The lack of interpretability in neural networks hinders fault diagnosis and optimal feature selection, leading to ineffective anomaly detection of gas turbines. This study proposes an anomaly detection framework for gas turbines that integrates explainable artificial intelligence with the physical intelligence of expert systems. This framework aims to enhance the interpretability of a neural network for optimizing the input features and localizing faults. The proposed framework features two key characteristics. First, the explainable training phase determines optimal features based on recursive feature elimination with reconstruction error and verification of expert systems. Second, the explainable inference phase executes real-time fault localization with an extremely low false alarm by addressing the cumulative alarm and feature-wise reconstruction error. The proposed framework is validated using extensive operational data including various fault modes, collected from combined cycle power plants in the Republic of Korea. Extensive experiments demonstrate that the proposed framework achieved exceptional accuracy with an F1-score of 0.961 and a false alarm rate of 1.98e-4, outperforming other anomaly detection methods. The proposed framework also effectively localizes the modes and causal effects of each fault even though the proposed neural network is trained only with normal data. Extensive case studies including sensitivity analysis and variations in training data size clearly reveal the reliability of the proposed framework. These systematic analyses confirm that the proposed framework is effective in detecting anomalies in real-world applications for efficient operation and maintenance of gas turbines.
RayaproluK. Boilers for power and process. Boca Raton: CRC Press, 2009.
2.
YuriMMasadaJTsukagoshiK, et al. Development of 1600 C-class high-efficiency gas turbine for power generation applying J-Type technology. Mitsubishi Heavy Ind Tech Rev2013; 50(3): 1–10.
3.
AbuduKIgieURoumeliotisI, et al. Impact of gas turbine flexibility improvements on combined cycle gas turbine performance. Appl Therm Eng2021; 189: 116703.
4.
ChandolaVBanerjeeAKumarV.Anomaly detection: a survey. ACM Comput Surv2009; 41(3): 1–58.
5.
TanYKWangYWNiYQ, et al. Parallel reservoir computing-based signal outlier detection and recovery method for structural health monitoring. Develop Built Environ2024; 18: 100463.
6.
HuangKYuenK.Hierarchical outlier detection approach for online distributed structural identification. Struct Control Health Monit2020; 27(11): e2632.
7.
XuXRenYHuangQ, et al. Anomaly detection for large span bridges during operational phase using structural health monitoring data. Smart Mater Struct2020; 29(4): 045029.
8.
BaoYTangZLiH, et al. Computer vision and deep learning-based data anomaly detection method for structural health monitoring. Struct Health Monit2018; 18(2): 401–421.
9.
LiQGaoJBeckJL, et al. Probabilistic outlier detection for robust regression modeling of structural response for high-speed railways track monitoring. Struct Health Monit2023; 23(2): 1280–1296.
10.
ShuXBaoTZhouY, et al. Unsupervised dam anomaly detection with spatial–temporal variational autoencoder. Struct Health Monit2022; 22(1): 39–55.
11.
HanachiHLiuJBanerjeeA, et al. A physics-based modeling approach for performance monitoring in gas turbine engines. IEEE Trans Reliab2014; 64(1): 197–205.
12.
LiY-G.Gas turbine performance and health status estimation using adaptive gas path analysis. J Eng Gas Turbines Power2010; 132(4): 041701.
13.
PourbabaeeBMeskinNKhorasaniK.Sensor fault detection, isolation, and identification using multiple-model-based hybrid Kalman filter for gas turbine engines. IEEE Trans Control Syst Technol2015; 24(4): 1184–1200.
14.
BelovSNikolaevSUzhinskyI.Hybrid data-driven and physics-based modeling for gas turbine prescriptive analytics. Int J Turbomachinery Propuls Power2020; 5(4): 29.
15.
NassifABTalibMANasirQ, et al. Machine learning for anomaly detection: a systematic review. IEEE Access2021; 9: 78658–78700.
16.
ZhouDZhangHWengS.A new gas path fault diagnostic method of gas turbine based on support vector machine. J Eng Gas Turbines Power2015; 137(10): 102605.
17.
ZhongSFuSLinL, et al. A novel unsupervised anomaly detection for gas turbine using isolation forest. In: 2019 IEEE international conference on prognostics and health management (ICPHM), San Francisco, California, 2019, pp. 1–6. IEEE.
18.
ChengTDairiAHarrouF, et al. Monitoring influent conditions of wastewater treatment plants by nonlinear data-based techniques. IEEE Access2019; 7: 108827–108837.
19.
XuHPangGWangY, et al. Deep isolation forest for anomaly detection. IEEE Trans Knowl Data Eng2023; 35(12): 12591–12604.
20.
ZhaoYPHuangGHuQK, et al. An improved weighted one class support vector machine for turboshaft engine fault detection. Eng Appl Artif Intell2020; 94: 103796.
