您好,欢迎来到五一七教育网。
搜索
您的当前位置:首页K-L分解在非平稳地震响应分析中的应用

K-L分解在非平稳地震响应分析中的应用

来源:五一七教育网
第1O卷第23期2010年8月 科学技术与工程 Vo1.10 No.23 Aug.2010 1671—1815(2010)23.5623-08 Science Technology and Engineering ⑥2010 Sci.Tech.Engng. 力 学 K-L分解在非平稳地震响应分析中的应用 徐瑞 (华南理工大学土木与交通学院,广州510640) 摘要介绍了一种非平稳地震激励下结构随机响应的分析方法。按照K—L分解理论,地震激励可以描述为一系列确定性 函数与随机系数乘积的线性组合形式。对于线性结构,地震激励的每项K—L展开级数对应于一项结构响应。因此,任何确定 性动力时程积分方法可以应用于求解与KL展开级数相对应的结构响应。这为商用有限元软件用于随机地震响应分析打开 了方便之门,同时也有利于提高随机振动分析在实际工程中的应用。一个受均匀调制非平稳地震激励下的2l层框架结构用 于阐明这种方法。 关键词地震 随机振动 K—L展开 非平稳随机激励 中图法分类号0324; 文献标志码A Earthquake is one of the most serious natural dis— practical to describe seismic ground motion as random asters to human beings.In the past three years,cata- process.Seismic ground motion is a type of non—sta- strophic earthquake has struck our country for two tionary random process,which is characterized by a times,one is the Wenchuan earthquake with magnitude building up portion in the beginning and a slowly deca— of 8.0 which occurred in the Sichuan Province on 12th ying tail portion at the end with strong—motion portions May 2008,and the other is Yushu earthquake with in between.Fig.1 shows a typical record of ground magnitude of 7.1 which occurred in the Gansu Prov- motion acceleration occurred in Wenchuan earthquake. ince only 3 weeks ago.There were great amounts of people lost lives due to the structural damage and col- lapse under those strong earthquakes.Although todayg state of the art develops rapidly,the occurrence time, site and intensity of ea ̄hquake are still unable to pre- dict.Therefore,efficient method for structural seismic response analysis should be developed to avoid earth- 萝 quake’S damage probably occurred in future to build. ings,bridges,dams,et.a1. Fig.1 A typical record of ground acceleration Due to the strong randomness of earthquake in occurred in Wenchuan earthquake terms of occurrence time,site and intensity,it is more A structure subjected to non・-stationary seismic ex・- citation produces non-stationary random response.So, 2009年5月20日收到 the analysis of a sturcture subjected to non—stationary 作者简介:徐瑞(198l一),男,博士生,研究方向:结构随机分析 seismic excitation is a non—stationary random vibration 与动力可靠度分析、大跨度桥梁抗风与抗震等。E—mail:xxurui@ 163.con。 problem.The various methods available to analyze lin— 5624 科学技术与工程 10卷 ear structure system subjected to random seismic exci. are sometimes treated as random processes.The Kar- tation can be distinguished into method operating in the frequency domain[ 一 ]and in the time domain[4--1o ̄. respectively.Those method rationally quantity the ran— domness of structural response in terms of the first two moment characteristics,i.e.the mean,covariance and/or the power spectral density function,respective- ly.Solutions for non—stationary seismic excitation,par— ticularly in the frequency domain,are rather involved and even intractable for larger structure systems. Direct integration procedures[ 一 。]operating in the time domain for the random response computation,ap— peared since the development of the finite element methods.However,all these procedures cannot be ap— plied efifciently for sturctures with thousands degree—of- rfeedom,because all of these procedures require the direct manipulation of the covafiance matirx with a full symmetric matirx of dimension≥2n.where n denotes the number of degree-of-freedom. In this paper,a method for calculating the linear structure subjected to non—stationary seismic excitation is introduced.In particular,the first two moment char- acteristics can be calculated efficiently based on K—L expansion used to represent the seismic excitation.For the case where the seismic excitation is described by a ifnite set of deterministic K—L terms,where to each de— tcrministic term of the response corresponding to deter— ministic K..L term of seismic excitation can be calculat.. ed by any available deterministic dynamic integration scheme.e.g.the wel1.known Newmark-lf scheme… or Wilson.0 scheme(‘ .Therefore.standard finite e1e. ment procedures as developed for deterministic dynam— ic analysis can be employed. 