由3D零件的幾何模型辨識加工特徵為自動化生產的一項重要工作，傳統的特徵辨識方法無法完整辨識出具有相交特徵的幾何模型，本論文以2D影像的深度學習進行3D CAD模型的加工特徵辨識，以解決傳統辨識方法對於相交特徵的辨識問題。在研究方法方面，本論文首先以體素網格的方式將零件的3D CAD模型轉換為網格陣列，並藉由改變網格陣列的結構排列獲得不同型態的特徵形式，接著以網格陣列建立深度學習訓練所需的圖像資料集，以利產生2D深度學習模型；將零件的CAD模型輸入至訓練完成的深度學習模型進行辨識，即可判別加工特徵的類別及其座標位置，接著藉由座標分割判定，將具有相交的座標進行分割，並以分割後的座標中心點在CAD模型中搜尋相對應的面，最後藉由凹凸邊搜尋的方法找出特徵的所有組成面。
本論文除了詳述如何將3D CAD 模型轉換為2D圖片、改變網格陣列結構產生不同形式的特徵圖形、建構及訓練深度學習模型、分割預測座標、以中心點及凹凸邊法搜尋特徵的組成面以外，也利用兩個案例驗證所開發系統的實用性。

Recognition of machining features is an essential task for manufacturing automation in industry. In general, traditional rule-based feature recognition approaches are inflexible and computationally expensive. It is also quite a challenging task to design proper rules to recognize intersecting features. In order to overcome these problems, this thesis approaches the recognition of machining features by using deep learning of 2D images. At first, the 3D CAD model of a part is converted to grid arrays through the introduction of volume pixels. Different types of machining features can also be obtained by changing the structural arrangement of grid arrays. Subsequently, using the grid arrays as data sets to train the deep learning network and the trained deep learning model can then be used to recognize from an input CAD model the types of machining features and their corresponding coordinates. The system will segment coordinates of intersecting features and then use the center points of coordinates to locate the corresponding surface in the CAD model. Finally, the entire composing surfaces of the feature can be found by using the concave and convex information among edges of surfaces.
In addition to proposing the methodologies for converting 3D CAD model into 2D images, changing grid arrays’ structural arrangement, generating training data sets for deep learning, segmenting feature coordinates and searching for feature surfaces, this thesis uses two case studies to verify the practicability of the developed system.

[1] Grayer, A.R. (1977), Computer link between design and manufacture, Ph.D Thesis, Department of Computer Science and Technology, University of Cambridge, Cambridge, UK.
[2] Shah, J.J., Anderson, D., Kim, Y.S., and Joshi, S. (2000), “A Discourse on Geometric Feature Recognition From CAD Models,” ASME. J. Comput. Inf. Sci. Eng.,Vol. 1, No. 1, pp. 41–51.
[3] Shi, Y., Zhang, Y., Xia, K., and Harik, R. (2020), “A critical review of feature recognition techniques,” Computer-Aided Design and Applications, Vol. 17, No. 5, pp. 861-899.
[4] Joshi, S. and Chang, T. (1988), ‘‘Graph-based heuristics for recognition of machined featured features from a 3D solid model,’’ Computer-Aided Design, Vol. 20, No. 2, pp. 58–66.
[5] Corney, J.R. (1993), Graph-based feature recognition, Ph.D Thesis, Heriot-Watt University, Edinburgh, UK.
[6] 張書庭，「自動建構木工樑柱之製程規劃系統」(2021)，碩士論文，臺灣科技大學機械工程系，台北市。
[7] Gavankar, P., and Henderson, M.R. (1990), “Graph-based extraction of protrusions and depressions from boundary representations,” Computer-Aided Design, Vol. 22, No. 7, pp. 442-450.
[8] 陳宇昇，「木屋樑柱之特徵辨識及特徵加工路徑的產生」(2021)，碩士論文，臺灣科技大學機械工程系，台北市。
[9] Verma, A.K., and Rajotia, S. (2008), “A hint-based machining feature recognition system for 2.5 D parts,” International Journal of Production Research, Vol. 46, No. 6, pp. 1515-1537.
[10] Kim, Y.S. (1992), “Recognition of form features using convex decomposition,” Computer-Aided Design, Vol. 24, No. 9, pp. 461-476.
[11] Sakurai, H. (1995), “Volume decomposition and feature recognition: Part 1—polyhedral objects,” Computer-Aided Design, Vol. 27, No. 11, pp. 833-843.
[12] Shi, Y., Zhang, Y., Xia, K., and Harik, R. (2020), “A critical review of feature recognition techniques,” Computer-Aided Design and Applications, Vol. 17, No. 5, pp. 861-899.
[13] Ding, L., and Matthews, J. (2009), “A contemporary study into the application of neural network techniques employed to automate CAD/CAM integration for die manufacture,” Computers & Industrial Engineering, Vol. 57, No. 4, pp. 1457-1471.
[14] Shi, P., Qi, Q., Qin, Y., Scott, P.J., and Jiang, X. (2020), “A novel learning-based feature recognition method using multiple sectional view representation,” Journal of Intelligent Manufacturing, Vol. 31, pp. 1201-1309.
[15] Prabhakar, S., and Henderson, M.R. (1992), “Automatic form -feature recognition using neural-network-based techniques on boundary representations of solid models,” Computer-Aided Design, Vol. 24, No. 7, pp. 381-393.
[16] Ding, L.; Yue, Y.: Novel (2004), “ANN-based feature recognition incorporating design by features, ” Computers in Industry, Vol. 55, No. 2, pp. 197-222.
[17] Lambourne, J.G., Willis, K.D., Jayaraman, P.K., Sanghi, A., Meltzer, P., and Shayani, H. (2021), “BRepNet: A topological message passing system for solid models,” In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 19-25, 2021, pp. 12773-12782.
[18] Babić, B.R., Nešić, N., and Miljković, Z. (2011), “Automatic feature recognition using artificial neural networks to integrate design and manufacturing: Review of automatic feature recognition systems,” AI EDAM, Vol. 25, No. 3, pp. 289-304.
[19] Y. Zhang, Y. Zhang, K. He, D. Li, X. Xu, Y. Gong (2021), “Intelligent feature recognition for STEP-NC-compliant manufacturing based on artificial bee colony algorithm and back propagation neural network”, J. Manuf. Syst.
[20] Shi, Y., Zhang, Y., and Harik, R. (2020), “Manufacturing feature recognition with a 2D convolutional neural network,” CIRP Journal of Manufacturing Science and Technology, Vol. 30, pp. 36-57.
[21] Z. Zhang, P. Jaiswal, R. Rai, (2018), “Featurenet: Machining feature recognition based on 3d convolution neural network, ” Comput Aided Des., Vol. 101, pp. 12-22,
[22] P. Shi, Q. Qi, Y. Qin, P.J. Scott, X. Jiang, (2021), “Intersecting Machining Feature Localization and Recognition via Single Shot Multibox Detector, ” IEEE Trans. Ind. Inf., Vol 17, NO. 5,pp. 3292-3302
[23] Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C-Y. and Berg, A.C. (2016), “SSD: Single Shot MultiBox Detector,” European conference on computer vision, Oct. 11-14, 2016, Amsterdam, The Netherlands, pp. 21-37.
[24] Train/Val Accuracy, (2019), Retrieved from https://cs231n.github.io/neural-networks-3/#accuracy
[25] Binvox, Retrieved from https://www.patrickmin.com/binvox/
[26] PeizhiShi, “SsdNet,” Retrieved from https://github.com/PeizhiShi/SsdNet, 2018
[27] NX Open (2014), Siemens PLM Software Inc., Plano, TX, USA.