為確保核電廠的安全運轉，在事故發生時，我們需能快速診斷出造成異常情事的肇始事件(initiating event, IE)，協助運轉員採取對應的舒緩措施；在穩定運轉中，能檢測感測器是否因老化而發生讀值飄移則是一個要點。本研究提出的核電廠在線智慧監控系統架構，主要由兩個子系統所組成：「偵測系統」及「診斷系統」，專注於肇始事件與感測器失能的偵測與診斷。
　　偵測系統旨在自動偵測未知異常事件，並鑑別該事件屬於肇始事件或失能感測器。偵測機制的實現分別透過「機器學習(machine learning)─Hotelling’s T2檢定」及「深度學習(deep learning)─長短期記憶」來完成。診斷系統內含「電廠運轉診斷模組」及「感測器診斷模組」，電廠運轉診斷模組的目標在辨識肇始事件，模組之實現也分別透過機器學習及深度學習方法來完成，肇始事件若屬於破管事件，則進一步評估該事件的嚴重程度；若電廠發生未建於訓練資料庫中的未見事件(unseen event)，也有能力隔離以免造成誤判。在機器學習方法中，用於事件辨識的特徵萃取器，使用感測器種類區塊投影法(sensor type-wise block projection, stBP)來擷取數據中的空時(spatio-temporal)資訊。基因演算法(genetic algorithm, GA)、循序向前選擇法(sequential forward selection, SFS)及可刪式向前選擇法(deflatable sequential forward selection, dSFS)則用於實現感測器偵選，刪除掉對辨識沒幫助的量測訊號。最終，未見事件隔離的議題，則透過四分位距(interquartile range, IQR)搭配機率計算的方法來完成。而在深度學習方面，我們憑藉著長短期記憶(long short-term memory, LSTM)能善於學習時間序列資料的動態變化的特性，來達成肇始事件偵測的需求；由於連續擷取數個時間點的多感測器數據，可視為一廣義的影像，故卷積神經網路(convolutional neural network, CNN)及全卷積網路(fully convolutional network, FCN)可被利用來解決肇始事件辨識及破管事件嚴重性評估的問題；在未見事件隔離的部分，我們則是用自動編碼器(autoencoder, AE)對於學習或未學習過的數據，其對應還原效果會有所不同的特性，讓未見事件可被有效隔離。
　　感測器診斷模組的部分，因感測器異常偵測中，仍無法確實得知是哪支感測器發生失能現象，為確切得知感測器為何，我們提出了類似循序向後選擇法的作法，迅速定位出失能的感測器。另外，由於失能感測器的訊號正確性已喪失，是以重建失能感測器的讀值讓運轉員參考使用是必要的。故而提出基於多變量自回歸模型(multivariate autoregressive model, MVAR)，於電廠穩定運轉抑或肇始事件下都能運用其他正常感測器的資訊，重建出失能感測器的量測讀值。另外，也採用卷積神經網路，協助解決失能感測器存在下的肇始事件辨識問題，並取代多變量自回歸模型，再一次重建肇始事件下失能感測器的訊號變化。
　　本研究以台灣核三廠的電廠模擬程式MAAP5 (Modular Accident Analysis Program 5)及PCTran (Personal Computer Transient Analyzer)生成電廠於穩定狀態及肇始事件情況下的數據，進行系統效能驗證。偵測系統的驗證結果顯示，當電廠有未知異常事件發生，我們的系統皆可迅速地偵測到，並成功鑑別為肇始事件或失能感測器。關於電廠運轉診斷模組的驗證結果，經辨識測試，依機器學習方法建立的辨識策略中，特徵萃取器採用感測器種類區塊投影法，且只挑選七支感測器的量測訊號進行辨識，擁有93.18%的辨識成功率。依深度學習方法建立的辨識策略，則有96.52%的辨識成功率；肇始事件的嚴重性評估，三大類破管事件(冷卻水流失事件、主蒸汽管斷裂事件及蒸汽產生器破管事件)的平均相對誤差皆在5%以內；長短期記憶為基礎的自動編碼器應用於未見事件隔離擁有不錯的表現，F1-score為0.9593。關於感測器診斷模組的驗證結果，當若干感測器發生失能時，它們也都能有效地被演算法定位。該感測器的訊號在電廠穩定狀態下可成功重建，且不論何種失能模式發生，皆能在啟動訊號重建後7個時間點之內將訊號修正回來；若在肇始事件下，考慮各感測器在各肇始事件下進行的重建，共145種測試情境組合，其中136種組合的相對誤差皆在0.1以下。

To maintain the safe operation of nuclear power plants (NPPs), fast detection and diagnosis of the initiating event (IE) that causes the abnormal condition is crucial. It is also essential that detecting whether the sensor has a drift or bias due to aging when NPP is under normal operation. A smart on-line monitoring system for NPPs is proposed to fulfill these needs.
