基于深度自编码器和支持向量数据描述的燃气轮机高温部件异常检测

doi:10.12096/j.2096-4528.pgt.21021

发电技术 ›› 2021, Vol. 42 ›› Issue (4): 422-430.DOI: 10.12096/j.2096-4528.pgt.21021

基于深度自编码器和支持向量数据描述的燃气轮机高温部件异常检测

白明亮¹(), 张冬雪²(), 刘金福²^,*(), 刘娇³(), 于达仁¹^,²()

¹ 哈尔滨工业大学控制科学与工程系, 黑龙江省哈尔滨市 150001
² 哈尔滨工业大学能源科学与工程学院, 黑龙江省哈尔滨市 150001
³ 中国航空工业集团公司沈阳飞机设计研究所, 辽宁省沈阳市 110035

收稿日期:2021-03-23 出版日期:2021-08-31 发布日期:2021-07-22
通讯作者: 刘金福
作者简介:白明亮(1996), 男, 博士研究生, 研究方向为智能故障诊断、时间序列数据预测与机器学习, mingliangbai@outlook.com
张冬雪(1996), 女, 硕士研究生, 研究方向为动力装置与能源系统的建模及系统运行优化调度, 19S102132@stu.hit.edu.cn
刘娇(1990), 女, 博士, 研究方向为发动机健康管理, liujiaohit@hotmail.com
于达仁(1966), 男, 教授, 博士生导师, 研究方向为动力装置建模仿真、控制与故障诊断, yudaren@hit.edu.cn
基金资助:
国家自然科学基金项目(51976042);国家重大科技专项(2017-I-0007-0008)

Anomaly Detection of Gas Turbine Hot Components Based on Deep Autoencoder and Support Vector Data Description

Mingliang BAI¹(), Dongxue ZHANG²(), Jinfu LIU²^,*(), Jiao LIU³(), Daren YU¹^,²()

¹ Department of Control Science and Engineering, Harbin Institute of Technology, Harbin 150001, Heilongjiang Province, China
² School of Energy Science and Engineering, Harbin Institute of Technology, Harbin 150001, Heilongjiang Province, China
³ AVIC Shenyang Aircraft Design & Research Institute, Shenyang 110035, Liaoning Province, China

Received:2021-03-23 Published:2021-08-31 Online:2021-07-22
Contact: Jinfu LIU
Supported by:
National Natural Science Foundation of China(51976042);National Science and Technology Major Project of China(2017-I-0007-0008)

摘要/Abstract

摘要：

开展燃气轮机高温部件的异常检测能有效提高其运行安全性和可靠性。随着人工智能技术的兴起，数据驱动的故障诊断方法已经越来越流行。然而，在实际应用中，燃气轮机故障数据很少甚至几乎没有。针对仅有正常数据场景下的燃气轮机高温部件异常检测问题，提出了一种基于深度自编码器(deep autoencoder，DAE)和支持向量数据描述(support vector data description，SVDD)融合的DAE-SVDD异常检测方法。该方法利用正常数据训练深度自编码器，并利用深度自编码器的重构误差来训练支持向量数据描述。与传统异常检测方法相比，该方法显著提高了异常检测精度，能实现更灵敏鲁棒的燃气轮机高温部件异常检测。

关键词: 燃气轮机, 高温部件, 深度自编码器(DAE), 支持向量数据描述(SVDD), 异常检测, 故障诊断

Abstract:

Anomaly detection of gas turbine hot components can ensure its operational safety and reliability. With the boom of artificial intelligence, data-driven fault diagnosis is becoming increasingly popular. However, in actual applications, fault data of gas turbines are rare or even unavailable. Aiming to solve the anomaly detection problem of gas turbine hot components in the case of only normal data available, this paper proposed an anomaly detection method based on the fusion of deep autoencoder and support vector data description. This method uses normal data to train deep autoencoder and then uses the reconstruction errors of deep autoencoder to train support vector data description. Experiments show that, compared with conventional anomaly detection methods, the proposed method can significantly improve the anomaly detection accuracy and realize more sensitive and robust anomaly detection of gas turbine hot components.

