為確保石化工廠管線流體輸送時的安全，即時偵測輸運狀態並檢查是否有異常事件的發生非常重要。及早發現管線異常，可協助運轉員盡早判斷並且採取解救措施，以免導致更大的危害。本研究提出一套石化管線監控系統，運用同一管線系統於各地所量測之流量及壓力訊號，來進行事件工況的辨識與偵測其是否為未知事件(unseen event)。在「工況辨識」方面，由於目前所獲取的數據並無運轉工況的類別標註，是以我們提出一套預標註演算法，先從原始數據萃取特徵，再利用k平均演算法將這些特徵分群並標注類別。接著，以此標註結果作為基準(ground-truth)來訓練工況辨識模型。在「未知事件檢測」方面，我們利用自動編碼器(autoencoder)的壓縮還原能力並搭配四分位距法(the interquartile range method)所訂出閾值來區分當下事件是否為未知事件。實驗結果顯示，我們所提出的石化管線監控系統在「工況辨識」上，其準確率可達99.7%。「未知事件檢測」我們先以不同運轉工況作為未知事件來進行驗證，所獲之偽陰性率(false negative rate, FNR)與偽陽性率(false positive rate, FPR)在停泵狀態時分別為0%及13.8%；而在全運轉狀態下則為0%及2.2%。最後，我們更以含有管線操作異常之數據進行驗證，所獲之未知事件檢測之FNR及FPR在全運輸狀態下分別為6.8%及3.8%。以上成果可顯示我們所提之系統對於工況辨識及未知事件檢測助益。

To ensure the safety of the fluids transporting through pipelines in the petrochemical plant, it is essential to constantly identify the working states and detect whether an abnormal event occurs. To achieve it, we proposed a petrochemical pipeline monitoring system that utilizes the flow rate and pressure signals measured at various locations for working state identification and unseen event isolation in this study. For working state identification, we first developed a pre-labeling method for automated working state labeling. Next, its results were used as the ground truth to train the state identification model. As for unseen event isolation, it was achieved by the auto-encoders (AE) of various working states and the interquartile range (IQR)-based outlier detection method. The experimental results demonstrated that the petrochemical pipeline monitoring system we proposed achieved a working state identification accuracy of 99.7%. While letting different working states take turns as unseen events for evaluating the unseen event isolators, the false negative rate (FNR) and the false positive rate (FPR) were 0% and 13.8% in the pump shut-in state, and 0% and 2.2% in the full operation state, respectively. We finally included a dataset containing abnormal events for evaluation, and the resulting FNR and FPR of the unseen event isolation were 6.8% and 3.8%, respectively, in the full operation state. All of these demonstrate the efficacy of the proposed petrochemical pipeline monitoring system.

[1] 行政院環境保護署毒物及化學物質局。電子報案例事故專欄。民101年8月15日，取自：https://www.tcsb.gov.tw/lp-138-1.html
[2] H. J. Liaw, "Lessons in process safety management learned in the Kaohsiung gas explosion accident in Taiwan," Process Safety Progress, vol. 35, no. 3, pp. 228-232, 2016.
[3] C.-H. Chen, Y.-N. Sheen, and H.-Y. Wang, "Case analysis of catastrophic underground pipeline gas explosion in Taiwan," Engineering Failure Analysis, vol. 65, pp. 39-47, 2016.
[4] I. Nimmo, "Adequately address abnormal operations," Chemical engineering progress, vol. 91, no. 9, 1995.
[5] P.-S. Murvay and I. Silea, "A survey on gas leak detection and localization techniques," Journal of Loss Prevention in the Process Industries, vol. 25, no. 6, pp. 966-973, 2012.
[6] H. Lu, T. Iseley, S. Behbahani, and L. Fu, "Leakage detection techniques for oil and gas pipelines: State-of-the-art," Tunnelling and Underground Space Technology, vol. 98, p. 103249, 2020.
[7] K. J. Garner et al., "Duty cycle of the detector dog: A baseline study," Institute for Biological Detection Systems, Auburn University, 2001.
[8] J. C. Liou, "Leak detection by mass balance effective for Norman wells line," Oil and gas journal, vol. 94, no. 17, 1996.
[9] J. Rougier, "Probabilistic leak detection in pipelines using the mass imbalance approach," Journal of Hydraulic Research, vol. 43, no. 5, pp. 556-566, 2005.
[10] L. Billmann and R. Isermann, "Leak detection methods for pipelines," Automatica, vol. 23, no. 3, pp. 381-385, 1987.
[11] C. Verde, "Multi-leak detection and isolation in fluid pipelines," Control Engineering Practice, vol. 9, no. 6, pp. 673-682, 2001.
