隨著全球能源危機和環境問題的日益加劇，綠色能源的需求逐漸成為全球焦點。振動能量擷取技術（Vibration Energy Harvesting, VEH）作為一種新興的發電方式，具有占地小、可應用範圍廣等優勢，特別適合應用於旋轉機械產生的振動能量回收。本研究針對不穩定旋轉機械的振動發電效益進行了深入探討，選用Jeffcott轉子作為系統模型，分析其在不同質量、轉速及偏心距離條件下，如何通過壓電材料將振動能量轉換為電能。
    本研究透過建立運動與壓電耦合方程，並以四階Runge-Kutta法進行數值模擬，共產生270萬筆資料，作為機器學習模型訓練用的資料庫。接著設計實驗裝置驗證理論電壓值，理論與實測誤差皆控制在5%以內，證實模型準確性。
在機器學習方面，採用監督式學習並建立三種模型：深度神經網路（DNN）、長短期記憶網路 （LSTM）與極限梯度提升模型（XGBoost）。結果顯示三種模型皆具高準確性（決定係數R²皆超過0.999），其中DNN預測表現最佳（R² = 0.99999893），而XGBoost則具備顯著的訓練效率優勢，適合即時預測應用。本研究亦發現高電壓區間因樣本較少而造成預測誤差較大，顯示數據分布對模型效能具關鍵影響。本研究不僅展示了機器學習在振動能擷取系統中的潛力，還為後續工業應用中的能量回收系統設計提供了寶貴的數據和理論支持。

With the intensifying global energy crisis and environmental concerns, the demand for green energy has become a central focus worldwide. Vibration Energy Harvesting (VEH) has emerged as a promising power generation technology due to its small footprint and broad applicability, particularly for recovering vibrational energy from rotating machinery. This study investigates the energy harvesting efficiency of unstable rotating systems using the Jeffcott rotor as the model, analyzing how vibration energy is converted into electricity through piezoelectric materials under varying mass ratios, rotational speeds, and eccentricities.
The system's motion and piezoelectric coupling equations are established and solved using the fourth-order Runge-Kutta method, producing a dataset of 2.7 million entries for machine learning model training. An experimental setup is also developed to verify the theoretical voltage output, showing an error margin within 5%, thus confirming the accuracy of the theoretical model.
Supervised machine learning is employed to construct three models: Deep Neural Network (DNN), Long Short-Term Memory (LSTM), and Extreme Gradient Boosting (XGBoost). Results show all models achieve high prediction accuracy (R² > 0.999), with DNN yielding the best performance (R² = 0.99999893). XGBoost offers superior training efficiency, making it suitable for real-time applications. It is observed that prediction errors increase in high-voltage ranges due to sparse data distribution, highlighting the importance of data balance.
This study not only demonstrates the potential of machine learning in VEH systems, but also provides valuable data and theoretical support for the future design of energy recovery systems in industrial applications.

目錄	IV
表目錄	VI
圖目錄	VII
第一章 緒論	1
一、1.研究動機	1
一、2.文獻回顧	2
一、3.研究方法	6
二 基本理論分析	8
二、1. Jeffcott系統運動方程之建立	8
二、2.壓電方程理論模型之建立	9
二、3. 轉子系統方程式之無因次化模式	11
二、4.四階Runge-Kutta建立資料庫	12
三 實驗設計與理論驗證	14
三、1實驗裝置設計	14
三、2實驗電壓量測與理論驗證	15
四 深度學習	20
四、1.1機器學習簡介	20
四、1.2深度學習簡介	21
四、1.3資料庫之建立	22
四、2.1深度神經網路架構	23
四、2.2長短期記憶模型架構	26
四、2.3 XGBoost模型架構	29
五 深度學習訓練分析與結果	31
五、1.硬體對模型訓練時間之比較	31
五、2.DNN模型建立與分析	32
五、3. LSTM模型建立與分析	36
五、4. XGBoost模型建立與分析	40
五、5.模型比較分析	43
六 結論	47
參考文獻	49
論文簡要版	51
表目錄
表 1 數據所使用之參數	12
表 2 實驗電壓與理論電壓對照表	19
表 3 不同硬體模型訓練總時間表	31
表 4 DNN不同隱藏層決定係數數值	33
表 5 DNN不同神經元決定係數數值	34
表 6 LSTM不同隱藏層決定係數數值	37
表 7 LSTM 不同神經元決定係數數值	38
表 8 不同樹深度決定係數數值	41
表 9 演算法最佳結果數值比較	43

圖目錄
圖 1 Jeffcott轉子系統示意圖	8
圖 2 PZT與底座結合示意圖	10
圖 3 PZT震動示意圖	10
圖 4 軸承處的電壓數值分布圖	13
圖 5實驗裝置示意圖	14
圖 6 振動擷能系統裝置圖	15
圖 7 圓盤螺絲示意圖	16
圖 8 三顆螺絲各一螺帽之理論與實驗電壓比較圖	16
圖 9 圓盤螺絲示意圖	17
圖 10 三顆螺絲各二螺帽之理論與實驗電壓比較圖	17
圖 11 圓盤螺絲示意圖	18
圖 12 四顆螺絲各二螺帽之理論與實驗電壓比較圖	18
圖 13 機器學習分類	20
圖 14 深度學習流程圖	22
圖 15 人工神經元構造	24
圖 16 深度神經網路架構	25
圖 17 RNN 模型	27
圖 18 長短期記憶(LSTM)模型	27
圖 19長短期記憶法(LSTM)細胞架構	29
圖 20 DNN一層隱藏層層預測結果	32
圖 21 DNN不同隱藏層決定係數數值	33
圖 22 DNN不同神經元決定係數數值	34
圖 23 DNN最佳預測結果	35
圖 24 DNN最佳預測殘差	35
圖 25 LSTM一層隱藏層預測結果	36
圖 26 LSTM不同隱藏層決定係數數值	37
圖 27 LSTM不同神經元決定係數數值	38
圖 28 LSTM最佳預測結果	39
圖 29 LSTM最佳預測殘差	39
圖 30 XGBoost一層隱藏層層預測結果	40
圖 31XGBoost不同樹深度決定係數數值	41
圖 32 XGBoost最佳預測結果	42
圖 33 XGBoost最佳預測殘差	42
圖 34 最佳預測結果	44
圖 35 最佳預測殘差	45

