臺灣博碩士論文加值系統

故障預測與健康管理（Prognostic and Health Management，PHM）是比傳統的維護技術更加卓越的方式，具有故障診斷、預測和健康管理技術的優點，此技術利用多感測技術監控，藉助各種演算法和模型來預測與評估系統的健康狀態，經由數據監控與分析，預測故障的發生，因此它是一種整合故障檢測、健康預測與評估及維護決策的綜合技術。
因應工業4.0及科技的發展，將故障預測與健康管理運用於工具機上已成為各產業的重要發展趨勢，在理論上此維護的方式突破了傳統維護方法，透過故障預測與健康管理能夠有效提高工具機的可靠性，減少不必要的成本浪費，且工具機運用故障預測與健康管理的技術，是具有顯著的經濟效益。
本文針對高單價工具機提出了工具機故障預測與健康管理模式建構，工具機主要的異常來源為主軸、移動平台以及周邊設備，因此本研究針對主軸轉速、主軸溫度以及XYZ滑台誤差位置來進行探討。研究方法分成四個步驟來進行：第一步蒐集工具機於運行過程中的相關資料；第二步將針對所蒐集到的工具機運行資料進行資料預處理與資料融合，以利下一步驟使用；第三步使用時間序列預測模型的方式對工具機進行預測未來趨勢，關於時間序列預測模型本文採用九種預測的方式對其進行預測，而後使用了五種模型評估方式來對預測模型進行評估，並選擇出最佳的預測模型來對工具機進行預測趨勢，經由分析評估後，顯示Prophet預言家模型為最適合之預測模型，因此利用Prophet預言家模型對工具機進行預測未來趨勢，Prophet預言家模型預測工具機資料的準確度達到88％以上；第四步驟依據預測結果對工具機進行預測診斷以及健康評估。

Prognostic and Health Management (PHM) technology is not only a troubleshooting, predictive and health management technology, but also that is superior to the traditional maintenance services. This technology uses multi-sensor technology to monitor and uses algorithms and models to predict and evaluate the health of the device. Through data monitoring and analysis, it can predict the occurrence of failures. Therefore, it is an integrated technology for integrated fault detection, health prediction and assessment, and maintenance decisions. Today, the Prognostic and Health Management applied to machine tool has become an important trend. Using Prognostic and Health Management can improve the machine tool reliability, availability, safety, and reduced maintenance costs.
With the rapid advancement of Industry 4.0 and technology, the application of PHM on machine tools has become an important development trend in various industries. In theory, this maintenance method breaks through the traditional maintenance method and can be effectively improved. Using the PHM on the machine tools can reduce costs, and the use of PHM on machine tools has significant economic benefits.
In this research we propose The Model of a Machine Tool Prognostic and Health Management for high-priced machine tool. The main sources of abnormalities for machine tool are spindles, lathe bench and other equipment. Therefore, this research discusses the spindle speed, spindle temperature and XYZ table error position. It can be divided into four steps. The first step is to collect machine tool information. The second step is data preprocessing and data fusion for use in the next step. The third step uses the time series prediction model to predict the future trend of the machine tool. About the time series prediction model, we use nine prediction methods to predict it, and then use five evaluation methods is used to evaluate the prediction model, and the best prediction model is selected to predict the trend of the machine tool. After analysis and evaluation, the Prophet model is displayed as the most suitable prediction model, so we use the Prophet model to predicts the future trend of the machine tool. The Prophet model predicts the accuracy of the machine tool data to be more than 88%. The last step is to evaluate the health of the machine tool according to the prediction results.

