AI-based real-time noise reduction of flow field pressure signals under plasma electromagnetic interference
-
摘要: 面向闭环主动流动控制的可靠传感需求,提出采用人工神经网络模型对不同等离子体电磁干扰下受污染流场信号进行实时降噪处理。以安装在圆柱表面的动态压力传感器为研究对象,分别采集了正弦交流介质阻挡放电(AC–DBD)的“稠密峰”型干扰信号和纳秒脉冲介质阻挡放电(NS–DBD)的“稀疏突刺”型干扰信号,对人工合成含干扰的流场压力数据进行了监督学习训练,并对模型的泛化能力进行了测试验证。结果表明:基于人工神经网络模型的实时降噪方法可以有效抑制等离子体激励带来的电磁干扰影响,还原真实的压力信号,且对AC–DBD“稠密峰”型干扰信号的降噪效果更好,降噪后信号波形平滑,与真实信号拟合度高;将模型用于真实的流场压力测量中,通过对比降噪后信号与真实信号均值,验证了降噪神经网络的信号还原精度。Abstract: For the reliable sensing requirements of closed-loop active flow control, a real-time noise reduction method based on the artificial neural network was proposed for solving the plasma actuation electromagnetic interference on flow field signals. Taking the dynamic pressure sensor installed on the cylinder surface as the experimental subject, the “dense peak” type noise signals of alternating current dielectric barrier discharge (AC–DBD) and the “sparse spike” type noise signals of nanosecond pulsed dielectric barrier discharge (NS–DBD) were collected respectively. Artificial synthetic noise signals were used for supervised learning, and the generalization of the artificial neural network model was tested and verified. The results show that this method can effectively suppress the influence of electromagnetic interference caused by plasma actuation and restore the real pressure signal. It has better denoising performance on the AC–DBD “dense peak” type noise signal. The denoised signal is smoother and better fitted with the real one. This model is also applied to the real flow field pressure measurement, and the accuracy of the denoising network prediction is further verified by comparing the mean value of the denoised signal and the real signal.
-
-
表 1 人工神经网络模型参数
Table 1 Parameters of artificial neural network model
网络模型参数 设置值 输入层节点数 402(NS–DBD),403(AC–DBD) 隐含层数 4 隐含层节点数 512,1024,2048,1024 输出层节点数 400 激活函数 Leaky ReLU 优化器 Adam 损失函数 均方根误差(RMSE) 初始学习率 0.001 学习率衰减系数 0.96 表 2 AC–DBD模型训练电参数
Table 2 Electrical parameters of training data for AC–DBD model
电压Up/kV 载波频率fbase/kHz 调制频率fm/Hz 4,6,8,10,12 6,8,10,12 100,300,500,700,900 表 3 NS–DBD模型训练电参数
Table 3 Electrical parameters of training data for NS–DBD model
电压Up/kV 脉冲频率fp/Hz 4,6,8,10,12 200,400,600,800,1000,1200,1400 表 4 AC–DBD模型测试电参数
Table 4 Electrical parameters of testing data for AC–DBD model
电压Up/kV 载波频率fbase/kHz 调制频率fm/Hz 5,7,9,11 5,7,9,11 200,400,600,800,1000 表 5 不同电参数下AC–DBD模型降噪前后均方根误差
Table 5 RMSE value of AC–DBD model under different testing electrical parameters
激励电参数 RMSE1 RMSE2 Up = 5 kV,fbase = 7 kHz,fm = 200 Hz 32.86 3.73 Up = 5 kV,fbase = 9 kHz,fm = 400 Hz 39.56 3.27 Up = 5 kV,fbase = 11 kHz,fm = 600 Hz 42.22 3.66 Up = 7 kV,fbase = 7 kHz,fm = 400 Hz 40.23 2.89 Up = 7 kV,fbase = 9 kHz,fm = 600 Hz 44.10 3.12 Up = 7 kV,fbase = 11 kHz,fm = 800 Hz 48.36 3.24 Up = 9 kV,fbase = 7 kHz,fm = 600 Hz 48.30 3.36 Up = 9 kV,fbase = 9 kHz,fm = 800 Hz 56.23 3.78 Up = 9 kV,fbase = 11 kHz,fm = 1000 Hz 62.35 3.82 Up = 11 kV,fbase = 7 kHz,fm = 800 Hz 66.23 2.99 Up = 11 kV,fbase = 9 kHz,fm = 1000 Hz 70.51 3.64 Up = 11 kV,fbase = 11 kHz,fm = 200 Hz 66.16 2.81 表 6 NS–DBD模型测试电参数