The system comprises two subsystems: the “detection system” and the “diagnostic system.” The detection system automatically detects abnormal events and discriminates this abnormality as IEs or faulty sensors. The implementation of the detection mechanism is carried out by machine learning and deep learning-based approach, which are Hotelling’s T2 test and long short-term memory (LSTM).
The diagnostic system includes the “plant operation diagnostic module” and the “sensor diagnostic module.” The module implementation is also done through machine learning and deep learning-based approaches. The goal of the plant operation diagnostic module is to identify the IE. If the IE belongs to the break event, the size of this break shall be further evaluated. Once an unseen event happens, this system also can avoid relating this event to those in the event database. To begin, the sensor type-wised block projection (stBP) is used to elicit the discriminant information in the data obtained from various NPP sensors to facilitate event identification. Moreover, the extracted features are subjected to a further dimensionality reduction by sensor selection. Several optimization algorithms, such as genetic algorithm (GA), sequential forward selection (SFS), and deflatable sequential forward selection (dSFS), are implemented for selecting sensors. Finally, for the unseen event isolation, the interquartile range (IQR) incorporates with probability calculation is considered. As for the deep learning-based approach, because the outputs of multiple sensors at several time instants are stacked as a matrix, a convolutional neural network (CNN) and fully convoluted network (FCN) can thus solve the problems of IE identification and the break size assessment. Since the data of known event categories train the proposed autoencoder, a significant reconstruction error is expected when feeding in data from the unseen event. LSTM-based autoencoder (LSTM-AE) is thus feasible for unseen event isolation.
For a sensor diagnostic module, it is still required to check all the sensors to isolate the defective ones, because only the existence of an anomaly in a sensor group is detected through the detection system. To achieve this, a sequential backward selection (SBS)-based approach is presented. Moreover, once a sensor is determined to be faulty, its readings are unreliable for providing correct information. However, when the IE occurs, the operator still refers to specific plant variables to confirm whether mitigation action is successfully executed following the emergency operating procedure (EOP). Therefore, an approach for reconstructing the readings of the faulty sensor is essential. In this study, a multivariate autoregressive (MVAR) model-based approach is proposed to reconstruct the faulty sensor readings using the information from other healthy sensors. If the data required for event identification is contaminated with faulty sensor readings, a reduction in the accuracy of event identification may be led to. Moreover, to reconstruct the faulty sensor readings under IE with the MVAR model, deterministic trends in sensor signal must be removed. It also depends on the correct identification of the abnormal event so that we can correctly remove the deterministic trends. For addressing these issues, CNNs are adopted again for event identification and signal reconstruction with sensor faults.
Various IEs and normal conditions considered in this study are to evaluate the performance of a smart on-line monitoring system. The simulations were performed on Personal Computer Transient Analyzer (PCTran) and Modular Accident Analysis Program 5 (MAAP5), which can simulate various accident and transient conditions for Taiwan Maanshan NPP. The experimental results of the detection system indicated that our system could timely detect and discriminate all IEs and faulty sensors of interest. The experimental results about evaluating the plant operation diagnostic system indicate that the identification strategy established based on the machine learning had an identification rate of 93.18%, while using stBP as a feature extractor and only selects seven sensors for event identification. The identification strategy established by deep learning has an identification rate of 96.52%. For break size assessment, all of the mean relative errors for three types of break event (i.e., loss of coolant accident, main steam line break, and steam generator tube rupture) were lower than 5%. LSTM-AE also had good performance when applied to unseen event isolation. The F1-score was 0.9593. The experimental results about evaluating the sensor diagnostic system indicate that this system could effectively detect and isolate the faulty sensors. Whatever sensor failure modes occurred under normal conditions, it corrected the faulty sensor readings within 7-time instants after signal reconstruction. Considering the reconstruction tasks under different sensor types and event categories, a total of 145 sensor type-event type combinations. The number of sensor type-event type combinations with relative errors lower than 0.1 was 136.

