齒輪製造工序一般從鋼胚加工出齒型，接著進行感應硬化熱處理 以提升齒部強度，最終再進行精加工以達到齒輪精準度。為了響應節 能減碳，省略最終進行精加工工序以達到簡化齒輪製造工序之目的， 針對齒輪於感應硬化熱處理後之特徵進行預測成為實現此目的之重 要工具。本研究將提出應用均勻實驗設計法於圓柱齒輪感應淬火實驗 規劃，並透過徑向基函數建立預測模型以預測齒輪於感應硬化熱處理 後之特徵，且推導出各目標預測方程式。此外，本研究利用有限元素 法模擬齒輪感應淬火，提出使用非均勻磁場模型以獲得更準確的結 果。隨後，透過粒子群最佳化方法求解特定齒輪特徵之加熱控制因數 最佳解。最後，應用最佳解於有限元素模擬出非均勻磁場模型之模擬 結果，並與預測模型預測結果進行比較以驗證非均勻磁場模型之正確 性。同時，透過非均勻磁場模型與均勻磁場模型之模擬結果與預測模 型預測結果進行比較，以驗證使用非均勻磁場模型之必要性。總結， 本研究提出預測模型較適用於預測齒輪感應硬化熱處理特徵之工具。;Gear manufacturing processes typically begin with shaping the teeth from a steel blank, followed by induction hardening to enhance tooth strength. Subsequently, precision machining is performed to achieve gear accuracy. In response to energy conservation and carbon reduction, the final precision machining step is omitted, simplifying the gear manufacturing process. To achieve this goal, predicting the characteristics of gears after induction hardening becomes a crucial tool. This study proposes the application of uniform experimental design for experimental planning of cylindrical gear induction hardening. A predictive model is established using radial basis functions to anticipate gear characteristics after induction hardening, with the derivation of respective target prediction equations. Additionally, finite element analysis is employed to simulate gear induction hardening, advocating the use of a non-uniform magnetic field model for enhanced accuracy. Subsequently, particle swarm optimization is utilized to determine optimal heating control factors for specific gear features. Finally, applying the optimal solution to finite element simulation produces results for the non-uniform magnetic field model, validated against predictive model outcomes. A comparison between non-uniform and uniform magnetic field models and predictive model results validates the necessity of employing a non-uniform magnetic field model. In conclusion, this study proposes that the predictive model is more suitable for forecasting characteristics in the induction.