Table 6 Electrical parameters of testing data for NS–DBD model
电压Up/kV 脉冲频率fp/Hz 5,7,9,11 300,500,700,900,1100 表 7 不同电参数下NS–DBD模型降噪前后均方根误差
Table 7 RMSE value of NS–DBD model under different testing electrical parameters
激励电参数 RMSE1 RMSE2 Up = 5 kV,fp = 300 Hz 40.36 4.36 Up = 5 kV,fp = 500 Hz 51.63 4.89 Up = 5 kV,fp = 700 Hz 62.22 4.78 Up = 7 kV,fp = 500 Hz 50.24 4.86 Up = 7 kV,fp = 700 Hz 59.33 4.88 Up = 7 kV,fp = 900 Hz 68.36 4.98 Up = 9 kV,fp = 700 Hz 66.32 4.48 Up = 9 kV,fp = 900 Hz 72.56 4.36 Up = 9 kV,fp = 1100 Hz 79.85 4.90 Up = 11 kV,fp = 300 Hz 68.49 4.22 Up = 11 kV,fp = 900 Hz 90.22 4.23 Up = 11 kV,fp = 1100 Hz 105.62 4.58 表 8 降噪后的压力信号均值与干净压力信号均值对比
Table 8 Mean value comparison between the denoised pressure signal and the clean pressure signal
激励电参数 降噪后信号与
真实信号的
均值差/PaAC–DBD Up = 5 kV,fbase = 6 kHz,fm = 200 Hz 1.26 Up = 5 kV,fbase = 8 kHz,fm = 400 Hz 1.15 Up = 7 kV,fbase = 6 kHz,fm = 400 Hz 1.36 Up = 7 kV,fbase = 8 kHz,fm = 600 Hz 1.44 Up = 9 kV,fbase = 6 kHz,fm = 600 Hz 1.52 Up = 9 kV,fbase = 8 kHz,fm = 800 Hz 1.21 Up = 11 kV,fbase = 6 kHz, fm = 800 Hz 1.58 Up = 11 kV,fbase = 8 kHz, fm = 1000 Hz 1.61 NS–DBD Up = 6 kV,fp = 200 Hz 2.98 Up = 7 kV,fp = 300 Hz 2.32 Up = 8 kV,fp = 400 Hz 2.36 Up = 9 kV,fp = 500 Hz 2.22 Up = 10 kV,fp = 600 Hz 2.43 Up = 11 kV,fp = 700 Hz 2.58 Up = 13 kV,fp = 800 Hz 2.46 Up = 14 kV,fp = 900 Hz 2.77 -
[1] 李应红, 吴云. 等离子体激励调控流动与燃烧的研究进展与展望[J]. 中国科学(技术科学), 2020, 50(10): 1252–1273. DOI: 10.1360/SST-2020-0111 LI Y H, WU Y. Research progress and outlook of flow control and combustion control using plasma actuation[J]. Scientia Sinica (Technologica), 2020, 50(10): 1252–1273. doi: 10.1360/SST-2020-0111
[2] 赵志杰, 罗振兵, 刘杰夫, 等. 基于分布式合成双射流的飞行器无舵面三轴姿态控制飞行试验[J]. 力学学报, 2022, 54(5): 1220–1228. DOI: 10.6052/0459-1879-21-586 ZHAO Z J, LUO Z B, LIU J F, et al. Flight test of aircraft three-axis attitude control without rudders based on distributed dual synthetic jets[J]. Chinese Journal of Theoretical and Applied Mechanics, 2022, 54(5): 1220–1228. doi: 10.6052/0459-1879-21-586
[3] 姚张奕, 史志伟, 董益章. 深度强化学习在翼型分离流动控制中的应用[J]. 实验流体力学, 2022, 36(3): 55–64. DOI: 10.11729/syltlx20210085 YAO Z Y, SHI Z W, DONG Y Z. Deep reinforcement learning for the control of airfoil flow separation[J]. Journal of Experiments in Fluid Mechanics, 2022, 36(3): 55–64. doi: 10.11729/syltlx20210085
[4] 兰子奇, 史志伟, 孙琪杰, 等. AC–DBD等离子体激励对L形截面钝体风荷载减阻的实验研究[J]. 实验流体力学, 2021, 35(2): 83–91. DOI: 10.11729/syltlx20200095 LAN Z Q, SHI Z W, SUN Q J, et al. Experimental study on drag reduction of L-shaped bluff body by AC–DBD plasma actuation[J]. Journal of Experiments in Fluid Mechanics, 2021, 35(2): 83–91. doi: 10.11729/syltlx20200095