[1] H. Kagermann, J. Helbig, A. Hellinger, and W. Wahlster, Recommendations for implementing the strategic initiative INDUSTRIE 4.0: Securing the future of German manufacturing industry; final report of the Industrie 4.0 Working Group: Forschungsunion, 2013.
[2] J. W. Hines, and E. Davis, “Lessons learned from the US nuclear power plant on-line monitoring programs,” Progress in Nuclear Energy, vol. 46, pp. 176-189, 2005.
[3] P. Rebstock, “On-Line Monitoring for Calibration Extension: an Overview and Introduction.”
[4] J. Ma, and J. Jiang, “Applications of fault detection and diagnosis methods in nuclear power plants: A review,” Progress in nuclear energy, vol. 53, pp. 255-266, 2011.
[5] G. Hessel, W. Schmitt, K. Van der Vorst, and F.-P. Weiss, “A neutral network approach for acoustic leak monitoring in the vver-440 pressure vessel head,” Progress in Nuclear Energy, vol. 34, pp. 173-183, 1999.
[6] C. J. Li, and S. Li, “Acoustic emission analysis for bearing condition monitoring,” Wear, vol. 185, pp. 67-74, 1995.
[7] J. Kim, I. Hwang, D. Lee, C. Ham, and T. Kim, “Development of a mass estimation algorithm using the impact test data of nuclear power plant,” Nuclear Engineering and Technology, vol. 32, pp. 227-234, 2000.
[8] G. Pokol, and G. Por, “The Advanced Loose Parts Monitoring System (ALPS) and wavelet analysis,” International Journal of Nuclear Energy Science Technology, vol. 2, pp. 241-252, 2006.
[9] R. M. G. Souza, and J. M. Moreira, “Neural network correlation for power peak factor estimation,” Annals of Nuclear Energy, vol. 33, pp. 594-608, 2006.
[10] T. Santosh, G. Vinod, R. Saraf, A. Ghosh, and H. Kushwaha, “Application of artificial neural networks to nuclear power plant transient diagnosis,” Reliability Engineering and System Safety, vol. 92, pp. 1468-1472, 2007.
[11] M. J. Embrechts, and S. Benedek, “Hybrid identification of nuclear power plant transients with artificial neural networks,” IEEE Transactions on Industrial Electronics, vol. 51, pp. 686-693, 2004.
[12] E. Zio, and P. Baraldi, “Evolutionary fuzzy clustering for the classification of transients in nuclear components,” progress in nuclear energy, vol. 46, pp. 282-296, 2005.
[13] K. C. Kwon, J. H. Kim, and P. H. Seong, “Hidden Markov model‐based real‐time transient identifications in nuclear power plants,” International journal of intelligent systems, vol. 17, pp. 791-811, 2002.
[14] J. A. C. C. Medeiros, and R. Schirru, “Identification of nuclear power plant transients using the Particle Swarm Optimization algorithm,” Annals of Nuclear Energy, vol. 35, pp. 576-582, 2008.
[15] K. Moshkbar-Bakhshayesh, and M. B. Ghofrani, “Transient identification in nuclear power plants: A review,” Progress in Nuclear Energy, vol. 67, pp. 23-32, 2013.
[16] M. G. Na, W. S. Park, and D. H. Lim, “Detection and diagnostics of loss of coolant accidents using support vector machines,” IEEE Transactions on Nuclear Science, vol. 55, pp. 628-636, 2008.
[17] M. G. Na, S. H. Shin, S. M. Lee, D. W. Jung, S. P. Kim, J. H. Jeong, and B. C. Lee, “Prediction of major transient scenarios for severe accidents of nuclear power plants,” IEEE transactions on nuclear science, vol. 51, pp. 313-321, 2004.
[18] K. Mo, S. J. Lee, and P. H. Seong, “A dynamic neural network aggregation model for transient diagnosis in nuclear power plants,” Progress in Nuclear Energy, vol. 49, pp. 262-272, 2007.
[19] M. Marseguerra, and A. Zoia, “The AutoAssociative Neural Network in signal analysis: II. Application to on-line monitoring of a simulated BWR component,” Annals of Nuclear Energy, vol. 32, pp. 1207-1223, 2005.