Key words: gas turbine, hot components, deep autoencoder (DAE), support vector data description (SVDD), anomaly detection, fault diagnosis

中图分类号:

TK05

白明亮, 张冬雪, 刘金福, 刘娇, 于达仁. 基于深度自编码器和支持向量数据描述的燃气轮机高温部件异常检测[J]. 发电技术, 2021, 42(4): 422-430.

Mingliang BAI, Dongxue ZHANG, Jinfu LIU, Jiao LIU, Daren YU. Anomaly Detection of Gas Turbine Hot Components Based on Deep Autoencoder and Support Vector Data Description[J]. Power Generation Technology, 2021, 42(4): 422-430.

图/表 14

图1 透平出口处的热电偶示意图

Fig.1 Thermocouples at turbine outlet

图2 热电偶的布置示意图

Fig.2 Layout of thermocouples

图3 燃气旋转示意图

Fig.3 Gas rotation

图4 燃气旋转对透平排气温度的影响

Fig.4 Effect of gas rotation on turbine exhaust gas temperature

图5 深度自编码器的原理图

Fig.5 Principle of deep autoencoder

图6 高维空间内的超球体

Fig.6 Hypersphere in high-dimensional space

图7 DAE-SVDD异常检测算法流程

Fig.7 Anomaly detection procedure of DAE-SVDD

表1 数据集描述

Tab. 1 Dataset description

参数	正常数据			异常数据
参数	训练集	验证集	测试集	异常数据
样本数目	2 100	450	450	500

图8 训练集中前500个样本的工况参数变化

Fig.8 The variation of working condition parameters of the first 500 samples in the training set

表2 深度自编码器训练结果

Tab. 2 training result of deep autoencoder

参数	正常数据			故障数据
参数	训练集	验证集	测试集	故障数据
E_RMSE	0.030 3	0.029 9	0.029 5	0.067 3

图9 不同数据中距离与超球体半径之间的关系

Fig.9 Relationship between distance and radius of hypersphere in different data

表3 DAE-SVDD异常检测精度

Tab. 3 Anomaly detection accuracy of DAE-SVDD

参数	正常数据			故障数据
参数	训练集	验证集	测试集	故障数据
检测精度	0.969 0	0.968 9	0.982 2	0.946 0

图10 正常数据的训练集、验证集、测试集以及故障数据中样本到球心距离与超球半径的关系

Fig.10 Relationship between the distance from the center of the hypersphere and the radius of the hypersphere in the training set, validation set and test set of normal data as well as fault data

表4 DAE-SVDD与SVDD的异常检测精度比较

Tab. 4 Anomaly detection accuracy comparison between of DAE-SVDD and SVDD

算法	正常数据			故障数据
算法	训练集	验证集	测试集	故障数据
SVDD	0.948 1	0.946 7	0.953 3	0.926 0
DAE-SVDD	0.969 0	0.968 9	0.982 2	0.946 0