[5] CHENG X Q, WONG C W, HUSSAIN F, et al. Flat plate drag reduction using plasma-generated streamwise vortices[J]. Journal of Fluid Mechanics, 2021, 918: A24. doi: 10.1017/jfm.2021.311
[6] DUONG A H, CORKE T C, THOMAS F O. Characteristics of drag-reduced turbulent boundary layers with pulsed-direct-current plasma actuation[J]. Journal of Fluid Mechanics, 2021, 915: A113. doi: 10.1017/jfm.2021.167
[7] LI Z, SHI Z W, DU H, et al. Analysis of flow separation control using nanosecond-pulse discharge plasma actuators on a flying wing[J]. Plasma Science and Technology, 2018, 20(11): 116–125.
[8] 周岩, 罗振兵, 王林, 等. 等离子体合成射流激励器及其流动控制技术研究进展[J]. 航空学报, 2022, 43(3): 90–132. DOI: 10.7498/aps.68.20190683 ZHOU Y, LUO Z B, WANG L, et al. Plasma synthetic jet actuator for flow control: review[J]. Acta Aeronautica et Astronautica Sinica, 2022, 43(3): 90–132. doi: 10.7498/aps.68.20190683
[9] WU B, GAO C, LIU F, et al. Reduction of turbulent boundary layer drag through dielectric-barrier-discharge plasma actuation based on the Spalding formula[J]. Plasma Science and Technology, 2019, 21(4): 107–114.
[10] 金元中, 郑博睿, 喻明浩, 等. 滑动放电等离子体控制细长体头部背风区非对称涡实验研究[J]. 实验流体力学, 2022, 36(5): 43–51. DOI: 10.11729/syltlx20210101 JIN Y Z, ZHENG B R, YU M H, et al. Experimental study on flow control of asymmetric vortex over the leeward region of the head of the slender body by sliding discharge plasma actuation[J]. Journal of Experiments in Fluid Mechanics, 2022, 36(5): 43–51. doi: 10.11729/syltlx20210101
[11] 张卫国, 史喆羽, 李国强, 等. 风力机翼型动态失速等离子体流动控制数值研究[J]. 力学学报, 2020, 52(6): 1678–1689. DOI: 10.6052/0459-1879-20-090 ZHANG W G, SHI Z Y, LI G Q, et al. Numerical study on dynamic stall flow control for wind turbine airfoil using plasma actuator[J]. Chinese Journal of Theoretical and Applied Mechanics, 2020, 52(6): 1678–1689. doi: 10.6052/0459-1879-20-090
[12] 黄广靖, 戴玉婷, 杨超. 低雷诺数俯仰振荡翼型等离子体流动控制[J]. 力学学报, 2021, 53(1): 136–155. DOI: 10.6052/0459-1879-20-183 HUANG G J, DAI Y T, YANG C. Plasma-based flow control on pitching airfoil at low Reynolds number[J]. Chinese Journal of Theoretical and Applied Mechanics, 2021, 53(1): 136–155. doi: 10.6052/0459-1879-20-183
[13] 梁华, 李应红, 程邦勤, 等. 等离子体气动激励抑制翼型失速分离的仿真研究[J]. 航空动力学报, 2008, 23(5): 777–783. DOI: 10.13224/j.cnki.jasp.2008.05.004 LIANG H, LI Y H, CHENG B Q, et al. Numerical simulation on airfoil stall separation suppression by plasma aerodynamic actuation[J]. Journal of Aerospace Power, 2008, 23(5): 777–783. doi: 10.13224/j.cnki.jasp.2008.05.004
[14] 杜海, 史志伟, 程克明, 等. 纳秒脉冲等离子体分离流控制频率优化及涡运动过程分析[J]. 航空学报, 2016, 37(7): 2102–2111. DU H, SHI Z W, CHENG K M, et al. Frequency optimization and vortex dynamic process analysis of separated flow control by nanosecond pulsed plasma discharge[J]. Acta Aeronautica et Astronautica Sinica, 2016, 37(7): 2102–2111.
