應用於石化業關鍵設備之集成式智慧預知維護系統

The Smart Predictive Maintenance System Based on Ensemble Methods for Key Equipment in Petrochemical Industry

郭至恩(Chih-En Kuo)；沈育霖(Yu-Lin Shen)；曹常成(Charng-Cheng Tsaur)；張純明(Chwen-Ming Chang)；高振山(Chen-Shan Kao)；許世希(Shih-Hsi Hsu)；邱俊憲(Chun-Hsien Chiu)；陳振和(Thomas Jeng-Ho Chen)；蔡瑜潔(Amanda Lih-Chuan Chen)；梁勝富(Sheng-Fu Liang)

壓縮機 ； 物聯網 ； 大數據 ； 機器學習 ； 集成式學習 ； Compressor ； Internet of things ； Big data ； Machine learning ； Ensemble learning methods

勞動及職業安全衛生研究季刊

26卷3期（2018 / 09 / 15）

141 - 150

繁體中文

壓縮機是石化製程中最關鍵的設備，在工業4.0的推動下，許多產業界機具的智慧狀態診斷與預知維護概念紛紛被提出中，預知維護（Predictive Maintenance，簡稱PdM）是以狀態為依據(Condition Based)來預測設備狀態未來的運作趨勢，預知維護不僅能優化設備的運作時間和性能，並減少預防性維護的時間和人力成本，提升企業在製程安全管理的能力。本研究的目標是提出一套應用於石化業關鍵設備壓縮機之智慧預知維護方法，以預知跳機時間。我們在之前的研究中已採用了五種不同的數值分類器，運用於氣體壓縮機上振動、壓力、溫度等三種不同類型的大數據感測資料上，以評估分類器的適用性與感測資料類型的重要性順序。本研究則由國內的石化業大廠取得4筆由開機運行至意外停機的空氣壓縮機感測資料，其總時間長度約185天的連續感測資料來開發本系統的預測維護模組。為了有效的達到預測成效，此階段訓練資料是利用集成學習(ensemble learning)的方法，通過構建並結合多個學習器來完成致慧預測維護系統。該系統包含不同類型的個體學習器，集成異質的(heterogeneous)學習算法，通過將多個學習器進行組合，獲得比單一學習器顯著優越的泛化性能與評估能力。最後選定ensemble learning bagged trees為準確度高且訓練速度快的分類器模型，能夠得到泛化性能強的效果，將集成中的個體學習器儘可能相互獨立，使各學習器儘可能具有較大的差異。該模型的預測準確率高達99.9%。此外，為了驗證此一開發策略的推廣可行性，我們用另一部類型相同但汽缸數不同的壓縮機實施相同的感測資料分析與模型建構策略，其總時間長度約93天的連續感測資料來開發該壓縮機的預測維護模組。模型最後的準確率亦高達99.99%，顯示此一開發策略能推廣至其他石化廠的壓縮機（或其他設備），是極佳的推廣策略。未來研究方向可應用於其他類似的設備，針對各類別的感測訊號做分析，希望不只能預測需要維護的時間，也能夠偵測可能故障的部位，節省停機維修或保養的時間，得到更通用的智慧預知維護方法。

