Abstract: To address the challenges of file transfer time and cost in large-scale engineering projects, a workflow-based file management optimization method for engineering projects is proposed. First, the file management environment of engineering projects and the file workflow model with logical order were established, and the transmission and caching of files were analyzed. Building on this, the file management optimization problem was modeled as a Markov process, and the joint optimization of task completion time and caching cost of file workflow was achieved by designing the state space, action space and reward function. Second, dueling double deep Q network (D3QN) was employed to reduce the training time and improve the training efficiency. Simulation results verify the effectiveness of the proposed scheme for file transfer under different parameter configurations, and it can still maintain a good optimization ability when the task volume increases.