臺灣博碩士論文加值系統

在切削加工中，刀具狀態為影響加工品質的關鍵因素，於前期研究已達到使用五軸同動控制與顯微攝影機，從各種位置、高度與角度量測不同種類的刀具尺寸與磨耗位置。然而現今刀具磨耗大部分以二維影像來計算其磨耗面積，如何以三維方式呈現實際磨耗曲面量，實為精密量測刀具磨耗的一大挑戰。
本研究精進一離線式刀具檢視系統 (Offline Tool Inspection System, fTIS)，在刀種辨識上，使用卷積神經網路 (Convolution Neural Network, CNN) 可自動分辨更多種刀具。在刀具拍攝上，使用基於卷積神經網路中的YOLO (You Only Look Once) 與更快速的區域卷積神經網路 (Faster Region-based Convolutional Neural Network, Faster R-CNN) 來修正因馬達機械誤差所導致的拍攝偏移，並比較兩種方法的補償結果。最後，在刀具磨耗量測上，使用現有磨耗位置的長度、寬度與像素量，限制其三維模型上的座標位置來標記出實際的三維刀具磨耗。
在研究成果上，精進後的fTIS可識別9種不同種類刀具，辨識正確率達99%。而在刀具檢測上，以兩刃鑽孔刀為例，可在1.5 ~ 2秒間快速修正到實際需拍攝位置，補償準確率達97%，也可依照原始計算的二維磨耗位置標記到三維刀具模型上。因此，精進後的fTIS可讓現場操作人員以及工程師更加直觀的查看刀具磨損情況，以便作為後續加工的參考依據。

In the cutting process, the tool state is a key factor that affects the machining quality. Previous research had achieved five-axis simultaneous control and a micro camera was used to measure different tool sizes and wear positions from various positions, heights and angles. However, current tool wear uses two-dimensional images to calculate its wear area. Therefore, accurately measuring tool wear and present the actual wear surface in three-dimensions is a major challenge.
This study improves an Offline Tool Inspection System (fTIS). For tool type identification, a Convolutional Neural Network (CNN) can be used to automatically identify more tool types. In tool filming, Faster Region-based Convolutional Neural Network (Faster R-CNN) and You Only Look Once (YOLO) which is based on convolutional neural network are used to correct the filming deviation caused by motor-based machinery and compare the compensation results of the two methods. Finally, in the measurement of tool wear, the length, width and pixel amount of the existing wear position are used to limit the coordinate position on the 3D model to mark the actual 3D tool wear.
Experimental results show, the improved fTIS can recognize 9 different types of tools, with a recognition accuracy of 99%. In terms of tool detection, taking a two-edged drilling tool as an example, it can be quickly corrected to the actual filming position in 1.5 to 2.0 seconds, and the compensation accuracy is 97%. The initially calculated 2D wear position can also be marked to the 3D tool model. Therefore, the improved fTIS allows field operators and engineers to view the tool wear more intuitively, and can be used as a reference for subsequent processing.

摘要 i
Abstract ii
誌謝 iii
目錄 iv
表目錄 vi
圖目錄 vii
中英文名詞對照表 x
第1章 緒論 1
1.1 研究背景 1
1.2 研究目的 5
1.3 研究架構 6
第2章 理論方法 7
2.1 刀具磨耗分析 7
2.2 特徵檢測方法 9
2.2.1 卷積神經網路 (Convolutional Neural Network, CNN) 9
2.2.2 You Only Look Once (YOLO) 15
2.2.3 更快速的區域卷積神經網路 (Faster R-CNN) 19
2.3 最佳化方法 22
2.3.1 粒子群演算法 (Particle Swarm Optimization, PSO) 22
第3章 系統開發 24
3.1 系統架構 24
3.1.1 系統概念 24
3.1.2 系統流程 25
3.2 影像資料蒐集 27
3.3 刀具種類辨識模組 29
3.3.1 影像前處理 29
3.3.2 卷積神經網路模型之建立 30
3.4 刀具拍攝位置補償模組 32
3.4.1 影像特徵標記 34
3.4.2 基於卷積神經網路之模型建立 35
3.4.3 特徵位置座標差 36
3.4.4 實際移動量換算 38
3.5 刀具三維磨耗標記模組 39
3.5.1 三維模型投影至二維平面 39
3.5.2 基於粒子群演算法之影像比對 40
3.5.3 三維刀具磨耗標記 41
第4章 案例研究與實驗分析 45
4.1 刀具分類分析 45
4.2 刀具拍攝位置補償 52
4.2.1 實際移動量換算結果 52
4.2.2 基於模型之補償分析 53
4.3 刀具磨耗量測與標記 73
4.3.1 二維刀具磨耗 73
4.3.2 三維模型投影至二維平面 77
4.3.3 三維刀具磨耗 78
4.3.4 結果與討論 81
第5章 結論與未來展望 83
5.1 結論 83
5.2 未來展望 84
參考文獻 85