参考文献 21

1	薛晓东, 韩巍, 王晓东, 等. 适合分布式冷热电联供系统的中小型发电装置[J]. 发电技术, 2020, 41 (3): 252- 260.
	XUE X D , HAN W , WANG X D , et al. Small and medium-scale power generation devices suiting for distributed combined cooling, heating and power system[J]. Power Generation Technology, 2020, 41 (3): 252- 260.
2	丁继伟, 李岩. 基于弹塑性理论的燃机转子寿命分析[J]. 发电技术, 2020, 41 (2): 202- 205.
	DING J W , LI Y . Analysis on gas turbine rotor life based on elastoplastic theory[J]. Power Generation Technology, 2020, 41 (2): 202- 205.
3	苏烨, 丁俊宏, 丁宁, 等. 全国燃气轮机联合循环机组热控系统典型故障分析及预控措施建议[J]. 浙江电力, 2020, 39 (8): 95- 102.
	SU Y , DING J H , DING N , et al. Typical fault analysis and proposals of pre-control measures for thermal control system of national gas turbine combined cycle units[J]. Zhejiang Electric Power, 2020, 39 (8): 95- 102.
4	杨卫国, 祝铁军, 王庆韧, 等. 燃气轮机组压气机失速引发的不稳定振动分析[J]. 广东电力, 2019, 32 (7): 37- 43.
	YANG W G , ZHU T J , WANG Q R , et al. Analysis on unstable vibration caused by stall of gas turbine compressor[J]. Guangdong Electric Power, 2019, 32 (7): 37- 43.
5	朱俊杰, 王晓维, 董玉亮, 等. 基于工况辨识的重型燃气轮机性能评价方法研究[J]. 智慧电力, 2020, 48 (7): 24- 29. DOI
	ZHU J J , WANG X W , DONG Y L , et al. Performance evaluation method of heavy duty gas turbine based on condition identification[J]. Smart Power, 2020, 48 (7): 24- 29. DOI
6	白明亮. 故障案例稀缺场景下的燃气轮机智能故障检测及诊断研究[D]. 哈尔滨: 哈尔滨工业大学, 2020.
	BAI M L. Study on intelligent fault detection and diagnosis of gas turbine under the scarcity of fault cases[D]. Harbin: Harbin Institute of Technology, 2020.
7	刘娇. 燃气轮机高温部件故障早期预警研究[D]. 哈尔滨: 哈尔滨工业大学, 2019.
	LIU J. Research on early warning of gas turbine high temperature components fault[D]. Harbin: Harbin Institute of Technology, 2019.
8	TAHAN M , TSOUTSANIS E , MUHAMMAD M , et al. Performance-based health monitoring, diagnostics and prognostics for condition-based maintenance of gas turbines: a review[J]. Applied Energy, 2017, 198, 122- 144. DOI
9	LIN X S, LI B W, YANG X Y. Engine components fault diagnosis using an improved method of deep belief networks[C]//7th International Conference on Mechanical and Aerospace Engineering(ICMAE). IEEE, 2016: 454-459.
10	WONG P K , YANG Z , VONG C M , et al. Real-time fault diagnosis for gas turbine generator systems using extreme learning machine[J]. Neurocomputing, 2014, 128 (3): 249- 257.
11	MARAGOUDAKIS M, LOUKIS E, PANTELIDES P P. Random forests identification of gas turbine faults[C]//19th International Conference on Systems Engineering. IEEE, 2008: 127-132.
12	WANG L , LI Y G , ABDUL GHAFIR M F , et al. A rough set-based gas turbine fault classification approach using enhanced fault signatures[J]. Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy, 2011, 225 (8): 1052- 1065. DOI
13	LECUN Y , BENGIO Y , HINTON G , et al. Deep learning[J]. Nature, 2015, 521 (7553): 436- 444. DOI
14	YAN L P , DONG X Z , WANG T , et al. A fault diagnosis method for gas turbines based on improved data preprocessing and an optimization deep belief network[J]. Measurement Science and Technology, 2019, 31 (1): 151- 162. DOI
15	FU X Y , LUO H , ZHONG S , et al. Aircraft engine fault detection based on grouped convolutional denoising autoencoders[J]. Chinese Journal of Aeronautics, 2019, 32 (2): 296- 307. DOI
16	MULEWICZ B , MARZEC M , MORKISZ P , et al. Failures prediction based on performance monitoring of a gas turbine: a binary classification approach[J]. Schedae Information, 2018, 38 (26): 9- 21.
17	LIU J , LIU J , YU D , et al. Fault detection for gas turbine hot components based on a convolutional neural network[J]. Energies, 2018, 11 (8): 2149- 2158. DOI
18	赵志宏, 李乐豪, 杨绍普, 等. 一种频域特征提取自编码器及其在故障诊断中的应用研究[J/OL]. 中国机械工程: 1-8[2021-03-16]. http://kns.cnki.net/kcms/detail/42.1294.th.20210316.1534.006.html.
	ZHAO Z H, LI L H, YANG S P, et al. A frequency domain feature extraction auto-encoder and its application on fault diagnosis[J/OL]. China Mechanical Engineering: 1-8[2021-03-16]. http://kns.cnki.net/kcms/detail/42.1294.th.20210316.1534.006.html.
19	顾珉, 施华君. 基于降噪自编码器降维的汽车行驶工况分析[J]. 计算机系统应用, 2021, 30 (1): 38- 44.
	GU M , SHI H J . Analysis of vehicle driving condition based on de-noise autoencoder[J]. Computer Systems & Applications, 2021, 30 (1): 38- 44.
20	周云成, 邓寒冰, 许童羽, 等. 基于稠密自编码器的无监督番茄植株图像深度估计模型[J]. 农业工程学报, 2020, 36 (11): 182- 192.
	ZHOU Y C , DENG H B , XU T Y , et al. Unsupervised deep estimation modeling for tomato plant image based on dense convolutional auto-encoder[J]. Transactions of the Chinese Society of Agricultural Engineering, 2020, 36 (11): 182- 192.
21	TAX D M J , DUIN R P W . Support vector data description[J]. Machine learning, 2004, 54 (1): 45- 66. DOI

