混接面(又稱圓角面)特徵為CAD(Computer-Aided Design)模型中常見特徵之一，雖然此特徵有利於解決部份工程問題，如應力集中、引導組裝、觸感等，但同時也衍生出其他的問題，在網格化時，容易產生劣質網格、在進行特徵辨識時，使拓樸結構複雜化。因此本研究擬發展一技術能夠辨識出CAD模型中的混接面，並應用於以上各方面。混接面辨識的程序中，使用者必須輸入閾值，之後由各面的平滑連續性重建拓樸結構並計算曲面的各式參數。根據以上的資訊，程式利用判斷條件篩選出混接面，而後經由模型中實際的混接情形來制定分類基準，並將混接面分類。最後藉著與商用辨識軟體比較測試結果來驗證本研究之正確性和探討混接面辨識演算法於眾多案例中的辨識情形，並點出目前辨識結果中具有問題處供後續研究參考。;Blend face (also called fillet) is one of the most viewed features in CAD models. While it benefits a CAD model for assembling or solving stress concentration, a fillet simultaneously causes other problems such as low quality in mesh generation or more complex topology in feature recognition. The aim of this study is to present a method for the recognition of blend faces in order to solve the problems as mentioned above. Through the input of a threshold, the proposed procedure of the blend faces recognition is to build the topological data of edges and faces, and compute some important parameters from faces. By these data, it can recognize blend faces from the CAD model and classify them. Several industrial parts are presented to verify the feasibility and accuracy of the proposed method.
