﻿ 应用支持向量回归机探索发动机VSV调节规律<sup>*</sup>
 文章快速检索 高级检索

1. 中国民航大学 航空工程学院, 天津 300300;
2. 中国民航大学 工程技术训练中心, 天津 300300

Exploration of engine VSV regulation law using support vector regression
CAO Huiling1, KAN Yuxiang1, XUE Peng2
1. College of Aeronautical Engineering, Civil Aviation University of China, Tianjin 300300, China;
2. Engineering Technology Training Center, Civil Aviation University of China, Tianjin 300300, China
Received: 2017-08-11; Accepted: 2017-12-04; Published online: 2018-01-15 17:18
Foundation item: the Fundamental Research Funds for the Central Universities (3122014D010)
Corresponding author. CAO Huiling.E-mail:hlcao@cauc.edu.cn
Abstract: The engine variable stator vane (VSV) regulation law is extremely complex, and through mining quick access recorder (QAR) data, the VSV regulation law is studied. Firstly, the support vector regre-ssion (SVR) model based on particle swarm optimization (PSO) is established through the QAR data of PW4077D engine health condition to explore the regulation law of VSV. Then, the PSO-SVR model is validated by the subsequent flight data, and the verification results are compared with the traditional PSO-BP neural network model. Finally, the PSO-SVR model is applied to engine fault diagnosis. The results show that the regression prediction accuracy of the PSO-SVR model is better than that of the PSO-BP neural network model, and it can accurately reflect the VSV regulation rule. It can be used in the condition monitoring and fault dia-gnosis of engine, and can also provide reference for the design of VSV control system.
Key words: engine variable stator vane (VSV)     regulation law     support vector regression (SVR)     particle swarm optimization (PSO) algorithm     quick access recorder(QAR) data     fault diagnosis

1 算法理论 1.1 支持向量回归机

 (1)

 (2)

 (3)

 (4)

 (5)

1.2 粒子群优化算法

 图 1 PSO算法流程图 Fig. 1 Flowchart of PSO algorithm
2 VSV调节规律模型建立 2.1 数据准备

2.1.1 数据选取

 参数 相关系数 N1 CMD 0.986 N1 0.986 Tt3 0.973 N2 0.939 Ma 0.901 Tt25 0.895 ALTC 0.871 ALT 0.871 TRA 0.820 BP 0.778 WF 0.702 TAT -0.750 Pt5 -0.767 Tt2 -0.809 SAT -0.851 Pt2 -0.864

2.1.2 数据预处理

 (6)
 (7)

 (8)

 (9)

2.2 支持向量回归机参数优化

 图 2 SVR模型训练相关系数和测试相关系数随惩罚参数的变化(σ=5) Fig. 2 Variation of SVR model training correlation coefficient and test correlation coefficient with penalty parameter(σ=5)
 图 3 SVR模型训练相关系数和测试相关系数随核参数的变化(C=300) Fig. 3 Variation of SVR model training correlation coefficient and test correlation coefficient with kernel function parameter(C=300)

 图 4 PSO算法进化曲线 Fig. 4 Evolution curves of PSO algorithm
2.3 模型建立与验证

 图 5 PSO-SVR模型训练的相对误差率 Fig. 5 Relative error rate of PSO-SVR model training

 验证航班 PSO-SVR模型 PSO-BP神经网络模型 MSE R2/% MSE R2/% a 8.20×10-4 99.57 1.05 97.60 b 6.08×10-4 99.76 0.65 99.03 c 1.00×10-3 99.67 1.34 98.69 d 7.80×10-3 99.69 0.97 99.08