1 K-L expansion Engineering parameters such as environmental loads and mechanical properties of the structural media hunen—Loeve(K—L)expansion[13' is a tool for the compact representation of arbitrary random processes, which can be used to characterize both weakly stationa. ry and non—stationary random functions.Therefore,the K—L expansion can be used to model seismic excitations and the responses. Consider a random process X(t)defined on a bounded domain t E[0,T],and assume that the process has a mean x(t)and a finite variance 2 (t).The process can be expressed as its mean and additional deterministic orthogonal functions muhiplied by uncor— related zero.mean random variables with unit standa deviation (£)= (£)+∑ ( ) (1) £=l in which A and (t)are the eigenvalues and eigen— functions of the covariance function Cx(tl,t2). According to random process theorem,the covari- ance function Cx(t1,t2)is bounded,symmetric and positive.Following Mercer’s theorem,it has the flow— ing eigen-decomposition: Cx(t1 ,t )=∑Ax (£ ) (f ) (2) I 1 and its eigenvalues and eigenfunctions are the solution of the homogeneous Fredholm integral equation of the second kind given by ( ,t ) ( t J)dt1=A (t ) (3) Eq.(3)arises from the fact that eigenfunctions form a complete orthogonal set satisfying the equation (t)xj(t)dt=8 (4) where is the Kronecker—delta function which satisfies ㈤ The random variable o/ in eq.(1)is a set of uncorre— lated random variables which can be expressed as ㈩ (6) 23期 徐瑞:K—L分解在非平稳地震响应分析中的应用 5625 with mean and covariance given by E(ot )=0 (7) 2 Analysis of random seismic response E( )= (8) The dynamic equation for n degree—of-freedom system under seismic excitation can be readily ex. pressed as Where E(。)denotes the expectation operator.The series expansion in eq.(1),referred to as the K—L ex. pansion,provides a second moment characterizati0n in terms of uncorrelated random variables and dete玎ninis. tic orthogonal functions.It is known that fhe K。L ex- pansion converges in the sense of mean square for any distirbution of X(t).For practical implementation,the series is approximated by finite terms,say m,giving (£)= (t)+∑ =1 i (f) (9) where Af is the largest m eigenvalues arranged in de. creasing order,say Al≥A2≥…≥A ≥A +1≥… (9) The accuracy of the eigen—pairs in representing the covariance function is hence govemed by the value of m・The corresponding covariance function is given by C (tl ,t2)=∑Ax ( ) ( ) (11) As ∞,eq.(11)converges to eq.(3).Ghanem and Spanos【l5 J have shown that the truncated series are optimal;that is,the mean square approximation error is minimized. The key to K・L expansion is therefore to obtain the eigenvalues and eigenfunctions of the cov卸riance flunc— tion by solving the homogeneous Fredholm integral equation of the second kind in eq.(3).The integral equation can be solved analytically only in special cir. cumstances,e.g.the correlation function is expressed as exponentiM function.In most cases.the anal ̄ical solution of the integral equation is not tractable and the numerical method is the only choice.Detailed introduce of numerical method can be found in references[ —。加]. In this paper,the procedure introduced in references ㈨is used to so1ve eq.(3). MY+CY+KY=一MLZ(t) (12) Where M,C and K are the global mass,damping and stiffness matrixes of the discretized structure,respec. tivety;Y, and denote the nodal displacement.ve— locity and acceleration vectors of the structure,respec— tively;L is the influence coefficient vector which re— presents the pseudo—elastic response in all degrees of rfeedom due to unit suppofl motion;and Z(t)is the rgound acceleration modeled as Gaussian non.stationary random process with zero mean.The K—L expansion of z(t)can be written as,see eq.