refinement levels, keeping models unchanged (i.e., without additions, deletions, or further refinement of elements), streamlining analysis and independent implementation across methods, scales, and platforms. (2) A smart replacement strategy ensures smooth transitions and prevents abrupt force changes between models at different refinement levels. (3) A ‘training process’ before performing the replacement and a semi-explicit method guarantee the accuracy and efficiency of the MSA method. (4) Parallel and distributed computing is seamlessly applied, significantly accelerating the analysis at each level. Implemented in the open-source software OpenSees, this method is illustrated through three examples that efficiently capture both the macroscopic mechanical responses and detailed local damage behaviors. This approach provides a valuable tool for the refined analysis of large-scale RC structures.