[1]T. Mikolajczyk, K. Nowicki, A. Klodowski and D.Yu. Pimenov, “Neural network approach for automatic image analysis of cutting edge wear”, Mechanical Systems and Signal Processing, Vol. 88, pp. 100–110, 2017.
[2]T. Mikolajczyk, K. Nowicki, A. Klodowski and D.Yu. Pimenov, “Predicting tool life in turning operations using neural networks and image processing”, Mechanical Systems and Signal Processing, Vol. 104, pp. 503–513, 2018.
[3]D.M. D’Addona, A.M.M.S. Ullah and D. Matarazzo, “Tool-wear prediction and pattern-recognition using artificial neural network and DNA-based computing”, Journal of Intelligent Manufacturing, Vol. 28, no. 6, pp. 1285–1301, 2017.
[4]M.T. Garcia-Ordas, E. Alegre, V. Gonzalez-Castro and R. Alaiz-Rodriguez, “A computer vision approach to analyze and classify tool wear level in milling processes using shape descriptors and machine learning techniques”, The International Journal of Advanced Manufacturing Technology, Vol. 90, no. 5-8, pp. 1947–1961, 2017.
[5]X. Wu, Y. Liu, X. Zhou and A. Mou, “Automatic Identification of Tool Wear Based on Convolutional Neural Network in Face Milling Process”, Sensors (Basel), vol. 19, no. 18, 2019.
[6]陳宇叡，“具自動分類與磨耗識別之刀具量測系統”，國立高雄科技大學電機工程研究所碩士論文，2018。
[7]C. Chen, T. Wang, D. Li and J. Hong, “Repetitive assembly action recognition based on object detection and pose estimation”, Journal of Manufacturing Systems, Vol. 55, pp. 325–333, 2020.
[8]J. Zhang, M. Karkee, Q. Zhang, X. Zhang, M. Yaqoob, L. Fu and S. Wang, “Multi-class object detection using faster R-CNN and estimation of shaking locations for automated shake-and-catch apple harvesting”, Computers and Electronics in Agriculture, Vol. 173, pp. 105384, 2020.
[9]J. Wang, P. Fu and R.X. Gao, L. Fu and S. Wang, “Machine vision intelligence for product defect inspection based on deep learning and Hough transform”, Journal of Manufacturing Systems, Vol. 51, pp. 52–60, 2019.
[10]K. Fu, Z. Chang, Y. Zhang, G. Xu, K. Zhang and X. Sun, “Rotation-aware and multi-scale convolutional neural network for object detection in remote sensing images”, ISPRS Journal of Photogrammetry and Remote Sensing, Vol. 161, pp.294–308, 2020.
[11]Z. Deng, H. Sun, S. Zhou, J. Zhao and H. Zou, “Toward Fast and Accurate Vehicle Detection in Aerial Images Using Coupled Region-Based Convolutional Neural Networks”, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 10, no. 8, pp. 3652–3664, 2017.
[12]D. Vucina, D. Bajic, S. Jozic, and I. Pehnec, “Evaluation of 3D tool wear in machining by successive stereo-photogrammetry and point cloud processing”, Tehnicki Vjesnik, Vol. 20, pp. 449–458, 2013.
[13]L. Cerce, F. Pusavec and J. Kopac, “3D cutting tool- wear monitoring in the process”, Journal of Mechanical Science and Technology. vol. 29, no. 9, pp. 3885–3895, 2015.
[14]J. Sandberg, “Development of a Laboratory Test Method for Assessment of Crater Wear Volume on Inserts for Steel Turning”, Mater Thesis of Lulea University of Technology, 2019.
[15]M.S.I. Chowdhury, B. Bose, K. Yamamoto, L.S. Shuster, J. Paiva, G.S. Fox-Rabinovich and S.C. Veldhuis, “Wear performance investigation of PVD coated and uncoated carbide tools during high-speed machining of TiAl6V4 aerospace alloy”, Wear, Vol. 446–447, p. 203168, 2020.
[16]S. Ren, K. He, R. Girshick and J. Sun, “Faster r-cnn: towards real-time object detection with region proposal networks”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137–1149, 2017.
[17]J. Redmon, S. Divvala, R. Girshick and A. Farhadi, “You only look once: unified, real-time object detection”, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 779–788, 2016.
[18]J. Redmon and A. Farhadi, “Yolov3: an incremental improvement”, 2018.
[19]C.W. Huang, L. Wei, L. Li and G. Anfu, “TF-YOLO: An Improved Incremental Network for Real-Time Object Detection”, Applied Sciences, Vol. 9, pp. 3225, 2019.
[20]R. Ramzi, E.A. Bakar and M.F. Mahmod, “Drill Bit Flank Wear Monitoring System in Composite Drilling Process Using Image Processing”, 10th International Conference on Robotics, Vision, Signal Processing and Power Applications, Vol. 547, pp. 551–557, 2019.
[21]K. He, X. Zhang, S. Ren and J. Sun, “Deep Residual Learning for Image Recognition”, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2016.
[22]G. Huang, Z. Liu, L. Van Der Maaten and K. Q. Weinberger, “Densely Connected Convolutional Networks”, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2261–2269, 2017.
[23]P. Murugan, “Feed Forward and Backward Run in Deep Convolution Neural Network”, School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore, 2017.
[24]S. Ruder, “An overview of gradient descent optimization algorithms”, 2016.
[25]C. Sullivan and A. Kaszynski, “PyVista: 3D plotting and mesh analysis through a streamlined interface for the Visualization Toolkit (VTK)”, Journal of Open Source Software, vol. 4, no. 37, p. 1450, 2019.