[15] 郝琳召, 张彬乾, 陈真利. 纳秒等离子体激励控制翼型流动分离机理研究[J]. 航空工程进展, 2014, 5(1): 25–32. DOI: 10.3969/j.issn.1674-8190.2014.01.005 HAO L Z, ZHANG B Q, CHEN Z L. Investigation on mechanisms of separation control over an airfoil using nanosecond pulsed plasma actuator[J]. Advances in Aeronautical Science and Engineering, 2014, 5(1): 25–32. doi: 10.3969/j.issn.1674-8190.2014.01.005
[16] MENG X S, LONG Y X, WANG J L, et al. Dynamics and control of the vortex flow behind a slender conical forebody by a pair of plasma actuators[J]. Physics of Fluids, 2018, 30(2): 024101. doi: 10.1063/1.5005514
[17] 孟宣市, 惠伟伟, 易贤, 等. AC–SDBD等离子体激励防/除冰研究现状与展望[J]. 空气动力学学报, 2022, 40(2): 31–49. DOI: 10.7638/kqdlxxb-2021.0159 MENG X S, HUI W W, YI X, et al. Anti-/ De-icing by AC–SDBD plasma actuators: status and outlook[J]. Acta Aerodynamica Sinica, 2022, 40(2): 31–49. doi: 10.7638/kqdlxxb-2021.0159
[18] CAI J S, TIAN Y Q, MENG X S, et al. An experimental study of icing control using DBD plasma actuator[J]. Experiments in Fluids, 2017, 58(8): 102. doi: 10.1007/s00348-017-2378-y
[19] WEI B, WU Y, LIANG H, et al. SDBD based plasma anti-icing: A stream-wise plasma heat knife configuration and criteria energy analysis[J]. International Journal of Heat and Mass Transfer, 2019, 138: 163–172. doi: 10.1016/j.ijheatmasstransfer.2019.04.051
[20] LATTARI F, GONZALEZ LEON B, ASARO F, et al. Deep learning for SAR image despeckling[J]. Remote Sensing, 2019, 11(13): 1532. doi: 10.3390/rs11131532
[21] ZHU W Q, MOUSAVI S M, BEROZA G C. Seismic signal denoising and decomposition using deep neural networks[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(11): 9476–9488. doi: 10.1109/TGRS.2019.2926772
[22] TIAN C W, FEI L K, ZHENG W X, et al. Deep learning on image denoising: an overview[J]. Neural Networks, 2020, 131: 251–275. doi: 10.1016/j.neunet.2020.07.025
[23] GOYAL B, DOGRA A, AGRAWAL S, et al. Image denoising review: from classical to state-of-the-art approaches[J]. Information Fusion, 2020, 55: 220–244. doi: 10.1016/j.inffus.2019.09.003
[24] ANWAR S, BARNES N. Real image denoising with feature attention[C]//Proc of the 2019 IEEE/CVF International Conference on Computer Vision (ICCV). IEEE, 2020: 3155-3164.