Fault Warning for Gas Turbine Combustion Chamber Based on Deep Transfer Learning

doi:10.16183/j.cnki.jsjtu.2024.200

Abstract

Abstract:

Aiming at the inefficiency of gas turbine combustion chamber fault warning due to the scarcity of sample data, a method based on transfer learning and convolutional neural network-bidirectional gated recurrent unit (CNN-BiGRU) network is proposed. First, the CNN-BiGRU network is pre-trained in the source domain using K-fold cross validation and the optimal model is filtered out. Then, the target domain fault warning model is obtained using transfer learning. The target domain data is imported into the model and a sliding window is used to reduce the false alarms caused by the abnormal data points. Finally, the fault occurrence is determined by the warning thresholds calculated using the triple standard deviation (3-sigma) method. The experimental results show that compared with the non-transfer learning model, the proposed method reduces the root mean square error by 87.5% and the mean absolute error by 89.05%, while improving the R-squared by 6.39% under conditions of insufficient sample data. In addition, compared with the system alarm moment, it can detect the signs of faults earlier, providing a fault warning for the gas turbine combustion chamber 82 min in advance.

Key words: gas turbine, combustion chamber, fault warning, transfer learning, convolutional neural network (CNN), bidirectional gated recurrent unit (BiGRU) network

CLC Number:

TK472

WU Yajun, KANG Yingwei. Fault Warning for Gas Turbine Combustion Chamber Based on Deep Transfer Learning[J]. Journal of Shanghai Jiao Tong University, 2026, 60(5): 848-859.

