In view of robustness of objective function and constraints in robust design, a new robust design method based on maximal variation analysis was proposed. Firstly, the principle of variations which have been generated in objective functions and constraints by considering the uncertain factors in design variables and design parameters were analyzed, then the maximal variations of objective function and constrains were estimated by using maximal variation analysis. The maximal variations of constraints were added to original constraints to guarantee feasibility of constraints; a new constraint has been added to original optimization problem to ensure the variation of objective function was less than the value which designer set, and then a bi-level mathematical optimal model was constructed. The top-level optimization was used to solve the original mathematical model; the lower-level optimization was used to judge the robustness of objective function and constraints. Example results showed that the proposed approach is feasible.