<p>The structure design and material selection of curved aluminum profiles for automotive body usually ignore the formability of subsequent bending process. Besides, hollow aluminium profile undergoes wall thickness variation and work-hardening in bending, which are ignored in crashworthiness analysis and optimization, causing redundancy in performance design. Collaborative optimization of structure, material, process and performance is an effective way to solve this issue. In this study, a 3D elasto-plastic FE model was developed for simulating the forming and unloading processes of bumper beam stretch-bending using dynamic explicit and static implicit algorithms. Then, the FE analysis results of stretch-bending were transferred to the impact FE simulation through isoparametric element transformation. The influences of forming effects on impact performances of bumper subsystem were identified. To control the forming defects in stretch-bending of bumper beam while meeting the requirements of impact performance and lightweight, a multi-objective collaborative optimization method based on sequential sampling approximate model and non-dominated sorting genetic algorithm was proposed. Three different methodologies of constructing approximate models for the optimization problem were compared and then a combined approximation technique for five constraints and objectives was proposed. The sensitivity of structure, material and process parameters on the bending forming defects and impact performance was determined by Sobol’s global sensitivity method. After optimization, the cross-section distortion and weight of bumper beam are respectively reduced by 70.22% and 13.55%, while most other performances are also slightly improved. The stretch-bending experiment and optimization results are in good agreement, which validates the effectiveness and feasibility of the proposed multi-objective collaborative optimization method.</p>