Abstract: The development of efficient seismic damage simulation methods for urban building clusters is critical for advancing urban resilience against earthquakes. This study addresses the fundamental challenges of balancing computational scale and efficiency in nonlinear dynamic time-history analysis of large-scale urban building clusters by proposing a distributed parallel computing framework based on the Finite Particle Method (FPM). The framework systematically resolves several critical aspects of building collapse simulation such as: geometric nonlinearities, material nonlinearities, and component fractures, etc. A novel parallel architecture is implemented through the Spark distributed computing platform, incorporating core innovations such as adaptive task scheduling strategy for heterogeneous building computations, chunk-based distributed storage mechanism for finite element data, and partitioned parallel solver for particle motion equations, etc. The resulting Spark-FPM system demonstrates exceptional scalability in numerical experiments involving a 100-building urban cluster model (68,328 elements), achieving computation times of 0.342 hours for 10⁴ time steps and 2.916 hours for 10⁵ time steps. Comparative experiments with a single-node FPM program reveals that Spark-FPM achieves a maximum speedup ratio of 1247 times, significantly enhancing the computational efficiency in seismic disaster simulations of urban building clusters. This research provides an efficient solution for nonlinear dynamic response analysis of city-scale building clusters.