摘要.......................................................i
Abstract...................................................ii
誌謝.......................................................iv
目錄.......................................................v
表目錄.....................................................viii
圖目錄.....................................................ix
第一章 緒論................................................1
1.1 研究背景與動機........................................1
1.2 研究方法..............................................2
1.3 研究流程..............................................3
1.4 論文章節..............................................4
第二章 文獻探討............................................5
2.1 工業4.0...............................................5
2.2 設備維護方法..........................................5
2.2.1 緊急搶修策略........................................8
2.2.2 預防維護策略........................................10
2.2.3 預兆式維護策略......................................11
2.3 故障預測與健康管理....................................13
2.3.1 故障預測與健康管理模式..............................14
2.3.2 資料融合............................................17
2.3.3 預測模型............................................18
2.4 本章小結..............................................22
第三章 工具機故障預測與健康管理模式建構....................23
3.1 資料預處理（Data Preprocessing）......................24
3.1.1 Z分數標準化(Z-Score Standardization)................24
3.2 資料融合..............................................25
3.3 時間序列預測..........................................27
3.3.1 天真預測模型........................................27
3.3.2 指數平滑模型........................................27
3.3.3 Holt模型............................................28
3.3.4 Holt-Winter模型.....................................29
3.3.5 自回歸模型..........................................31
3.3.6 移動平均模型........................................31
3.3.7 自回歸移動平均模型..................................32
3.3.8 差分整合移動平均自回歸模型..........................33
3.3.9 Prophet預言家模型...................................34
第四章 工具機故障預測與健康管理模式建構實作................38
4.1 Python................................................39
4.2 模擬工具機運行........................................40
4.3 資料預處理............................................43
4.4 資料融合..............................................48
4.5 時間序列預測模型......................................49
4.5.1 天真預測模型........................................51
4.5.2 指數平滑模型........................................52
4.5.3 Holt模型............................................54
4.5.4 Holt-Winter模型.....................................56
4.5.5 自回歸模型..........................................58
4.5.6 移動平均模型........................................60
4.5.7 自回歸移動平均模型..................................61
4.5.8 差分整合移動平均自回歸模型..........................63
4.5.9 Prophet預言家模型...................................65
4.6 模型評估..............................................67
第五章 模型驗證與討論......................................69
5.1 驗證一................................................71
5.2 驗證二................................................73
5.2.1 以normal模擬方式產生資料............................73
5.2.2 以logistic模擬方式產生資料..........................76
第六章 結論................................................79
參考文獻...................................................80
Extended Abstract..........................................82

[1]南樺 電子報 148，2009
[2]DALE R. BLANN, Maximizing the P-F Interval Through Condition-Based
Maintenance, Marshall Institute Inc., 2013
[3]Andy Hess, The Joint Strike Fighter (JSF) Prognostics and Health Management, NDIA 4th Annual Systems Engineering Conference, 2001
[4]Gill, JJ; Lessons learned from rotary- and fixed-wing HUMS applications,Aerospace Conference Proceedings, 2000
[5]Byington,CS; Watson,M.; Roemer,MJ; Galic, TR;McGroarty, JJ; Prognostic enhancements to gas turbine diagnostic systems，Aerospace Conference, 2003
[6]Byington,CS; Roemer,MJ; Galie,T.; Prognostic enhancements to diagnostic systems for improved condition-based maintenance， Aerospace Conference Proceedings, 2002
[7]Araiza,ML; Kent,R.; Espinosa,R.; Real-time, embedded diagnostics and prognostics in advanced artillery systems, AUTOTESTCON Proceedings,2002
[8]F. Castanedo, "A review of data fusion techniques", The Scientific World Journal, 2013
[9]楊森，基於視情況維修的電子裝備PHM體系統結構分析，北京航空航天大學自動化科學與電氣工程
學院，中國，2017
[10]Kacprzynski G J, Roemer M J. Health management strategies for 21st century condition-based maintenance systems J. In- ternational COMADEM Congress,Houston, 2000
[11]Haghighat, Mohammad; Abdel-Mottaleb, Mohamed; Alhalabi, Wadee"Discriminant Correlation Analysis: Real-Time Feature Level Fusion for Multimodal Biometric Recognition", 2016
[12]Talyor SJ, Letham B. "Forecasting at Scale", Facebook, Menlo Park,California, United States, 2017
[13]YaakovBar-Shalom, EdisonTse“Tracking in a cluttered environment with probabilistic data association”, Systems Control, Inc., Palo Alto,California 94304, United States,1975
[14]薛吉順，使用智慧型計算於不確定動態環境之多感測器資料融合系統，元智大學電機工程學系，
台灣，2014
[15]許展源，應用雙季節指數平滑模型於來電人數預測之研究，國立政治大學統計系，台灣，2018
[16]江政儒，應用ARIMA建立部門航空油料需求預測模型，國防大學運籌管理學系，台灣，2019
[17]Hugh F. Durrant-Whyte, Sensor Models and Multisensor Integration,Department of Engineering Science University of Oxford, U.K,1988
[18]Hall D L, Llinas J. An introduction to multisensor data fusion.Proceedings of the IEEE, 1997
[19]Kalgren, P. W., Byington, C. S., Roemer, M. J., & Watson, M. J. (2007).Defining PHM, a lexical evolution of maintenance and logistics. In 2006 IEEE AUTOTESTCON - IEEE Systems Readiness Technology Conference