[12] C. Verde and N. Visairo, "Bank of nonlinear observers for the detection of multiple leaks in a pipeline," in Proceedings of the 2001 IEEE International Conference on Control Applications (CCA'01)(Cat. No. 01CH37204), 2001: IEEE, pp. 714-719.
[13] O. M. Aamo, J. Salvesen, and B. A. Foss, "Observer design using boundary injections for pipeline monitoring and leak detection," IFAC Proceedings Volumes, vol. 39, no. 2, pp. 53-58, 2006.
[14] E. Hauge, O. M. Aamo, and J.-M. Godhavn, "Model based pipeline monitoring with leak detection," IFAC Proceedings Volumes, vol. 40, no. 12, pp. 318-323, 2007.
[15] R. A. Silva, C. M. Buiatti, S. L. Cruz, and J. A. Pereira, "Pressure wave behaviour and leak detection in pipelines," Computers & chemical engineering, vol. 20, pp. S491-S496, 1996.
[16] H. Chen, H. Ye, L. Chen, and H. Su, "Application of support vector machine learning to leak detection and location in pipelines," in Proceedings of the 21st IEEE Instrumentation and Measurement Technology Conference (IEEE cat. no. 04ch37510), 2004, vol. 3: IEEE, pp. 2273-2277.
[17] A. S. de Joode and A. Hoffman, "Pipeline leak detection and theft detection using rarefaction waves," in 6th pipeline technology conference, 2011.
[18] W. Mpesha, Leak detection in pipes by frequency response method. University of South Carolina, 1999.
[19] T. R. Sheltami, A. Bala, and E. M. Shakshuki, "Wireless sensor networks for leak detection in pipelines: a survey," Journal of Ambient Intelligence and Humanized Computing, vol. 7, no. 3, pp. 347-356, 2016.
[20] Z. Qu, H. Feng, Z. Zeng, J. Zhuge, and S. Jin, "A SVM-based pipeline leakage detection and pre-warning system," Measurement, vol. 43, no. 4, pp. 513-519, 2010.
[21] R. R. Sharma, "Gas leakage detection in pipeline by svm classifier with automatic eddy current based defect recognition method," Journal of Ubiquitous Computing and Communication Technologies (UCCT), vol. 3, no. 03, pp. 196-212, 2021.
[22] R. Xiao, Q. Hu, and J. Li, "Leak detection of gas pipelines using acoustic signals based on wavelet transform and Support Vector Machine," Measurement, vol. 146, pp. 479-489, 2019.
[23] J. Wan, Y. Yu, Y. Wu, R. Feng, and N. Yu, "Hierarchical leak detection and localization method in natural gas pipeline monitoring sensor networks," Sensors, vol. 12, no. 1, pp. 189-214, 2011.
[24] S. Rashid, U. Akram, S. Qaisar, S. A. Khan, and E. Felemban, "Wireless sensor network for distributed event detection based on machine learning," in 2014 IEEE International Conference on Internet of Things (iThings), and IEEE Green Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom), 2014: IEEE, pp. 540-545.
[25] I. Brodetsky and M. Savic, "Leak monitoring system for gas pipelines," in 1993 IEEE International Conference on Acoustics, Speech, and Signal Processing, 1993, vol. 3: IEEE, pp. 17-20.
[26] T. B. Quy and J.-M. Kim, "Real-Time Leak Detection for a Gas Pipeline Using a k-NN Classifier and Hybrid AE Features," Sensors, vol. 21, no. 2, p. 367, 2021.
[27] G.-O. Glentis, K. Georgoulaki, and K. Angelopoulos, "Leak detection in the pipeline network of an oil refinery-A pattern classification paradigm," in 24th Pan-Hellenic Conference on Informatics, 2020, pp. 178-182.
[28] X.-f. Wang, Y. Wang, C.-l. Jiang, and H.-w. Liang, "Natural gas pipeline leak detection based on data mining," in 2011 International Conference on Consumer Electronics, Communications and Networks (CECNet), 2011: IEEE, pp. 492-494.
[29] S. Valizadeh, B. Moshiri, and K. Salahshoor, "Leak detection in transportation pipelines using feature extraction and KNN classification," in Pipelines 2009: Infrastructure's Hidden Assets, 2009, pp. 580-589.
[30] M. Munir, S. A. Siddiqui, A. Dengel, and S. Ahmed, "DeepAnT: A deep learning approach for unsupervised anomaly detection in time series," Ieee Access, vol. 7, pp. 1991-2005, 2018.
[31] Y. Song and S. Li, "Gas leak detection in galvanised steel pipe with internal flow noise using convolutional neural network," Process Safety and Environmental Protection, vol. 146, pp. 736-744, 2021.
