Thin-walled tubes are classic energy-absorbing structures, but the objective of crashworthiness topology optimization for energy-absorbing structures is to determine the optimal configuration under limited mass and specific boundary conditions. However, the direct optimization process often requires specific mass constraints. Furthermore, the lack of explicit identification of plastic hinges leads to an unstable energy absorption process, which may fail to satisfy the requirements for progressive deformation. To achieve progressive energy absorption, this study proposes a hybrid crashworthiness topology optimization method for thin-walled tubes based on relative element density. In the first step, under given multi-stage force boundary, the thin-walled tube is segmented according to progressive intervals. The mass constraints of the different segments are calibrated to determine the feasible range of multi-stage stiffness. For the manufacturing constraints, ESLM is utilized to perform stiffness-based topology optimization on the different segments. In the second step, elements within the initial topology optimization results are removed based on relative density sorting, and the optimized configuration is determined through multi-sample active learning. This hybrid optimization process is validated based on the numerical examples and optimization results, which indicate the feasibility of the proposed method and focus on exploring the convergence rate of crashworthiness optimization for energy-absorbing structures.