As the numbers and running distance of Chinese high-speed trains increase, many electric multiple units (EMU) gradually enter into overhaul stage, EMU maintenance bases face challenges of transition from practical exploration, regular production to lean production. In accordance of business requirement, built a model of dynamic resource allocation, task splitting, and soft precedence constraints. By the design of nonlinear decline inertia weight factor, a refined particle swarm optimization (PSO), as well as the corresponding parallel transformation scheme from particle to schedule, is presented. Finally, computational analysis is performed to validate the model and algorithm on optimization capabilities, resource utilization and performance.