張杰 史治宇
摘要: 提出了同步壓縮小波時頻脊提取結(jié)合自適應時域濾波的時變系統(tǒng)參數(shù)識別方法。同步壓縮小波相比傳統(tǒng)小波具有優(yōu)異的時頻分辨率,基于該小波時頻脊提取可以獲得時變結(jié)構(gòu)的瞬時模態(tài)頻率,在此基礎上可構(gòu)造各階分量信號的載波矩陣,并應用自適應時域濾波求解分量信號的幅值包絡,進而識別結(jié)構(gòu)的阻尼比。該方法能對時變系統(tǒng)結(jié)構(gòu)響應進行各階分解,相比經(jīng)驗模態(tài)分解方法具有優(yōu)異的時頻提取能力、較強的抗噪性能和識別復雜時變問題的能力。在理論推導基礎上,首先通過一個3自由度時變仿真算例驗證了方法的正確性和抗噪性,再應用該算例構(gòu)造了一個復雜時變算例(分量信號在頻域重疊且突變),以此驗證方法對各類復雜時變情況的適用性和準確性。
關(guān)鍵詞: 參數(shù)識別; 同步壓縮小波; 時頻脊提取; 自適應濾波; 時變系統(tǒng)
中圖分類號: TB123; O327? 文獻標志碼: A? 文章編號: 1004-4523(2019)03-0462-09
DOI:10.16385/j.cnki.issn.1004-4523.2019.03.011
引 言
現(xiàn)代工程結(jié)構(gòu)逐漸邁向大型化、智能化、微型化的發(fā)展方向,結(jié)構(gòu)時變問題隨之凸顯,例如:運載火箭發(fā)射時質(zhì)量減小、高超聲速飛行器氣動加熱引起結(jié)構(gòu)剛度變化、空間太陽能板伸展引起結(jié)構(gòu)時變等,因此結(jié)構(gòu)時變問題研究無論在理論層面還是在實際應用上都有重大研究價值。
國內(nèi)外現(xiàn)有線性時變系統(tǒng)參數(shù)識別的研究思路主要分兩類[1]:第一類是基于短時時不變假設,對信號進行局部處理,應用信號處理方法(Gabor變換、Wigner-Ville分布[2]、小波變換[3-7])或子空間(Subspace)[8-10]進行參數(shù)識別;第二類是先將響應信號作為整體進行自適應分解,再對各階響應信號進行參數(shù)識別[11-14]。第一類方法,短時時不變假設要求時域信號分析區(qū)間選取足夠小但同時增加了計算量,因此算法都存在計算效率低、抗噪性能差、實用性不足的問題。第二類方法中較有代表性的是希爾伯特黃變換(HHT)[11],HHT方法通過經(jīng)驗模態(tài)分解可以自適應、高效地將信號分解為多個本征模態(tài)函數(shù),進一步通過對每一個分量做希爾伯特變換,可在時頻域得到良好的信號表示。然而,該方法難以處理多分量頻率重疊或交叉信號,且存在模式混淆、“邊界效應”等問題。
由于時變系統(tǒng)產(chǎn)生非平穩(wěn)的振動響應,振動信號具有明顯的時變調(diào)制特性,因此時頻分析工具是一種較為理想的研究方法,常用時頻分析工具都是由傅里葉變換發(fā)展而來,都存在上文所述的精度低、自適應不夠、抗噪性能差等問題。
近年來,Daubechie等[15]提出了壓縮小波變換(Synchrosqueezed Wavelet Transform, SWT)。該算法基于連續(xù)小波變換,通過對小波變換的復數(shù)譜沿頻率軸方向壓縮重排,具有較高的頻率分辨率,能得到頻率曲線更加集中的時頻表達。同時自適應濾波在最小化結(jié)構(gòu)和數(shù)據(jù)誤差基礎上,可有效分離在時頻域內(nèi)鄰近甚至交叉的信號分量[16],自適應濾波在非平穩(wěn)信號處理中尤其是故障診斷分析中具有廣泛應用[17-18],然而時域濾波須獲得待分離信號分量的瞬時頻率載波矩陣。因此本文提出了將同步壓縮小波時頻脊提取引入到自適應濾波中,得到改進的自適應時頻分解方法,并將其應用到時變系統(tǒng)參數(shù)識別中。該方法首先應用同步壓縮小波時頻脊提取脈沖響應信號的時頻分布,根據(jù)其時頻結(jié)果構(gòu)造載波矩陣再應用時域濾波將響應信號自適應分解成多階響應信號,最后再對各階響應進行瞬時模態(tài)參數(shù)的識別。改進算法可有效提高識別的精度并增加其適用范圍。
首先計算響應的壓縮小波時頻脊,結(jié)果如圖7所示,可以看出結(jié)構(gòu)的第2階和第3階分量在5-8 Hz存在重疊,且在20 s時頻率發(fā)生突變。分別用本文方法和EMD方法對加速度響應進行分解,得到各個分量的時域信號,再對分量信號作短時傅里葉分析(STFT)來驗證分解信號的頻域成分。本文方法分解結(jié)果如圖8所示,將結(jié)果和圖7比較,可以看出3個分量的STFT結(jié)果和理論值是吻合的。EMD分解結(jié)果如圖9所示,由于EMD存在較多虛假成分,因此按頻率由低到高取能量較大的前4階成分。將分解結(jié)果和圖7比較,可以發(fā)現(xiàn)EMD分解在無改進情況下較難處理此種時變問題,第1分量和第4分量頻率成分較混亂,第2分量含較多突變前頻率成分,第3分量則含較多突變后頻率成分,因此傳統(tǒng)的EMD方法較難處理多分量信號重頻或交叉的情況。
