臺灣博碩士論文加值系統

由於現代科技發展快速，傳統使用有限元素法(Finite Element Method,FEM)求取主軸軸承系統自然頻率相當的耗時與耗力，目前少有研究運用人工智能(Artificial Intelligence,AI)系統探討自然頻率這方面的問題。因此本研究擬建立一套人工智能系統，系統本身以機器學習中的迴歸為基本架構，運用Matlab建立神經網路系統，將設計變數輸入，透過神經網路訓練，最後獲得自然頻率值。並利用Matlab建構有限元素法模型，求出自然頻率值，兩者加以作比較，來驗證人工智能系統的準確性。
本研究先使用統計的迴歸探討六種軸承組組合中，整體設計變數與自然頻率的關係為何，接著再使用兩種迴歸模型為線性迴歸、高斯過程迴歸，並將資料75%作為訓練資料，其餘當作驗證資料，探討兩種迴歸模型的預測準確度。

As technology develops, it is time-consuming to measure natural frequency of the spindle bearing system with finite element method (FEM). Currently, there are few research in measuring natural frequency with artificial intelligence (AI). The purpose of this research is to establish an AI system which framework is regression analysis for machine learning. We build a neural network system by using Matlab, input related parameters, and get natural frequency as output through neural network training. We also calculate the frequency with FEM by the same software.
This study first uses statistical regression to explore the relationship between the overall design variables and natural frequencies in the six bearing group combinations, and then uses two regression models as linear regression and Gaussian process regres-sion. A total of 75% is used as training data, and the rest are considered verify the da-ta and discuss the prediction accuracy of the two regression models.

目 錄

第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 1
1.3 研究流程 2
第二章 文獻探討 4
2.1 主軸軸承系統自然頻率 4
2.2 機器學習 4
2.3 交叉驗證 5
第三章 研究方法 6
3.1 研究步驟 6
3.2 迴歸模型 6
3.3 有限元素法動態模型 7
3.4 參數設定 9
3.5 K等分交叉驗證 9
3.6 數據轉換 10
第四章 研究結果與分析 12
4.1 數據來源 15
4.2 軸承組整體設計變數與自然頻率關係 16
4.3 線性迴歸預測模型 16
4.4 高斯過程迴歸預測模型 20
4.5 結果分析與比較 24
第五章 結論 25
參考文獻 26
附錄 28

[1]行政院，生產力4.0發展方案，2016。
[2]林巧盈，探討K等分交叉驗證法對於分類器錯選率之研究，國立成功大學，碩士論文，2014。
[3]羅華強，類神經網路－MATLAB的應用。高立圖書，2011。
[4]A. L. Samuel,“Some studies in Machine Learning using the game of checkers,” IBM. Journal 3, pp. 211-229, 1959.
[5]C.-W. Lin and J. F. Tu,“Model-based design of motorized spindle systems to im-prove dynamic performance at high speeds, ” J. Manuf. Process, vol. 9, no. 2, pp. 94-108, 2007.
[6]C.-W. Lin,Y.-K. Lin, and C.-H. Chu, “Dynamic models and design of spindle-bearing systems of machine tools: A review, ” Int. J. Precis. Eng. Manuf, vol. 14, no. 3, pp. 513-521, 2013.
[7]Crosby, H., P. Davis, T. Damoulas, and S.A. Jarvis. “A Spatio-Temporal, Gaussi-an Process Regression, Real-State Price Predictor.” In Proceedings of the 24th ACM SIGSPATIAL International Conference on Advances in Geographic Infor-mation Systems. New York: ACM, 2016.
[8]D. J. Inman, Engineering Vibrations. Pearson Education, Inc., 2008.
[9]Geisser, Seymour,“Predictive Inference：An Introduction. Florida,”CRC Press, 1933.
[10]Haiqin Yang, Laiwan Chan, and Irwin King, “Support Vector Machine Regression for Volatile Stock Market Prediction, ” Intelligent Data Engineering and Automated Learning, pp. 391-396, 2002.
[11]K. J. H. Al-Shareef and J. A. Brandon, “On the effects of variations in the design parameters on the dynamic performance of machine tool spindle-bearing systems, ” Int. J. Mach. Tools Manuf, vol.30,no. 3, pp. 431-445,1990.
[12]M. I. Jordan and T. M. Mitchell, “Machine learning: Trends, perspectives, and prospects,”Science, vol. 349, no. 6245, pp. 255-260, 2015.
[13]P. Kim, MATLAB Deep Learning–With Machine Learning, Neural Networks and Artificial Intelligence. Apress, 2017.
[14]W. S. McCulloch and W. Pitts, “A logical calculus of the ideas immanent in nervous activity, ” Bull. Math. Biophys, vol. 5, no. 4, pp. 115-133, 1943.