[20] D. Roverso, “Soft computing tools for transient classification,” Information Sciences, vol. 127, pp. 137-156, 2000.
[21] D. Roverso, “Plant diagnostics by transient classification: The aladdin approach,” International Journal of Intelligent Systems and Information Sciences, vol. 17, pp. 767-790, 2002.
[22] K.-C. Kwon, “HMM-based transient identification in dynamic process,” Trans Control Automat Syst Eng, vol. 2, pp. 40-46, 2000.
[23] K.-C. Kwon, and J.-H. Kim, “Accident identification in nuclear power plants using hidden Markov models,” Engineering Applications of Artificial Intelligence, vol. 12, pp. 491-501, 1999.
[24] C. Gottlieb, V. Arzhanov, W. Gudowski, and N. Garis, “Feasibility study on transient identification in nuclear power plants using support vector machines,” Nuclear technology, vol. 155, pp. 67-77, 2006.
[25] E. Zio, and P. Baraldi, “Identification of nuclear transients via optimized fuzzy clustering,” Annals of Nuclear Energy, vol. 32, pp. 1068-1080, 2005.
[26] C. Lin, and H.-J. Chang, “Identification of pressurized water reactor transient using template matching,” Annals of Nuclear Energy, vol. 38, pp. 1662-1666, 2011.
[27] Y.-G. No, J.-H. Kim, M.-G. Na, D.-H. Lim, and K.-I. Ahn, “Monitoring severe accidents using AI techniques,” Nuclear Engineering and Technology, vol. 44, pp. 393-404, 2012.
[28] W. J. Scheirer, A. de Rezende Rocha, A. Sapkota, and T. E. Boult, “Toward open set recognition,” IEEE transactions on pattern analysis and machine intelligence, vol. 35, pp. 1757-1772, 2012.
[29] S.-C. Wu, K.-Y. Chen, T.-H. Lin, and H.-P. Chou, “Multivariate algorithms for initiating event detection and identification in nuclear power plants,” Annals of Nuclear Energy, vol. 111, pp. 127-135, 2018.
[30] Y. Bartal, J. Lin, and R. E. Uhrig, “Nuclear power plant transient diagnostics using artificial neural networks that allow “don’t-know” classifications,” Nuclear Technology, vol. 110, pp. 436-449, 1995.
[31] X. Tian, V. Becerra, N. Bausch, T. Santhosh, and G. Vinod, “A study on the robustness of neural network models for predicting the break size in LOCA,” Progress in Nuclear Energy, vol. 109, pp. 12-28, 2018.
[32] M. Saghafi, and M. B. Ghofrani, “Real-time estimation of break sizes during LOCA in nuclear power plants using NARX neural network,” Nuclear Engineering and Technology, vol. 51, pp. 702-708, 2019.
[33] T. Santhosh, M. Kumar, I. Thangamani, A. Srivastava, A. Dutta, V. Verma, D. Mukhopadhyay, S. Ganju, B. Chatterjee, and V. Rao, “A diagnostic system for identifying accident conditions in a nuclear reactor,” Nuclear engineering and design, vol. 241, pp. 177-184, 2011.
[34] F. Jamil, M. Abid, I. Haq, A. Q. Khan, and M. Iqbal, “Fault diagnosis of Pakistan Research Reactor-2 with data-driven techniques,” Annals of Nuclear Energy, vol. 90, pp. 433-440, 2016.
[35] W. Li, M. Peng, Y. Liu, N. Jiang, H. Wang, and Z. Duan, “Fault detection, identification and reconstruction of sensors in nuclear power plant with optimized PCA method,” Annals of Nuclear Energy, vol. 113, pp. 105-117, 2018.
[36] A. Ray, and S. Phoha, “Calibration and estimation of redundant signals,” Automatica, vol. 36, pp. 1525-1534, 2000.
[37] R. Dorr, F. Kratz, J. Ragot, F. Loisy, and J.-L. Germain, “Detection, isolation, and identification of sensor faults in nuclear power plants,” IEEE Transactions on Control Systems Technology, vol. 5, pp. 42-60, 1997.
[38] G. Ablay, and T. Aldemir, “Fault detection in nuclear systems using sliding mode observers,” Nuclear Science and Technology, vol. 173, pp. 82-98, 2013.