(9) z(f)=∑ i=1 (f) (13) Where (f)denotes the th term of K—L expansion 0f Z(t);and m is the order of truncation of the series ex— pansion.Let F (£)=一ML (£) (14) F(£)=∑otiF (t) (15) Where (t) denotes the seismic loading.Consequent. ly,eq.(12) can be rewritten as +Cf+KY=F(£)=∑ ( )(16) Based on the superposition principle of linear sys. tems,for each deterministic terms F (t)of the load F(t),the associated deterministic structural resDonse (t)can be calculated by MYi+Cr,+gr,:F (t)(i=1,2,…,m)(17) Symbolically,this can be written as Fi( ) (t)(i=1,2,…,m) (18) Finally,the slims of all terms Yi(t)yields the struc. tural response 5626 科学技术与工程 ∑ / 10卷 y= 、,. L (19) St (∞)= ± ( 一602) + ~ ( ~ ) + ,.-2 2 2 = Based on eq.(19) ,the mean and the c0variance ma- / ,f●●●J,、●●trix of the structural response can be calcuIated as _●_f 、J (24) bove two cases can be expressed as (1 , Accordingly,the correlation function of Z(f)for the a— (t)=E[y(t)]=0 y(20) and ∑ 2 P =cov[y( ),y(f)]: E{[y(f)一tty(t)][Y( )一tty(f) : (21) Where T denotes matrix transpose 3 Numerical example The introduced procedure is applied to a 2 1一st0rev frame structure with 4 spans under non—stationan seis. mic excitation shown in fig.2.The Young modu1us E and density P of all the members are 20.0 GPa and 2 500 kg/m ,respectively.The section sizes 0f the member 1—3 are 0.7 m×0.7 m,0.8 m x 0.8 m. and 0.3 m x0.8 m,respectively.The mass 0f member is lumped on noda1.Damping ratios are assumed to be 5%for all vibration modes. Due to the lack of experimental data,gr0und ac— celeration Z(t)is modeled as uniformly modu1ated Gaussian random process that is expressed as Z(t)= g(t)U(t).The modulated function is given by ) ,0≤f≤ 】 g(f) tl≤t≤t2 (22) ) 2≤f≤T U(t)is a zero—mean stationary random Dr0cess.Fl0r iI. 1umination of convergence of K—L expansion,U(£)is chosetl as Kanai—T ̄imi(K—T)model[ ]and Clough— Penzien (C-P) model㈣, respectively.The p0wer spectrum function of K-T model and C—P mode1 are ex- pressed as s (∞)= ( ∞:+4 一 ) +4  ̄.2 —2 。 ̄2g~g 2 S。 (23) ( ,t+r)=g(t)g(t+r) (tr): g(t)g(t+ r) s )e (25) (f,£+ )=g(t)g(t+7.) (r): g(t)g(t s ∞)e (26) The expression of (丁)and (7-)can be found in reference[24]。For the numerical example.the values ti=8 s,t2=20 s,T=40 s,口:0.157 2. g= 4 rad/s, :0.6, =1.4 rad/s, =0.6 and 3 2 3 2 3 2 3 1 3 2 3 2 3 2 3 1 3 2 3 2 3 2 3 1 3 33 2 3 2 3 2 3 1 i3 1 1 3 2 3 2 3 1 1 3 2 3 2 3 l I 3 2 3 2 3 l 3 J g 寸 3 2 3 2 3 2 3 1 ii 3 2 3 2 3 2 3 1 × g 1 3 2 3 2 3 2 寸 i 3 2-3 2 3 2 1 3 2 3 2 3 2 I 1 3 2 3 2 3 2 1 3 2 3 2 3 2 1 3 2 3 2 3 2 3 2 3 2 3 2 3 l 3 2 3 2 3 2 3 l 3 2 3 2 3 2 3 3 2 3 2 3 2 3 (Z】t 三5 上5 工5卫上5 Fig.2 A frlame structure subjected to seismic excitation  _【Ilu\I10一 1 5 ● 2 0 9 0 6 gII ∽ O 3 O O 23期 徐瑞:K—L分解在非平稳地震响应分析中的应用 5627 S0=0.6 cm /s have been used.The Newmark-lf regarding the number of K—L expansion terms used for scheme for deterministic dynamic analysis is chosen to representation of the two cases,i.e. Kanai-Tajimi calculate eq.(17). (K—T)model and Clough—Penzien(C—P)mode1.In The introduced procedure is adopted for the non— particular,the standard deviation of horizontal displace— stationary solution of seismic responses of structure.In ment of the 7th floor,14th floor and the 21st floor,are order to demonstrate the accuracy of the procedure,the shown in fig.3 and ng.4,respectively.The maximum non——stationary solution of structure response is com-・ standard deviation of horizontal displacement of each pared with the exact solution obtained by the power lfoor is shown in fig.5 and fig.6,respectively. spectrum method .A convergence study is conducted 【_IIu\口0一 _l 口DJ鲁暑 力 Time/s Time/s fa) (b) g g ‘0 .曼 昌 蠢 Time/s fc1 Fig・3 ( )is in。deled as K-T In。de1:deviation of horizontal di placement。