3 基于PSO-SVR模型的故障诊断

 图 6 2个航班的监控结果 Fig. 6 Monitoring results for two flights

4 结论

1) 支持向量回归机惩罚参数和高斯核参数对回归模型的预测精度有较大影响，过大或过小都会降低模型的学习推广能力，本文应用PSO算法实现了对两者的优化。

2) PSO-SVR模型的回归预测精度优于传统的PSO-BP神经网络模型，具有更好的泛化性能，在一定范围内能够准确反映VSV调节规律。

3) PSO-SVR模型能够实现VSV位置监控和VSV系统故障的初步诊断，为发动机状态监控和故障诊断提供帮助。

 [1] 唐庆如, 孔萌. CFM56-7B发动机VSV结构损伤分析[J]. 航空维修与工程, 2011 (4): 31–33. TANG Q R, KONG M. Analysis of CFM56-7B VSV structural damage[J]. Aviation Maintenance Engineering, 2011 (4): 31–33. (in Chinese) [2] 李世林. VSV系统对CFM56发动机喘振的影响分析[J]. 科学技术与工程, 2011, 11 (20): 4934–4936. LI S L. Research on VSV faults based CFM56 engine surge[J]. Science Technology and Engineering, 2011, 11 (20): 4934–4936. DOI:10.3969/j.issn.1671-1815.2011.20.063 (in Chinese) [3] 黄爱华. 涡扇发动机可调静子叶片控制规律研究[J]. 燃气涡轮试验与研究, 2017, 30 (1): 48–51. HUANG A H. Control law of variable stator vane for turbofan engine[J]. Gas Turbine Experiment and Research, 2017, 30 (1): 48–51. (in Chinese) [4] 吴秀宽, 林森. 某涡扇发动机风扇进口可调导流叶片调节规律分析[C]//第五届中国航空学会青年科技论坛文集(第5集). 北京: 北京航空航天大学出版社, 2012: 227-231. WU X K, LIN S. The analyse about the control law of turbofan'IGV[C]//Proceedings of the Fifth China Aviation Society Youth Science and Technology Forum (Fifth Episodes). Beijing: Beihang University Press, 2012: 227-231(in Chinese). [5] 曹志鹏, 刘波, 丁伟. 静叶角度调节对组合压气机性能优化机理[J]. 北京航空航天大学学报, 2007, 33 (8): 878–881. CAO Z P, LIU B, DING W. Stator setting angles adjustment on performance improvement of axial-centrifugal compressor[J]. Journal of Beijing University of Aeronautics and Astronautics, 2007, 33 (8): 878–881. (in Chinese) [6] 张健, 任铭林. 静叶角度调节对压气机性能影响的试验研究[J]. 航空动力学报, 2000, 15 (1): 27–30. ZHANG J, REN M L. Experimental investigation on effect of stator vane angle adjustment on compressor performance[J]. Journal of Aerospace Power, 2000, 15 (1): 27–30. (in Chinese) [7] 张宇飞, 么子云, 唐松林, 等. 一种基于主成分分析和支持向量机的发动机故障诊断方法[J]. 中国机械工程, 2016, 27 (24): 3307–3311. ZHANG Y F, YAO Z Y, TANG S L, et al. An engine fault diagnosis method based on PCL and SVR[J]. China Mechanical Engineering, 2016, 27 (24): 3307–3311. DOI:10.3969/j.issn.1004-132X.2016.24.008 (in Chinese) [8] BI F R, LIU Y P. Fault diagnosis of valve clearance in diesel engine based on BP neural network and support vector machine[J]. Transactions of Tianjin University, 2016, 22 (6): 536–543. DOI:10.1007/s12209-016-2675-1 [9] 殷锴, 钟诗胜, 那媛, 等. 基于BP神经网络的航空发动机故障检测技术研究[J]. 航空发动机, 2017, 43 (1): 53–57. YIN K, ZHONG S S, NA Y, et al. Research on aeroengine fault detection technology based on BP neural network[J]. Aeroengine, 2017, 43 (1): 53–57. (in Chinese) [10] 栾圣罡. 基于气路参数样本的航空发动机状态监视方法与系统研究[D]. 哈尔滨: 哈尔滨工业大学, 2008. LUAN S G. Aeroengine condition monitoring technique and system based on gas path parameter sample[D]. Harbin: Harbin Institute of Technology, 2008(in Chinese). [11] 刘永建. 基于改进神经网络的民机发动机故障诊断与性能预测研究[D]. 南京: 南京航空航天大学, 2012. LIU Y J. Research on modified neural network for fault diagnosis and performance prediction of aeroengine[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2012(in Chinese). [12] 王芳. 基于支持向量机的沪深300指数回归预测[D]. 济南: 山东大学, 2015. WANG F. CSI 300 index regression prediction based on support vector machine[D]. Jinan: Shandong University, 2015(in Chinese). [13] 史峰, 王辉, 郁磊, 等. 智能算法30个案例分析[M]. 北京: 北京航空航天大学出版社, 2011. SHI F, WANG H, YU L, et al. 30 cases analysis of intelligent algorithm[M]. Beijing: Beihang University Press, 2011. (in Chinese) [14] 崔智全. 民航发动机气路参数偏差值挖掘方法及其应用研究[D]. 哈尔滨: 哈尔滨工业大学, 2013. CUI Z Q. Civil aeroengine gas path parameter deviation mining method with application[D]. Harbin: Harbin Institute of Technology, 2013(in Chinese). [15] Boeing. 777 aircraft maintenance manual[Z]. Chicago: Boeing, 2015. [16] 彭泽琰, 刘刚. 航空燃气轮机原理[M]. 北京: 国防工业出版社, 2000. PENG Z Y, LIU G. Principles of aviation gas turbines[M]. Beijing: National Defense Industry Press, 2000. (in Chinese) [17] Pratt & Whitney Company. ECMⅡtraining manual[Z]. Hartford: Pratt & Whitney Company, 1994. [18] 周百政, 曹惠玲. 基于EHM软件思路的QAR数据处理[J]. 航空维修与工程, 2010 (4): 60–62. ZHOU B Z, CAO H L. QAR data processing based on the method of EHM software[J]. Aviation Maintenance & Engineering, 2010 (4): 60–62. (in Chinese)

#### 文章信息

CAO Huiling, KAN Yuxiang, XUE Peng

Exploration of engine VSV regulation law using support vector regression

Journal of Beijing University of Aeronautics and Astronsutics, 2018, 44(7): 1371-1377
http://dx.doi.org/10.13700/j.bh.1001-5965.2017.0523