[39] Y. Pang, and H. Xia, “Fault detection of nuclear reactors by estimation of unknown input for systems with disturbances,” Nuclear Engineering and Design, vol. 275, pp. 91-95, 2014.
[40] P. Baraldi, A. Cammi, F. Mangili, and E. Zio, “An ensemble approach to sensor fault detection and signal reconstruction for nuclear system control,” Annals of Nuclear Energy, vol. 37, pp. 778-790, 2010.
[41] R. Dunia, S. J. Qin, T. F. Edgar, and T. J. McAvoy, “Identification of faulty sensors using principal component analysis,” AIChE Journal, vol. 42, no. 10, pp. 2797-2812, 1996.
[42] N. Kaistha, and B. R. Upadhyaya], “Incipient fault detection and isolation of field devices in nuclear power systems using principal component analysis,” Nuclear Technology, vol. 136, pp. 221-230, 2001.
[43] B. Lu, and B. Upadhyaya, “Monitoring and fault diagnosis of the steam generator system of a nuclear power plant using data-driven modeling and residual space analysis,” Annals of Nuclear Energy, vol. 32, pp. 897-912, 2005.
[44] B. Upadhyaya, K. Zhao, and B. Lu, “Fault monitoring of nuclear power plant sensors and field devices,” Progress in nuclear energy, vol. 43, pp. 337-342, 2003.
[45] K. R. Uren, G. Van Schoor, P. Carel, and A. Botha, “An integrated approach to sensor FDI and signal reconstruction in HTGRs–Part I: Theoretical framework,” Annals of Nuclear Energy, vol. 87, pp. 750-760, 2016.
[46] K. R. Uren, G. Van Schoor, P. Carel, and A. Botha, “An integrated approach to sensor FDI and signal reconstruction in HTGRs–Part II: Case studies,” Annals of Nuclear Energy, vol. 87, pp. 739-749, 2016.
[47] N. Zavaljevski, and K. C. Gross, Sensor fault detection in nuclear power plants using multivariate state estimation technique and support vector machines, Argonne National Lab., Argonne, IL (US), 2000.
[48] P. F. Fantoni, “Experiences and applications of PEANO for online monitoring in power plants,” Progress in nuclear energy, vol. 46, pp. 206-225, 2005.
[49] P. F. Fantoni, and A. Mazzola, “Multiple-failure signal validation in nuclear power plants using artificial neural networks,” Nuclear technology, vol. 113, pp. 368-374, 1996.
[50] S. Martinez-Martinez, N. Messai, J.-P. Jeannot, and D. Nuzillard, “Two neural network based strategies for the detection of a total instantaneous blockage of a sodium-cooled fast reactor,” Reliability Engineering and System Safety, vol. 137, pp. 50-57, 2015.
[51] A. Messai, A. Mellit, I. Abdellani, and A. M. Pavan, “On-line fault detection of a fuel rod temperature measurement sensor in a nuclear reactor core using ANNs,” Progress in Nuclear Energy, vol. 79, pp. 8-21, 2015.
[52] U. N. R. Commission, Guidelines for the preparation of emergency operating procedures, 1982.
[53] 陳光佑，“壓水式反應器肇始事件偵測與辨識之研究”，國立清華大學，2017。
[54] L. H. Chiang, E. L. Russell, and R. D. Braatz, Fault detection and diagnosis in industrial systems: Springer Science & Business Media, 2000.
[55] 王夢蝶，“核電廠未知事件之偵測與辨識”，2020.
[56] P. Sanguansat, Principal component analysis: BoD–Books on Demand, 2012.
[57] 林廷翰，“訊號特徵萃取及感測器偵選於核能電廠肇始事件辨識之研究”，國立清華大學，2016。
[58] T.-H. Lin, S.-C. Wu, K.-Y. Chen, and H.-P. Chou, “Feature extraction and sensor selection for NPP initiating event identification,” Annals of Nuclear Energy, vol. 103, pp. 384-392, 2017.
[59] O. Kramer, "K-nearest neighbors," Dimensionality reduction with unsupervised nearest neighbors, pp. 13-23: Springer, 2013.
[60] Y. LeCun, and Y. Bengio, “Convolutional networks for images, speech, and time series,” The handbook of brain theory and neural networks, vol. 3361, pp. 1995, 1995.