f the 7th floor,14th lfoor and 2lst lfoor.obtained by the introduced procedure with the number m of K—L terms used as 100,200,220,and 240. It can be seen from fig.3 to fig.6,about 240 K— performed for the case of K—T model,and 220 dynamic L terms are required in order to obtain accurate second integrations for the C—P mode1.It is quite obvious,that moment characteristics of the structure as f t)is mod— the computational effort of the introduced procedure is eled as K.T model,and about 220 K.L terms are re. different from different random mode1.and the compu. quired to obtain the second moment characteristics of tational effort strongly relies on the number of K.L the sturcture while U(t)is modeled as C.P mode1. terms Used. That is,240 dynamic integrations must also are 5628 科学技术与工程 l0卷 吕u/【10一苗一A3々胃口ⅡBls 吕u\【10 【^0 pJ弓【Ig∞吕T1吕【x 苫 l l l 1 O O O 0 0 3 3 2 2 1 l 0 0 6 4 2 0 8 6 4 2 O 5 O 5 O 5 0 5 0 暑 ,【10一 q p朋q哥1s 0 5 10 15 20 25 3O 3 35 3 40 2 1 1 O O Time/s 6 O 4 8 2 6 O O 5 10 l5 20 25 30 35 40 (a) Time/s fb) g/u0焉一 p朋 窝 3 2 2 1 O 0 0 5 O 5 O 5 O g0/u0 B1> 口pJ时 量∞宣;吕一 苫 3 3 2 1 1 O O 6 0 4 8 2 6 0 ,- -.__ ,  _十 Fig.4 U(t)is modeled as C-P model:deviation of horizontal displacement of the 7th floor,14th floor and 2lst,  floor, obtained by the introduced procedure with the number m of K—L terms used as 1 00,1 80,200,and 220 f x x. . 薹t . 一… 霎a 。c 毒:,。:量・: fft。l 1X.X t soluti   . X. ,× x. X X. 1 3 5 7 9 11 13 15 17 19 2l l 3 5 7 9 l1 13 15 17 19 21 Floor number Floor number Fig.5 U(t)is modeled as K—T model:maximum devia— Fig.6 U(t)is modeled as K—T model:maximum devia・ tion of horizontal displacement of each floor,obtained by tion of horizontal displacement of each floor,obtained by the introduced procedure with the number m of K—L terms the introduced procedure with the number m of K—L terms used as 100,200,220,and 240. used as 100,180,200,and 220. variance matrix of the response can be easily calculated 4 Conclusions by the procedure which is valid for any type of non—・sta・・ tionary seismic excitation.In fact,the procedure can Base on K—L exPansion,a procedure is introduced be extended to any non—stationary response analysis to calculate the random response of structure subjected where the second moment characteristics of the excita— to non・stationary seismic excitation.The mean and CO— tions are known(including non-Gaussian excitation). 23期 徐瑞:K—L分解在非平稳地震响应分析中的应用 5629 The important advantage of this procedure is that the well known deterministic integration procedures can be applied.Due to this feature the procedure is easily im— plemented in commercial finite element packages.By this,the use of random vibration analysis in engineer- ing practice is significantly enhanced. Although the procedure’s advantage is bright on not calling for modiifcation of existing commercial finite element packages,it’s worth to note that the conver- gence and accuracy of seismic random response strong- ly depend on the truncated terms of K—L expansion, and the number of terms of K—L expansion required dif- fers from various non—stationary seismic model chosen. Another problem is that more than two hundreds of terms of K—-L expansion are required to obtain the ra-- tional results,for complex and large structures with ten thousands of degree—of-freedom or even more,the COB— putational cost is still unacceptable.As an outlook,the author’s future work will be concentrated on develo— ping efficient methods for determining rational scale of the terms of K—L expansion and reducing the terms of K—L expansion with the reasonable accuracy guarantee of random response. 