Figures/Tables 15

Fig.1

Fig.2

Fig.3

Fig.4

Tab.1

Fig.5

Tab.2

Fig.6

Tab.3

Fig.7

Tab.4

Tab.5

Fig.8

Fig.9

Fig.10

References 24

[1]	ZHOU D J, YAO Q B, WU H, et al. Fault diagnosis of gas turbine based on partly interpretable convolutional neural networks[J]. Energy, 2020, 200: 117467. doi: 10.1016/j.energy.2020.117467 URL
[2]	尹杰, 刘博, 孙国兵, 等. 基于迁移学习和降噪自编码器-长短时间记忆的锂离子电池剩余寿命预测[J]. 电工技术学报, 2024, 39(1): 289-302.
	YIN Jie, LIU Bo, SUN Guobing, et al. Transfer learning denoising autoencoder-long short term memory for remaining useful life prediction of li-ion batteries[J]. Transactions of China Electrotechnical Society, 2024, 39(1): 289-302.
[3]	黄伟, 张泽发. 基于多元状态估计的燃烧室故障预警研究[J]. 汽轮机技术, 2020, 62(1): 38-42.
	HUANG Wei, ZHANG Zefa. Early warning of combustion chamber faults based on multivariate state estimation[J]. Turbine Technology, 2020, 62(1): 38-42.
[4]	SARWAR U, MUHAMMAD M, MOKHTAR A A, et al. Hybrid intelligence for enhanced fault detection and diagnosis for industrial gas turbine engine[J]. Results in Engineering, 2024, 21: 101841. doi: 10.1016/j.rineng.2024.101841 URL
[5]	CHENG K R, WANG Y Z, YANG X L, et al. An intelligent online fault diagnosis system for gas turbine sensors based on unsupervised learning method LOF and KELM[J]. Sensors and Actuators A: Physical, 2024, 365: 114872. doi: 10.1016/j.sna.2023.114872 URL
[6]	HU M G, HE Y, LIN X Z, et al. Digital twin model of gas turbine and its application in warning of performance fault[J]. Chinese Journal of Aeronautics, 2023, 36(3): 449-470. doi: 10.1016/j.cja.2022.07.021 URL
[7]	彭道刚, 姬传晟, 涂煊, 等. 基于LSTM-SVM的燃气轮机压气机故障预警研究[J]. 动力工程学报, 2021, 41(5): 394-399. doi: 10.19805/j.cnkij.cspe.2021.05.008
	PENG Daogang, JI Chuansheng, TU Xuan, et al. Research on gas turbine compressor fault early warning based on LSTM-SVM[J]. Journal of Chinese Society of Power Engineering, 2021, 41(5): 394-399. doi: 10.19805/j.cnkij.cspe.2021.05.008
[8]	周锐, 康英伟. 基于CNN-LSTM的燃气轮机燃烧室故障预警[J]. 热能动力工程, 2024, 39(1): 191-197.
	ZHOU Rui, KANG Yingwei. Fault warning of gas turbine combustor based on CNN-LSTM[J]. Journal of Engineering for Thermal Energy and Power, 2024, 39(1): 191-197.
[9]	刘广孚, 姜霄, 杜玉龙, 等. 基于长短时记忆神经网络的潜油电泵故障预警[J]. 中国石油大学学报(自然科学版), 2022, 46(5): 170-176.
	LIU Guangfu, JIANG Xiao, DU Yulong, et al. Fault early warning of electric submersible pump based on long short-term memory neural network[J]. Journal of China University of Petroleum (Edition of Natural Science), 2022, 46(5): 170-176.
[10]	谢岳生, 万震天, 李俊昆. 融合热力学模型与人工智能的燃气轮机压气机典型故障预警方法研究[J]. 热力发电, 2024, 53(3): 117-125.
	XIE Yuesheng, WAN Zhentian, LI Junkun. Typical fault warning method of gas turbine compressor combining thermodynamic model with artificial neural network[J]. Thermal Power Generation, 2024, 53(3): 117-125.
[11]	苏静雷, 王红军, 王政博, 等. 多通道卷积神经网络和迁移学习的燃气轮机转子故障诊断方法[J]. 电子测量与仪器学报, 2023, 37(3): 132-140.
	SU Jinglei, WANG Hongjun, WANG Zhengbo, et al. Fault diagnosis method of gas turbine rotor with multi-channel convolutional neural network and transfer learning[J]. Journal of Electronic Measurement and Instrumentation, 2023, 37(3): 132-140.
[12]	唐竞鹏, 王红军, 钟建琳, 等. 基于WDCNN-SVM深度迁移学习的燃气轮机转子故障诊断方法[J]. 电子测量与仪器学报, 2021, 35(11): 115-123.
	TANG Jingpeng, WANG Hongjun, ZHONG Jianlin, et al. Gas turbine rotor fault diagnosis method based on WDCNN-SVM deep transfer learning[J]. Journal of Electronic Measurement and Instrumentation, 2021, 35(11): 115-123.
[13]	ZHUANG J C, JIA M P, HUANG C G, et al. Health prognosis of bearings based on transferable autoregressive recurrent adaptation with few-shot learning[J]. Mechanical Systems and Signal Processing, 2024, 211: 111186. doi: 10.1016/j.ymssp.2024.111186 URL
[14]	CHEN Y Q, WU Z Q, ZHANG B J, et al. Transfer learning and pretraining enhanced physics-informed machine learning for closed-circuit reverse osmosis modeling[J]. Desalination, 2024, 580: 117557. doi: 10.1016/j.desal.2024.117557 URL
[15]	LZADI J M, HASSANI P, RAEESI M, et al. A novel WaveNet-GRU deep learning model for PEM fuel cells degradation prediction based on transfer learning[J]. Energy, 2024, 293: 130602. doi: 10.1016/j.energy.2024.130602 URL
[16]	MENG J H, YOU Y Q, LIN M Q, et al. Multi-scenarios transferable learning framework with few-shot for early lithium-ion battery lifespan trajectory prediction[J]. Energy, 2024, 286: 129682. doi: 10.1016/j.energy.2023.129682 URL
[17]	ZHANG Y P, ACHYUT S. Enhancing supply chain transparency and risk management using CNN-LSTM with transfer learning[J]. Journal of Organizational and End User Computing, 2023, 35(1): 1-22.
[18]	王克定, 李敬兆, 石晴, 等. 基于深度迁移学习的矿井通风机轴承故障诊断[J]. 机床与液压, 2023, 51(22): 209-214.
	WANG Keding, LI Jingzhao, SHI Qing, et al. Bearing fault diagnosis for mine ventilator based on deep transfer learning[J]. Machine Tool & Hydraulics, 2023, 51(22): 209-214.
[19]	郭俊锋, 王淼生, 王智明. 基于对不平衡数据集进行二次迁移学习的滚动轴承剥落类故障诊断方法[J]. 上海交通大学学报, 2023, 57(11): 1512-1521. doi: 10.16183/j.cnki.jsjtu.2022.008
	GUO Junfeng, WANG Miaosheng, WANG Zhiming. Fault diagnosis of rolling bearing with roller spalling based on two-step transfer learning on unbalanced dataset[J]. Journal of Shanghai Jiao Tong University, 2023, 57(11): 1512-1521.
[20]	詹可, 朱仁传. 一种CNN-LSTM船舶运动极值预报模型[J]. 上海交通大学学报, 2023, 57(8): 963-971. doi: 10.16183/j.cnki.jsjtu.2022.089
	ZHAN Ke, ZHU Renchuan. A CNN-LSTM ship motion extreme value prediction model[J]. Journal of Shanghai Jiao Tong University, 2023, 57(8): 963-971.
[21]	杨婷婷, 高乾, 李浩千, 等. 基于卷积神经网络-长短时记忆神经网络的磨煤机故障预警[J]. 热力发电, 2022, 51(10): 122-129.
	YANG Tingting, GAO Qian, LI Haoqian, et al. Coal mill fault warning technology based on CNN-LSTM network[J]. Thermal Power Generation, 2022, 51(10): 122-129.
[22]	王瑞, 徐新超, 逯静. 基于特征选择及ISSA-CNN-BiGRU的短期风功率预测[J]. 工程科学与技术, 2024, 56(3): 228-239.
	WANG Rui, XU Xinchao, LU Jing. Short-term wind power prediction: feature selection and ISSA-CNN-BiGRU approach[J]. Advanced Engineering Sciences, 2024, 56(3): 228-239.
[23]	WANG W P, WANG Z R, ZHOU Z F, et al. Anomaly detection of industrial control systems based on transfer learning[J]. Tsinghua Science and Technology, 2021, 26(6): 821-832. doi: 10.26599/TST.2020.9010041 URL
[24]	黄伟, 张泽发. 基于相似度分析的电站燃气轮机燃烧室故障预警研究[J]. 上海电力大学学报, 2020, 36(3): 220-224.
	HUANG Wei, ZHANG Zefa. Research on early warning of gas turbine chamber based on similarity analysis[J]. Journal of Shanghai University of Electric Power, 2020, 36(3): 220-224.