In the petrochemical industry, an unplanned stop causes extremely high costs. It results in an unscheduled downtime with no possibility to continue production, unplanned maintenance costs are a lot higher than planned maintenance. A shutdown would cause production downtime, and therefore, forces the companies to buy the products from competitors, causing major costs. Through condition monitoring analyses, the lifespan and future maintenance of machines are determined. This ensures that the necessary small repairs do not grow into major repairs, which requires a prolonged production stoppage. In this study, we developed an artificial intelligence predictive maintenance methods. In previous study, we finished an advance development and evaluation of an intelligent predictive maintenance method for key equipment in petrochemical industry. Five different numerical classifier were used in compressors on the vibration, pressure, temperature types of sensing data in order to assess the suitability of the classifier and the importance of the type of data. In this study, we use the data from petrochemical plant compressor which length of approximate 185 days to different numerical classifier were used in compressors on the vibration, pressure, temperature types of sensing data in order to assess the suitability of the classifier and the importance of the type of data. For the further achievement, through the ensemble learning classifier to help us to evaluate the equipment performance. Finally, the result of trained model, bagged trees classifier (one of ensemble learning classifiers) who present the highest accuracy and learn faster than most of other classifiers. Furthermore, in order to prove our methodology is easy to implement on other similar equipment. Our group get the data from petrochemical plant compressor (same type, but it with six cylinders) which length of approximate 93 days. Using of the same methodology, the accuracy can also achieve 99.99%. It means that the proposed modeling strategy is very suitable for compressor. In future, in order to develop a generalized predictive maintenance system with high accuracy, more short-term data or features should be considered and tested. The results of this study should be able to expand to other similar equipment. Try to find out the failure mode and the failure causes, to accomplish real artificial intelligence predictive maintenance.

1. 郭至恩、張純明、高振山、許世希、陳振和、蔡瑜潔、梁勝富(2018)。先期開發與評估應用於石化業關鍵設備之智慧預知維護方法。勞動及職業安全衛生研究季刊，26(1)，1-80。
   連結：
2. El-thalji, I,Jantunen, E(2015).A summary of fault modelling and predictive health monitoring of rolling element bearings.Mechanical Systems and Signal Processing,60,252-72.
3. Grimm, L. G.(Ed.),Yarnold, P. R.(Ed.)(1995).Reading and understanding multivariate statistics.American Psychological Association.
4. Hegde, GP,Seetha, M,Hegde, N(2016).Kernel Locality Preserving Symmetrical Weighted Fisher Discriminant Analysis based subspace approach for expression recognition.Engineering Science and Technology, an International Journal,19,1321-33.
5. Mobley, RK(2002).An introduction to predictive maintenance.Butterworth-Heinemann.
6. Mounce, SR,Ellis, K,Edwards, JM,Speight, VL,Jakomis, N,Boxall, JB(2017).Ensemble Decision Tree Models Using RUSBoost for Estimating Risk of Iron Failure in Drinking Water Distribution Systems.Water Resources Management,31,1575-90.
7. Northerton, D(2000).RCM Standard.Maintenance & Asset Management,15,12-20.
8. Ponomareva, N,Colthurst, T,Hendry, G,Haykal, S,Radpour, S(2017).Compact multi-class boosted trees. 2017 IEEE International Conference on Big Data (Big Data) Big Data (Big Data).2017 IEEE International Conference on.
9. Sherwin, D(2000).A Review of Overall Models for Maintenance Management.Journal of Quality in Maintenance Engineering,6,138-64.
10. Soualhi, A,Medjaher, K,Zerhouni, N(2015).Bearing health monitoring based on Hilbert-Huang transform, support vector machine, and regression.IEEE Transactions on Instrumentation and Measurement,64,52-62.
11. Sun, J,Yan, C,Wen, J(2018).Intelligent bearing fault diagnosis method combining compressed data acquisition and deep learning.IEEE Transactions on Instrumentation and Measurement,67,185-95.
12. TOMAK, O,KAYIKÇIOLU, T(2018).Bagged tree classification of arrhythmia using wavelets for denoising, compression, and feature extraction.Turkish Journal of Electrical Engineering & Computer Sciences,26,1555-71.
13. Xu, X,et, al(2017).A belief rule-based expert system for fault diagnosis of marine diesel engines.IEEE Transactions on Systems, Man, and Cybernetics: Systems
14. Zhang, B,Qin, KA,Sellis, T(2018).Evolutionary feature subspaces generation for ensemble classification.Genetic & Evolutionary Computation Conference,Kyoto, Japan:
15. Zhang, Q(2015).Vibration sensor based intelligent fault diagnosis system for large machine unit in petrochemical industries.International Journal of Distributed Sensor Networks,11,239405.

1. 趙仁德,陳彥銘,呂政芳(2022)。利用人工智慧建立石油煉製業壓縮機之異常診斷機制。石油季刊，58(1)，83-103。