21.
FuSZhongSLinL, et al. A re-optimized deep auto-encoder for gas turbine unsupervised anomaly detection. Eng Appl Artif Intell2021; 101: 104199.
22.
LeeGJungMSongM, et al. Unsupervised anomaly detection of the gas turbine operation via convolutional auto-encoder. In: 2020 IEEE International conference on prognostics and health management (ICPHM), Detroit, Michigan, 2020, pp. 1–6. IEEE.
23.
FahmiATWKKashyzadehKRGhorbaniS. Fault detection in the gas turbine of the Kirkuk power plant: an anomaly detection approach using DLSTM-Autoencoder. Eng Fail Anal2024; 160: 108213.
24.
HarrouFDairiATaghezouitB, et al. Automatic fault detection in grid-connected photovoltaic systems via variational autoencoder-based monitoring. Energy Convers Manag2024; 314: 118665.
25.
Crespo MárquezA. The curse of dimensionality. In: Digital maintenance management: guiding digital transformation in maintenance, 2022, pp. 67–86. Springer International Publishing.
26.
SamekWMüllerKR.Towards explainable artificial intelligence. In: Explainable AI: interpreting, explaining and visualizing deep learningCham: Springer International Publishing, 2019, pp. 5–22.
27.
YaoCYueyunXJinweiC, et al. A novel gas path fault diagnostic model for gas turbine based on explainable convolutional neural network with LIME method. In: Turbo expo: power for land, sea, and air. ASME, 2021, p. V004T05A008.
28.
NorAKMPedapatiSRMuhammadM.Application of explainable AI (XAI) for anomaly detection and prognostic of gas turbines with uncertainty quantification. Preprints. Epub ahead of print September2021. DOI: 10.20944/preprints202109.0034.v1.
29.
HuangJPengYHuL.A multilayer stacking method base on RFE-SHAP feature selection strategy for recognition of driver’s mental load and emotional state. Expert Syst Appl2024; 238: 121729.
30.
KimINaJYunJP, et al. Deep feature selection framework for quality prediction in injection molding process. IEEE Trans Ind Inf2023; 20(1): 503–512.
31.
LiQJiYZhuM, et al. Unsupervised feature selection using chronological fitting with Shapley Additive explanation (SHAP) for industrial time-series anomaly detection. Appl Soft Comput2024; 155: 111426.
32.
RajaPSThangavelKJSC. Missing value imputation using unsupervised machine learning techniques. Soft Comput2020; 24(6): 4361–4392.
33.
MorshedizadehMKordestaniMCarriveauR, et al. Application of imputation techniques and adaptive neuro-fuzzy inference system to predict wind turbine power production. Energy2017; 138: 394–404.
34.
QuFLiuJMaY, et al. A novel wind turbine data imputation method with multiple optimizations based on GANs. Mech Syst Signal Process2020; 139: 106610.
35.
KornelsenKCoulibalyP.Comparison of interpolation, statistical, and data-driven methods for imputation of missing values in a distributed soil moisture dataset. J Hydrol Eng2014; 19(1): 26–43.
36.
ParkJMüllerJAroraB, et al. Long-term missing value imputation for time series data using deep neural networks. Neural Comput Appl2023; 35(12): 9071–9091.
37.
LiuJBaiMLongZ, et al. Early fault detection of gas turbine hot components based on exhaust gas temperature profile continuous distribution estimation. Energies2020; 13(22): 5950.
38.
YuLJinJ.Study on deflection laws of 9FA gas-turbine exhaust temperature field. Electr Power Constr2009; 3: 26.
39.
GülenSCGriffinPRPaolucciS.Real-time on-line performance diagnostics of heavy-duty industrial gas turbines. J Eng Gas Turbines Power2002; 124(4): 910–921.
40.
JinfuLJiaoLJieW, et al. Early fault detection of hot components in gas turbines. J Eng Gas Turbines Power2017; 139(2): 021201.
41.
HeitorMVWhitelawJH.Velocity, temperature, and species characteristics of the flow in a gas-turbine combustor. Combust Flame1986; 64(1): 1–32.
42.
ZhaoHLiuHHuW, et al. Anomaly detection and fault analysis of wind turbine components based on deep learning network. Renew Energy2018; 127: 825–834.
43.
LiXDingQSunJQ.Remaining useful life estimation in prognostics using deep convolution neural networks. Reliab Eng Syst Saf2018; 172: 1–11.
44.
KoJUNaKOhJS, et al. A new auto-encoder-based dynamic threshold to reduce false alarm rate for anomaly detection of steam turbines. Expert Syst Appl2022; 189: 116094.
45.
AliPJMFarajRHKoyaE, et al. Data normalization and standardization: A technical report. Mach Learn Tech Rep2014; 1(1): 1–6.