[1]	崔则阳, 孔祥玲, 付经伦, 施佳君. 一种基于图像的燃气轮机叶型参数测量方法[J]. 发电技术, 2024, 45(1): 106-112.
[2]	杨旸, 李耀强, 张金琦. 基于数值方法的燃气轮机贫预混旋流燃烧室单头部结构设计[J]. 发电技术, 2023, 44(5): 712-721.
[3]	杨旸, 郭德三, 李耀强, 张金琦. 燃气轮机贫预混多旋流组合燃烧室头部结构设计[J]. 发电技术, 2023, 44(2): 183-192.
[4]	张弘毅, 曲立涛. 9F级燃机选择性催化还原脱硝数值模拟研究与应用[J]. 发电技术, 2023, 44(1): 78-84.
[5]	张航, 周传杰, 张林, 陈节涛, 徐春梅, 彭道刚. 基于概率神经网络-小波神经网络-DS信息融合的电厂引风机故障诊断[J]. 发电技术, 2022, 43(6): 951-958.
[6]	金云峰, 刘超, 邓高峰, 关运龙, 郝建刚, 黄海舟, 蒋东翔. 燃气轮机进气过滤系统维修周期经济性分析[J]. 发电技术, 2022, 43(1): 119-125.
[7]	孔祥玲, 付经伦. 基于计算机视觉的三维重建技术在燃气轮机行业的应用及展望[J]. 发电技术, 2021, 42(4): 454-463.
[8]	邱彬, 付经伦. 燃气轮机排气扩压器研究现状[J]. 发电技术, 2021, 42(4): 437-446.
[9]	魏凯, 罗忠, 孙永航, 王宇. 燃气轮机带阀片引射混合器内流特性分析[J]. 发电技术, 2021, 42(4): 431-436.
[10]	朱华, 严彪, 刘雨松, 李亮. 湿空气透平冷却技术研究[J]. 发电技术, 2021, 42(4): 412-421.
[11]	管金, 何宗泽, 吕小静, 翁一武. 30kW微型燃气轮机发电机组启动实验研究[J]. 发电技术, 2021, 42(4): 404-411.
[12]	金云峰, 刘超, 邓高峰, 关运龙, 田鑫, 黄海舟, 蒋东翔. 燃气轮机进气压差建模方法研究[J]. 发电技术, 2021, 42(4): 395-403.
[13]	陈尚年, 李录平, 张世海, 欧阳敏南, 樊昂, 文贤馗. 汽轮发电机组振动故障诊断技术研究进展[J]. 发电技术, 2021, 42(4): 489-499.
[14]	黄友华, 马善为, 刘吉, 武正华. 燃气轮机烟气SCR脱硝系统优化设计与工程应用[J]. 发电技术, 2021, 42(3): 350-356.
[15]	李亚光, 李蒙. 基于深度小世界神经网络的风电机组异常检测[J]. 发电技术, 2021, 42(3): 313-321.

基于深度自编码器和支持向量数据描述的燃气轮机高温部件异常检测

Anomaly Detection of Gas Turbine Hot Components Based on Deep Autoencoder and Support Vector Data Description

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 21

相关文章 15

编辑推荐

Metrics

本文评价