参数	参数描述	正常工况(0.8)	故障工况 (燃烧室喷嘴堵塞)	变化原因
燃烧室压力/MPa	燃烧室内气体压力,反映燃烧效率	3.5	4.0	燃烧室喷嘴堵塞导致气体无法顺利排出,增加了燃烧室内压力
CSO	控制阀门的开度	60	68	为了补偿堵塞导致的燃气流量减少,控制系统增加了阀门开度
压气机出口温度/℃	空气或气体在通过压气机后,从压气机出口处测得的温度	400	443	压气机负荷增加导致温度上升
压气机出口压力/MPa	空气或气体在通过压气机后,从压气机出口处测得的压力	1.5	1.8	增加的负荷导致压气机出口压力升高
燃气透平膨胀比	燃气透平中气体膨胀的比例,反映了透平的工作效率	12	10.8	由于气流不足,透平的膨胀效率下降,导致膨胀比降低
压比	进气压力与排气压力之比	10	9.5	压气机效率下降导致压比略有下降
压气机排气温度/℃	压缩后的气体离开压气机时的温度	450	475	压气机负荷增加导致温度升高

状态参数	源域数值	目标域数值
CSO	70	47
压气机出口温度/℃	424	—
压气机出口压力/MPa	1.8	—
透平排气温度/℃	674	570
燃气透平膨胀比	16	—
压比	—	15
压气机排气温度/℃	—	412
燃烧室压力/MPa	—	3.3

参数	参数配置
序列长度(Sequence)	Source_sequence为3
	Target_sequence为3
学习率(Lr)	Source_lr为0.001
	Target_lr为0.000 1
迭代次数(Epoch)	Source_epoch为100
	Target_epoch为30
K折数(K_folds)	5
批量大小(Batchsize)	Source_batchsize为128
	Target_batchsize为64
激活函数(Activation function)	ReLU
L2正则化系数(L2_reg)	0.000 1
隐藏层大小(Hidden_size)	64
层数(Num_layers)	2

模型	e_RMSE/℃	e_MAE/℃	R²
CNN	0.009 2	0.008 4	0.971 9
BiGRU	0.076 4	0.064 0	0.826 9
TCN	0.028 9	0.019 7	0.941 5
TCN-BiGRU	0.006 3	0.005 6	0.984 6
CNN-BiGRU	0.004 7	0.003 3	0.995 0

模型	e_RMSE/℃	e_MAE/℃	R²
无K折交叉验证	0.006 8	0.006 1	0.980 7
3层CNN	0.005 9	0.005 2	0.987 8
5层CNN	0.010 4	0.008 9	0.969 0
1层BiGRU	0.019 9	0.015 8	0.953 0
3层BiGRU	0.015 8	0.012 2	0.961 6
完整模型	0.004 7	0.003 3	0.995 0