[1]	H. H. Jeffcott, "XXVII. The lateral vibration of loaded shafts in the neighbourhood of a whirling speed.—The effect of want of balance," The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, vol. 37, no. 219, pp. 304-314, 1919.
[2]	N. A. Saeed et al., "Rub‐Impact Force Induces Periodic, Quasiperiodic, and Chaotic Motions of a Controlled Asymmetric Rotor System," Shock and Vibration, vol. 2021, no. 1, 2021, doi: 10.1155/2021/1800022.
[3]	Y. Ishida, "Cracked rotors: industrial machine case histories and nonlinear effects shown by simple Jeffcott rotor," Mechanical Systems and Signal Processing, vol. 22, no. 4, pp. 805-817, 2008.
[4]	L. Weng, Z. Yang, and Y. Cao, "The nonlinear dynamic characteristics of a cracked rotor-bearing system with initial bend deformation," in 2015 7th International Conference on Intelligent Human-Machine Systems and Cybernetics, 2015, vol. 2: IEEE, pp. 341-344. 
[5]	H. Hassenpflug, R. Flack, and E. Gunter, "Influence of acceleration on the critical speed of a Jeffcott rotor," 1981.
[6]	S. Piramoon, M. Ayoubi, and S. Bashash, "Modeling and Vibration Suppression of Rotating Machines Using the Sparse Identification of Nonlinear Dynamics and Terminal Sliding Mode Control," IEEE Access, 2024.
[7]	L. Zhang, L. Qin, Z. Qin, and F. Chu, "Energy harvesting from gravity-induced deformation of rotating shaft for long-term monitoring of rotating machinery," Smart Materials and Structures, vol. 31, no. 12, p. 125008, 2022.
[8]	S. Han, T. Yang, Q. Zhu, Y. Zhao, and Q. Han, "Unbalance position of aeroengine flexible rotor analysis and identification based on dynamic model and deep learning," Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, vol. 237, no. 19, pp. 4410-4429, 2023.
[9]	Y. LeCun, Y. Bengio, and G. Hinton, "Deep learning," nature, vol. 521, no. 7553, pp. 436-444, 2015.
[10]	C. Janiesch, P. Zschech, and K. Heinrich, "Machine learning and deep learning," Electronic Markets, vol. 31, no. 3, pp. 685-695, 2021.
[11]	J.-W. Baek and K. Chung, "Context deep neural network model for predicting depression risk using multiple regression," IEEE Access, vol. 8, pp. 18171-18181, 2020.
[12]	O. Avci, O. Abdeljaber, S. Kiranyaz, M. Hussein, M. Gabbouj, and D. J. Inman, "A review of vibration-based damage detection in civil structures: From traditional methods to Machine Learning and Deep Learning applications," Mechanical systems and signal processing, vol. 147, p. 107077, 2021.
[13]	S. Hochreiter, "Long Short-term Memory," Neural Computation MIT-Press, 1997.
[14]	S. Akhtar, M. Adeel, M. Iqbal, A. Namoun, A. Tufail, and K.-H. Kim, "Deep learning methods utilization in electric power systems," Energy Reports, vol. 10, pp. 2138-2151, 2023.
[15]	M. Sundermeyer, R. Schlüter, and H. Ney, "Lstm neural networks for language modeling," in Interspeech, 2012, vol. 2012, pp. 194-197. 
[16]	S. Karsoliya, "Approximating number of hidden layer neurons in multiple hidden layer BPNN architecture," International journal of engineering trends and technology, vol. 3, no. 6, pp. 714-717, 2012.
[17]	H. Jabbar and R. Z. Khan, "Methods to avoid over-fitting and under-fitting in supervised machine learning (comparative study)," Computer Science, Communication and Instrumentation Devices, vol. 70, no. 10.3850, pp. 978-981, 2015.
[18]	E. Buber and D. Banu, "Performance analysis and CPU vs GPU comparison for deep learning," in 2018 6th International Conference on Control Engineering & Information Technology (CEIT), 2018: IEEE, pp. 1-6. 
[19]	L. N. Smith, "A disciplined approach to neural network hyper-parameters: Part 1--learning rate, batch size, momentum, and weight decay," arXiv preprint arXiv:1803.09820, 2018.
[20]	D. R. Wilson and T. R. Martinez, "The need for small learning rates on large problems," in IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No. 01CH37222), 2001, vol. 1: IEEE, pp. 115-119. 
[21]	M. M. Lau and K. H. Lim, "Review of adaptive activation function in deep neural network," in 2018 IEEE-EMBS Conference on Biomedical Engineering and Sciences (IECBES), 2018: IEEE, pp. 686-690. 
[22]	N. Herisanu and V. Marinca, "Analytical Study of Nonlinear Vibration in a Rub-Impact Jeffcott Rotor," Energies, vol. 14, no. 24, 2021, doi: 10.3390/en14248298.
[23]	R. L. Harne and K. Wang, "A review of the recent research on vibration energy harvesting via bistable systems," Smart materials and structures, vol. 22, no. 2, p. 023001, 2013.