影響係數法的平衡面及量測點的位置及數量，是否影響計算校正質量的精度，使複雜轉子之影響係數構成病態矩陣，導致計算結果對量測誤差過於敏感。由於條件數可表達矩陣病態，因此，得以影響係數矩陣的條件數為動平衡計算誤差最小化的目標函數。基於影響係矩陣本身為目標尋優，可以先行得到最優的平衡面及量測點位置及數目，並免於計算平衡誤差之程序，符省優步驟，提高效率，因此本文提出以影響係數條件數最小化為目標函數，以降低動平衡結果誤差的方法，並使用遺傳演算法及禁忘搜尋法求解，得到最佳精度之平衡面位置，並以數值模擬及實驗驗證撓性多盤轉子，分析其平衡前後之穩態反應與平衡面及矩陣條件數之關係，得到在相同矩陣維度下，最小條件之影響係數矩陣，可獲得最佳的動平衡結果。

For the influence coefficient method the influences of locations of balancing planes and/or measurement sensors, on the balancing accuracy are very significant, while the influence coefficient matrix for a complex rotor-bearing systems has an inclination to ill condition and then lead to unavailable computation. Besides, the computation of correction masses would be excessively sensitive to measurement errors. This paper utilizes Genetic algorithm and Tabu search to minimize the condition number of influence coefficient matrix for the optimal locations of balancing planes, and utilizes finite element method to simulate multidisk and flexible rotor-bearing systems to determine the steady-state response. Finally, the experiments and numerical simulations verify the promotion of balancing accuracy by proposed optimization for flexible rotor-bearing systems.