Abstract Although the bladed disks of turbomachinery are nominally designed to be cyclically symmetric (tuned system), the vibration characteristics of individual blades on a disk vary slightly owing to manufacturing tolerances, deviations in material properties, wear during operation, etc. These small variations break the cyclic symmetry and split eigenvalue pairs. Actual bladed disks with small variations are called mistuned systems. The resonant stress on a mistuned bladed disk may become large and cause a blade failure due to high cycle fatigue. Especially for a bladed disk with a free-standing blade structure, it is essential to use a sufficiently large safety factor at the design stage, considering the mistuning effect. Traditionally, blade designers have adopted various countermeasures to reduce the resonant stress at the design stage. Besides the countermeasures adopted in blade design, for a variable speed engine, which cannot avoid resonance, it is thought that a practical optimization method for a mistuned system (bladed disk with a free-standing blade structure) is to sort the blades so that the resonant stress is minimized. In this study, a simultaneous method for optimizing the blade resonant stress and the amount of unbalance causing rotor vibration is proposed. In this method, first, the natural frequencies and weights of all blades on a disk are measured. Then, a mistuned system is assembled and the analysis model is generated. Next, the resonant stress and the amount of unbalance in the mistuned system are analyzed. To reduce the computation time, the reduced-order model known as fundamental mistuning model (FMM) is used to calculate the resonant stress in the mistuned system. The analyses of the resonant stress and the amount of unbalance are carried out repeatedly, sorting the blades on the disk, and the optimal solution is explored using Monte Carlo simulations (MCS) or discrete differential evolution (DDE). As an example, a mistuned bladed disk of an aero-engine was analyzed, and the validity of the proposed method was verified.