[61] M. A. Wani, F. A. Bhat, S. Afzal, and A. I. Khan, Advances in deep learning: Springer, 2019.
[62] R. Atienza, Advanced Deep Learning with Keras: Apply deep learning techniques, autoencoders, GANs, variational autoencoders, deep reinforcement learning, policy gradients, and more: Packt Publishing Ltd, 2018.
[63] U. Stańczyk, and L. C. Jain, Feature selection for data and pattern recognition: Springer, 2015.
[64] H. Hytti, R. Takalo, and H. Ihalainen, “Tutorial on multivariate autoregressive modelling,” Journal of clinical monitoring and computing, vol. 20, pp. 101-108, 2006.
[65] G. P. Nason, “Stationary and non-stationary time series,” Statistics in volcanology, vol. 60, 2006.
[66] B. Carter, and R. Mancini, Op Amps for everyone: Newnes, 2017.
[67] J. Schmidhuber, “Deep learning in neural networks: An overview,” Neural networks, vol. 61, pp. 85-117, 2015.
[68] R. Hecht-Nielsen, "Theory of the backpropagation neural network," Neural networks for perception, pp. 65-93: Elsevier, 1992.
[69] S. Hochreiter, and J. Schmidhuber, “Long short-term memory,” Neural computation, vol. 9, pp. 1735-1780, 1997.
[70] Y.-L. Kong, Q. Huang, C. Wang, J. Chen, J. Chen, and D. He, “Long short-term memory neural networks for online disturbance detection in satellite image time series,” Remote Sensing, vol. 10, pp. 452, 2018.
[71] L. Prechelt, "Early stopping-but when?," Neural Networks: Tricks of the trade, pp. 55-69: Springer, 1998.
[72] J. Han, J. Pei, and M. Kamber, Data mining: concepts and techniques: Elsevier, 2011.
[73] C. S. Burrus, R. A. Gopinath, and H. Guo, Introduction to wavelets and wavelet transforms: a primer, 1997.
[74] J. Yang, D. Zhang, A. F. Frangi, and J.-y. Yang, “Two-dimensional PCA: a new approach to appearance-based face representation and recognition,” IEEE transactions on pattern analysis and machine intelligence, vol. 26, pp. 131-137, 2004.
[75] M. D. Vose, The simple genetic algorithm: foundations and theory: MIT press, 1999.
[76] D. F. Specht, “Probabilistic neural networks,” Neural networks, vol. 3, pp. 109-118, 1990.
[77] S. Theodoridis, A. Pikrakis, K. Koutroumbas, and D. Cavouras, Introduction to pattern recognition: a matlab approach: Academic Press, 2010.
[78] M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, and M. Isard, "Tensorflow: A system for large-scale machine learning." pp. 265-283.
[79] D. P. Kingma, and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014.
[80] V. Dumoulin, and F. Visin, “A guide to convolution arithmetic for deep learning,” arXiv preprint arXiv:1603.07285, 2016.
[81] M. Lin, Q. Chen, and S. Yan, “Network in network,” arXiv preprint arXiv:1312.4400, 2013.
[82] R. Takalo, H. Hytti, and H. Ihalainen, “Tutorial on univariate autoregressive spectral analysis,” Journal of clinical monitoring and computing, vol. 19, pp. 401-410, 2005.
[83] K. Sameshima, and L. A. Baccala, Methods in brain connectivity inference through multivariate time series analysis: CRC press, 2016.
[84] Y. Sakamoto, M. Ishiguro, and G. Kitagawa, “Akaike information criterion statistics,” Dordrecht, The Netherlands: D. Reidel, vol. 81, pp. 26853, 1986.
[85] S. Yue, and P. Pilon, “Interaction between deterministic trend and autoregressive process,” Water Resources Research, vol. 39, 2003.
[86] L. Yin, R. Yang, M. Gabbouj, and Y. Neuvo, “Weighted median filters: a tutorial,” IEEE Transactions on circuits and systems II: analog and digital signal processing, vol. 43, pp. 157-192, 1996.
[87] 台灣電力公司第三核能發電廠，“核能發電訓練基本教材-壓水式反應器系統介紹”，1983.
[88] Y. Yu, W. Li, D. Sheng, and J. Chen, “A novel sensor fault diagnosis method based on modified ensemble empirical mode decomposition and probabilistic neural network,” Measurement, vol. 68, pp. 328-336, 2015.