[32] Y. Xie, Y. Xiao, X. Liu, G. Liu, W. Jiang, and J. Qin, "Time-frequency distribution map-based convolutional neural network (CNN) model for underwater pipeline leakage detection using acoustic signals," Sensors, vol. 20, no. 18, p. 5040, 2020.
[33] M. Zhou, Z. Pan, Y. Liu, Q. Zhang, Y. Cai, and H. Pan, "Leak Detection and Location Based on ISLMD and CNN in a Pipeline," IEEE Access, vol. 7, pp. 30457-30464, 2019.
[34] R. O. Melo, M. G. F. Costa, and C. F. Costa Filho, "Applying convolutional neural networks to detect natural gas leaks in wellhead images," IEEE Access, vol. 8, pp. 191775-191784, 2020.
[35] J. Li, Y. Liu, Y. Chai, H. He, and M. Gao, "A Small Leakage Detection Approach for Gas Pipelines based on CNN," in 2019 CAA Symposium on Fault Detection, Supervision and Safety for Technical Processes (SAFEPROCESS), 2019: IEEE, pp. 390-394.
[36] T.-K. Wang, Y.-H. Lin, and J.-Y. Shen, "Developing and Implementing an AI-Based Leak Detection System in a Long-Distance Gas Pipeline."
[37] L. Yang and Q. Zhao, "A novel PPA method for fluid pipeline leak detection based on OPELM and bidirectional LSTM," IEEE Access, vol. 8, pp. 107185-107199, 2020.
[38] H. Kim, M. Park, C. W. Kim, and D. Shin, "Source localization for hazardous material release in an outdoor chemical plant via a combination of LSTM-RNN and CFD simulation," Computers & Chemical Engineering, vol. 125, pp. 476-489, 2019.
[39] A. Kopbayev, F. Khan, M. Yang, and S. Z. Halim, "Gas leakage detection using spatial and temporal neural network model," Process Safety and Environmental Protection, vol. 160, pp. 968-975, 2022.
[40] B. Wang, Y. Guo, D. Wang, Y. Zhang, R. He, and J. Chen, "Prediction model of natural gas pipeline crack evolution based on optimized DCNN-LSTM," Mechanical Systems and Signal Processing, vol. 181, p. 109557, 2022.
[41] Z. Zuo, L. Ma, S. Liang, J. Liang, H. Zhang, and T. Liu, "A semi-supervised leakage detection method driven by multivariate time series for natural gas gathering pipeline," Process Safety and Environmental Protection, vol. 164, pp. 468-478, 2022.
[42] J. Na, K. Jeon, and W. B. Lee, "Toxic gas release modeling for real-time analysis using variational autoencoder with convolutional neural networks," Chemical Engineering Science, vol. 181, pp. 68-78, 2018.
[43] S. Ahmad, Z. Ahmad, C.-H. Kim, and J.-M. Kim, "A Method for Pipeline Leak Detection Based on Acoustic Imaging and Deep Learning," Sensors, vol. 22, no. 4, p. 1562, 2022.
[44] C. Spandonidis, P. Theodoropoulos, F. Giannopoulos, N. Galiatsatos, and A. Petsa, "Evaluation of deep learning approaches for oil & gas pipeline leak detection using wireless sensor networks," Engineering Applications of Artificial Intelligence, vol. 113, p. 104890, 2022.
[45] M. Dix et al., "Anomaly detection in the time-series data of industrial plants using neural network architectures," in 2021 IEEE Seventh International Conference on Big Data Computing Service and Applications (BigDataService), 2021: IEEE, pp. 222-228.
[46] S. Hansun, "A new approach of moving average method in time series analysis," in 2013 conference on new media studies (CoNMedia), 2013: IEEE, pp. 1-4.
[47] T. Salimans and D. P. Kingma, "Weight normalization: A simple reparameterization to accelerate training of deep neural networks," Advances in neural information processing systems, vol. 29, 2016.
[48] T. G. Dietterich, "Machine learning for sequential data: A review," in Joint IAPR international workshops on statistical techniques in pattern recognition (SPR) and structural and syntactic pattern recognition (SSPR), 2002: Springer, pp. 15-30.
[49] Q. Meng, D. Catchpoole, D. Skillicom, and P. J. Kennedy, "Relational autoencoder for feature extraction," in 2017 International Joint Conference on Neural Networks (IJCNN), 2017: IEEE, pp. 364-371.
[50] Z. Chen, C. K. Yeo, B. S. Lee, and C. T. Lau, "Autoencoder-based network anomaly detection," in 2018 Wireless telecommunications symposium (WTS), 2018: IEEE, pp. 1-5.
[51] D. J. MacKay and D. J. Mac Kay, Information theory, inference and learning algorithms. Cambridge university press, 2003.