然后根據(jù)本文方法分解結(jié)果進行瞬時頻率的識別,識別結(jié)果如圖10所示。3階頻率識別結(jié)果都能較好地貼近理論值,僅在數(shù)據(jù)兩端局部會誤差稍大,并且方法對突變點也能較好追蹤,總體來看本文方法相比EMD方法更適用在復雜時變情況的參數(shù)識別。同樣地對信號添加不同信噪比的噪聲考察方法的抗噪性,頻率識別誤差如表3所示,可以看出復雜時變情況下參數(shù)識別也幾乎不受噪聲影響。
5 結(jié) 論
(1) 本文提出了同步壓縮小波時頻脊提取結(jié)合自適應時域濾波的時變系統(tǒng)參數(shù)識別方法,該方法基于加速度響應信號進行數(shù)據(jù)的整體分解,具有較高的自適應性和實用性。
(2) 該方法利用壓縮小波較高的時頻分辨率能精確提取響應信號的各階瞬時頻率。
(3) 由于自適應時域濾波適用于非平穩(wěn)信號處理,因此改進時頻分解方法能得到精確的時變響應幅值包絡,基于包絡結(jié)果進行的阻尼識別也更準確,同時算法具有很好的抗噪性能,識別結(jié)果幾乎不受噪聲影響,抗噪性能優(yōu)異。
(4) 相比EMD方法,本文方法適用范圍更廣,能應用在各類復雜時變結(jié)構(gòu)中,并且可以追蹤瞬時頻率的各類變化(線性、周期和突變等)。
參考文獻:
[1] 于開平, 龐世偉, 趙 婕. 時變線性/非線性結(jié)構(gòu)參數(shù)識別及系統(tǒng)辨識方法研究進展[J]. 科學通報, 2009, 54(20): 3147-3156.
Yu Kaiping, Pang Shiwei, Zhao Jie. Advances in method of time-varying linear/nonlinear structural system identification and parameter estimate[J]. Chinese Science Bulletion (Chinese Version), 2009, 54(20): 3147-3156.
[2] 續(xù)秀忠, 華宏星, 張志誼, 等. 應用時頻表示進行結(jié)構(gòu)時變模態(tài)參數(shù)辨識[J]. 振動與沖擊, 2002, 21(2): 36-40.
Xu Xiuzhong, Hua Hongxing, Zhang Zhiyi,et al. Time-varying modal frequency identification by using time-frequency representation[J]. Journal of Vibration and Shock, 2002, 21(2): 36-40.
[3] Ghanem R, Romeo F. A wavelet-based approach for the identification of linear time-varying dynamic systems[J]. Journal of Sound and Vibration, 2000, 234(4): 555-576.
[4] 鄒甲軍, 馮志化, 陸維生. 基于小波的LTV系統(tǒng)的參數(shù)識別[J]. 蘇州大學學報(工科版), 2005, 25: 41-46.
Zou Jiajun, Feng Zhihua, Lu Weisheng. Model parameter identification of LTV systems based on Daubechies wavelet[J]. Journal of Soochow University (Engineering Science Edition), 2005, 25: 41-46.
[5] Xu X, Shi Z Y, You Q. Identification of linear time-varying systems using a wavelet-based state-space method[J]. Mechanical Systems and Signal Processing, 2012, 26: 91-103.
[6] Hera A, Shinde A, Hou Z K. Issues in tracking instantaneous modal parameters for structural health monitoring using wavelet approach[C]. Proceedings of 23rd International Modal Analysis Conference (IMAC XXIII), Orlando, Florida, USA, 2005: 338-347.
[7] Dziedziech K, Staszewski W J, Basu B, et al. Wavelet-based detection of abrupt changes in natural frequencies of time-variant systems[J]. Mechanical Systems and Signal Processing, 2015, 64-65:347-359.
[8] Liu K. Identification of linear time-varying systems[J]. Journal of Sound and Vibration, 1997, 204: 487-500.