46.
HigginsIMattheyLPalA, et al. Beta-VAE: learning basic visual concepts with a constrained variational framework. In: ICLR (Poster), Tuolon, France, 2017, p. 3.
47.
AnJChoS.Variational autoencoder based anomaly detection using reconstruction probability. Spec Lect IE2015; 2(1): 1–18.
48.
XieJSageMZhaoYF.Feature selection and feature learning in machine learning applications for gas turbines: a review. Eng Appl Artif Intell2023; 117: 105591.
49.
CostaNSánchezL.Variational encoding approach for interpretable assessment of remaining useful life estimation. Reliab Eng Syst Saf2022; 222: 108353.
WangYBleiDCunninghamJP.Posterior collapse and latent variable non-identifiability. Adv Neural Inf Process Syst2021; 34: 5443–5455.
52.
WangHLiangQHancockJT, et al. Feature selection strategies: a comparative analysis of SHAP-value and importance-based methods. J Big Data2024; 11(1): 44.
53.
SnoekJLarochelleHAdamsRP.Practical Bayesian optimization of machine learning algorithms. Adv Neural Inf Process Syst2012; 25: 2951–2959.
54.
WangJYangZChenK, et al. Practices of detecting and removing nuisance alarms for alarm overloading in thermal power plants. Control Eng Pract2017; 67: 21–30.
55.
ISAA.ISA-18.2: Management of alarm systems for the process industries. Durham, NC, USA: International Society of Automation, 2009.
56.
MoonJKimMGKangOH, et al. Automatic high-frequency induction brazing through an ensembled detection with heterogeneous sensor measurements. J Intell Manuf2024; 36: 2439–2459.
57.
TulsyanAAlrowaieFGopaluniB.Design and assessment of delay timer alarm systems for nonlinear chemical processes. AIChE J2018; 64(1): 77–90.
58.
OliveiraDFVismariLFNascimentoAM, et al. A new interpretable unsupervised anomaly detection method based on residual explanation. IEEE Access2021; 10: 1401–1409.
59.
PrasovAAKhalilHK.A nonlinear high-gain observer for systems with measurement noise in a feedback control framework. IEEE Trans Autom Control2012; 58(3): 569–580.
60.
BoydKEngKHPageCD. Area under the precision-recall curve: point estimates and confidence intervals. In: Machine learning and knowledge discovery in databases: European conference, ECML PKDD 2013, Prague, Czech Republic, 23–27 September 2013, Proceedings, Part III. 2013, pp. 451–466. Springer Berlin Heidelberg.
61.
KotsiantisSKanellopoulosDPintelasP.Handling imbalanced datasets: a review. GESTS Int Trans Comput Sci Eng2006; 30(1): 25–36.
62.
JakubowskiJStaniszPBobekS, et al. Anomaly detection in asset degradation process using variational autoencoder and explanations. Sensors2021; 22(1): 291.
63.
KimDAntariksaGHandayaniMP, et al. Explainable anomaly detection framework for maritime main engine sensor data. Sensors2021; 21(15): 5200.
64.
AntwargLMillerRMShapiraB, et al. Explaining anomalies detected by autoencoders using Shapley additive explanations. Expert Syst Appl2021; 186: 115736.
65.
SalamMAAzarATElgendyMS, et al. The effect of different dimensionality reduction techniques on machine learning overfitting problem. Int J Adv Comput Sci Appl2021; 12(4): 641–655.
66.
SchölkopfBWilliamsonRCSmolaA, et al. Support vector method for novelty detection. Adv Neural Inf Process Syst1999; 12: 582–588.
67.
LiuFTTingKMZhouZH.Isolation forest. In: 2008 Eighth IEEE international conference on data mining, Pisa, Italy, 2008, pp. 413–422. IEEE.
68.
RuffLVandermeulenRGoernitzN, et al. Deep one-class classification. In: International conference on machine learning, Stockholm, Sweden, 2018, pp. 4393–4402. PMLR.
69.
LiuYLiZZhouC, et al. Generative adversarial active learning for unsupervised outlier detection. IEEE Trans Knowl Data Eng2019; 32(8): 1517–1528.
70.
AkcaySAtapour-AbarghoueiABreckonTP.GANomaly: semi-supervised anomaly detection via adversarial training. In: Computer vision – ACCV 2018: 14th asian conference on computer vision, Perth, Australia, 2–6 December 2018, Revised Selected Papers, Part III, 2019, pp. 622–637. Springer International Publishing.
71.
BankDKoenigsteinNGiryesR.Autoencoders. In: RokachLMaimonOShmueliE (eds): Machine learning for data science handbook: data mining and knowledge discovery handbook. Cham, Switzerland: Springer, 2023, pp. 353–374.
