本研究開發可搭配一部全自動自行車架檢測及校正機器所迫切需要的高效率專家系統，協助判斷包含金屬彈塑性現象的自行車車架校正過程中實際對應不同變異量所需要的車架校正調整量，以突破自動化自行車車架檢測及校正設備開發的瓶頸。
本系統於建置階段以有限元素法進行彈塑性分析，模擬自行車製造廠中人工對歪斜的車架進行校正的過程，並以分析結果在資料庫中建立數種不同的自行車車架尺寸，其對應的歪斜量及應校正量，以此作為資料樣本建立預測用的數學模型。此數學模型除輸出建議校正調整量以外同時接收來自數控設備的回授資料，以機構對各種尺寸車架校正的結果即時在資料庫中修正相應的校正量預測模型。在累積足夠實驗次數和車架種類後即可受益於資料庫內存有大量實驗結果，而達到使設備校正車架時間縮短之效果。
本系統以通訊連接合作廠商提供的數值化控制設備，並設計實驗對三種廠商提供之不同尺寸之鋁合金自行車車架各至少30隻及電動車車架20隻以檢視系統之校正效果。經實驗結果證實，本系統校正速度良好，實驗同時證實車架前製程品質對系統校正速度影響極小，即便兩批車架校正前之尺寸精度(CP)及準度(CK)有所差異，校正平均時間差異仍不足20%，顯示本系統校正速度不受前製程品質之影響。相較於人工的經驗累積，本系統以資料庫儲存校正資料校正時所記錄數據，因此不論員工經驗多寡皆可以相同的速度完成車架校正，且校正前後的尺寸數據亦可以產生每批車架的詳細尺寸資料及歪斜狀況，用於管制品質並改善前製程，對生產排程管理為相當大的助益。

In this research, an automatic bicycle frame measurement and alignment expert system for efficiently supporting a fully automated bicycle frame inspection and fixing machine is developed. This expert system solved the uncertainty of required application forces to match the varied amount of dimensional errors during the bicycle frame alignment process, which is complicated due to involvement of elastoplastic problems.
In the research, Finite Element Analysis is used to simulate the elastoplastic problem of bicycle frame alignment, and the result of the simulation shows the relationship between the dimension of bicycle frames and force used to correct askew frames. After simulating, the data is recorded in the database as a sample, which is used for establishing a correction force prediction function. This prediction function is the main function of expert system, and system connected to a numerical bicycle frame alignment machine. The machine is able to implement the correction force which is calculated by the expert system during the process of alignment. After implementing the correction force, the result of correction will be sent back to the expert system and replace the simulated data in the database. Benefit from the feedback data, the expert system is expected to become more accurate after enough experiments.
To verify the ability of the system, 3 types of bicycle frames and 2 types of electronic bicycle frame with different dimension is used in the test. Test has been done on 30 samples of each types of bicycle and 20 samples of each electronic bicycle, the result shows that the system performs stably during the test, and the time consumption to alignment each single sample remains to be short. A continuous test on 3 batches of samples with same dimension but different askew is done. The result shows that the system is able to finish the task when the accuracy and precision is poor, and time consumption only increases 20 percent when the quality decline.
Compare to manual alignment, this system is free from effected by experience and skill of the workers. In the other hand, aliment expert system records the data during the process, and the data can be used to improve the weld process. The system is expected to be a helpful tool to the management of production schedule and the quality control.

[1] B. Kehoe, S. Patil, P. Abbeel, and K. Goldberg, "A Survey of Research on Cloud Robotics and Automation," IEEE Transactions on Automation Science and Engineering, vol. 12, no. 2, pp. 398-409, 2015.
[2] 謝宗恩, "六軸手臂自動循線運動系統開發," 碩士, 機械工程系, 國立臺灣科技大學, 台北市, 2018.
[3] 林冠甫, "基於影像伺服之輸送帶上物件自動追蹤系統," 碩士, 機械工程系, 國立臺灣科技大學, 台北市, 2016.
[4] M. Wilkinson, Winston Churchill Travelling Fellow. (2011). Bicycle Frame Building in the USA. Available: https://www.wcmt.org.uk/sites/default/files/migrated-reports/891_1.pdf
[5] 張紹勳, 人工智慧與專家系統. 台灣: 松崗電腦圖書資料股份有限公司, 1993.
[6] 江孟峰, 專家系統. 台灣: 文魁資訊股份有限公司, 2002.
[7] C.-C. Lin, S.-J. Huang, and C.-C. Liu, "Structural analysis and optimization of bicycle frame designs," (in English), Advances in Mechanical Engineering, vol. 9, no. 12, 2017/12/16/ 2017.
[8] L. Huei-Huang, Finite Element Simulations with ANSYS Workbench 15. 台灣: 全華圖書股份有限公司, 2014.
[9] 楊榮顯, 工程材料學. 台灣: 全華圖書, 2012.
[10] 劉晉奇 and 褚晴暉, 有限元素分析與ANSYS的工程應用. 台灣: 滄海書局, 2016.
[11] D.L.Logan, A First Course in the Finite Element Method. Boston: PWS-KENT, 2012.
[12] R.C.Hibbeler, Engineering Mechanics STATICS. U.S.A.: Pearson, 2017.
[13] M. Taya, K. Lulay, K. Wakashima, and D. Lloyd, "Bauschinger Effect in Particulate Sic-6061 Aluminum Composites," (in English), Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, vol. 124, no. 2, pp. 103-111, 1990/04/20/ 1990.
[14] T. W. Kirkman. (1996). Least Squares Fitting (Regression).
[15] 成功大學統計學系. (2002). The Method of Least Square. Available: http://estat.ncku.edu.tw/topic/line_stat/base/lse.html
[16] 王栢村. (2017). 質塊彈簧系統「自然頻率」(natural frequency)？. Available: http://aitanvh.blogspot.com/2017/07/natural-frequency.html?m=0