[9] 龐世偉, 于開平, 鄒經(jīng)湘. 識別時變結(jié)構(gòu)模態(tài)參數(shù)的改進子空間方法[J]. 應用力學學報, 2005, 22: 184-188.
Pang Shiwei, Yu Kaiping, Zou Jingxiang. Improved subspace method with application in linear time-varying structural modal parameter identification[J]. Chinese Journal of Applied Mechanics, 2005, 22: 184-188.
[10] Tasker F, Bosse A, Fisher S. Real-time modal parameter estimation using subspace methods: Theory[J]. Mechanical Systems and Signal Processing, 1998, 12: 797-808.
[11] Feldman M. Nonlinear system vibration analysis using Hilbert transform II: Forced vibration analysis method[J]. Mechanical Systems and Signal Processing, 1994, 8(3): 309-318.
[12] Shi Z Y, Law S S. Identification of linear time-varying dynamical systems using Hilbert transform and EMD[J]. Journal of Applied Mechanics, 2007, 74(2): 223-230.
[13] Bao C X, Hao H, Li Z X, et al. Time-varying system identification using a newly improved HHT algorithm[J]. Computers and Structures, 2009, 87:1611-1623.
[14] 程軍圣, 張 亢, 楊 宇, 等. 局部均值分解與經(jīng)驗模式分解的對比研究[J]. 振動與沖擊, 2009, 28(5):13-16.
Cheng Junsheng, Zhang Kang, Yang Yu, et al. Comparison between the methods of local mean decomposition and empirical mode decomposition[J]. Journal of Vibration and Shock, 2009, 28(5):13-16.
[15] Daubechies I, Lu F J, Wu H T. Synchrosqueezed wavelet transforms: An empirical mode decomposition like tool[J]. Applied and Computational Harmonic Analysis, 2011, 30:243-261.
[16] Vold H,Herlufsen H,Mains M.Multi axle order tracking with the Vold-Kalman tracking filter[J].Journal of Sound and Vibration, 1997, 13(5):30-34.
[17] 秦嗣峰,馮志鵬,Liang Ming. Vold-Kalman濾波和高階能量分離在時變工況行星齒輪箱故障診斷中的應用研究[J]. 振動工程學報,2015,28(5):839-845.
Qin Sifeng, Feng Zhipeng, Liang Ming. Application of Vold-Kalman filter and higher order energy separation to fault diagnosis of planetary gearbox under time-varying conditions[J]. Journal of Vibration Engineering, 2015, 28(5):839-845.
[18] 馮 珂,王科盛,宋理偉,等. 基于階次譜的Vold-Kalman濾波帶寬優(yōu)選方法[J]. 振動工程學報,2017,30(02):319-324.
Feng Ke, Wang Kesheng, Song Liwei, et al. An order spectrum based selection method to Vold-Kalman filter bandwidth[J]. Journal of Vibration Engineering, 2017, 30(02):319-324.
[19] Vold H, Mains M, Blough J. Theoretical foundations for high performance order tracking with the Vold-Kalman tracking filter[C]. SAE Noise and Vibration Conference and Exposition, 1997:1083-1088.
[20] Feldauer Ch, Hldrich R. Realisation of a Vold-Kalman tracking filter——A least square problem[C]. Proceedings of the COST G-6 Conference on Digital Audio Effects (DAFX-410800), Verona, Italy, 2000: 7-9.
[21] Tuma J. Setting the pass bandwidth in the Vold-Kalman order tracking filter[C]. Twelfth International Congress on Sound and Vibration, Lisbon, 2005, Paper 719.
Abstract: In this paper, an instantaneous modal parameter identification method for time-varying structures based on synchrosqueezed wavelet transform (SWT) and adaptive filtering is proposed. The time-frequency ridges of SWT are applied to the instantaneous frequency extraction of time-varying structures for their excellent time-frequency resolution. Then the carrier matrix of each order component signal can be constructed and the component signal amplitude envelope is calculated based on the adaptive filter. On this basis, the structural damping ratio is identified. Compared with the empirical mode decomposition, this method has excellent time-frequency extraction capability, strong noise immunity and strong applicability for various time-varying conditions. Based on the theory, the results of a three-degree-of-freedom time-varying simulation verify the correctness and anti-noise ability of the method. The example is also used to construct a complex time-varying situation in order to verify the applicability of the method, and the results show that this method can be applied to time-varying situations that component signals overlap or even intersect in frequency domain.
Key words: parameter identification; synchrosqueezed wavelet transform; time-frequency ridges extraction; adaptive filtering; time-varying system
作者簡介: 張 杰(1988-),男,博士研究生。電話:18551670428;E-mail:jzhang1988@nuaa.edu.cn
通訊作者: 史治宇(1967-),男,博士,教授,博士生導師。電話:18914704215;E-mail:zyshi@nuaa.edu.cnZ ··y^