参考文献 1 Zhu W Q.Random vibration keferences.Beijing:Science Press,1992 2 Soong T T,Grigoriu M.Random vibration of mechanical and structur- la systems.New Jersey:Prentice・Hall,1993 3 Lin J H.Zhang Y H Pseudo-excitation method of random vibration.Beijing:Science Press,2004 4 Lin J H,Zhong W X,Zhang W S.Precise integration of the varinace matrix of structural non stationary random responses.Journal of Vibra— tion Engineering,1999;12(1):1m8 5 Sun Z Y.Wang H.Calculation method of the variance matirx of struc— tural non-stationary Iltndom responses.Journal of Vibration Engineer- ing,2009;22(3):325—028 6 Hoshiya M,Ishii K,Nagata S.Reeursive covarinace of structural re— sponses.Journal of Engineering Mechanics,1984;110(12): l743—1755 7 To C W S.A stochastic version ofthe Newmark family of algorithms for discretized dynamic systems.Computers and Sturctures,1992; 44(33):667—673 8 To C W S.Iju M L Recursive expressions for time dependent mesns and mean square responses of a multi—degree—of-freedom non—。linear system.Computem and Structures,1993,48(6):993—1oo0 9 Miao B Q Direct integration varinace prediction of random response of nonlinear systems.Computers and Structures,1993;46(6):97 3 10 Miao B Q,Hu X X,Xu L S.Recursive Newmark algorithm for the analysis of nonstationary random structurla response.Structure and Environment Engineering.1996;23(1):34—09 1 1 Newmark N M.A method of computation for structural dynamics. Journal of Engineering Mechanics,1959;85(3):67 12 Wilson E L A computer program for dynamic stress analysis of an— derground structures.SESM Report(No 68—1),University of Cal— ifomia.1968 13 Karhunen k Uber lineare methoden in der nung.Annales Academiae Seientiamm Fennicae(Series A1):Math・ ematica—Physics,1947 14 Loeve M.Probability theory.Berlin:Springer—Verlag,1977 15 Ghanem R,Spanos P D.Stochastic finite element:a spectral印一 proach.New York:Springer,1991 16 Phoon K K,Huang S P,Quek S Implementation of Karhunen— Loeve expansion for simulation using wavelet—Gale/kin scheme.Probabilistie Engineering Mechanics,2002,17(3):293— 303 17 Phoon KK。HuangHW。Quek S Comparison betweenKarhunen— Loeve and wavelet expansions for simulation of Gaussian proces— ses.Computers and Structure,2004;82(5):985—991 18 Zhang J, Ellingwood B.Orthogonal series expansions of rnadom ifelds in reliability analysis.Journal of Engineering Mechanics, 1994;120(12):2660--2667 l9 Liu T Y.Wu Z H.Zhao G Projective expansion method of rail- dom fields.Chinese Journal of Computational Mechanics,1997, l4(4):484--489 2O Li J.Liu Z J.Expansion method of stochastic processes based on normalized oflhogonal bases.Journal of Ton ̄i University,2006; 36(10):1279--1282 21 Sehueller G I.Pradlwater H .Uncertainty analysis of complex sturt— rural systems.International Journal for Numerical Methods in Engi- neefing,80(6_7):881__913 22 Kanai k Semi-empirical formula ofr the seismic characteristics of the ground motion.Bulletin of the Earthquake Research Institute,Uni— versity of Tokyo,1957;35(2):308—325 (下转第5643页) 23期 马一玫:聚驱高矿化度采出污水配制聚合物溶液方法研究 5643 Preparing Polymer Solution with Higher Salinity Sewage of Polymer Flooding Field MA Yi.mei (Northeast Petroleum University,Petroleam Engineering College,Daqing 163318,P.R.China) l Abstract]Aims at the problems of high salinity in sewerage in the 4th Oil Rec。very P1ant。f Daqing Oil Field, the influencing factors。f polymer s。lution Viscosity by sl mu1ating kinds。f ions and salinity are researched.A con— clusion is get that the higher the salinity。f water,the lowe polymer s。luti。n viscosity;the m。re the c。ntents of ca“,Mg and Na ,the lower the p。lymer s。luti。n viscosity.Demu1sifers,anti-scaIing additive,and pammn in. hibi10r have n。influence on polymer soluti。n Viscosity,flocculant and bactericide can reduce the viscosity.By re. searching the reduce of the salinity of sewerage with the method of eletrodialysis in the 4th Oil Recovery Plant,it indicates that electrodialysis can lower the salinity more, and also can wipe off all ions,the viscosity of polymer so— lution disP。sed by the freshwater which has been disP。sed by the way。f electrodialys1。s is higher than viscosity hasn't been disP。sed,SO the disP。sed freshwater c。uld be used when c。nfecting p01ymer s。luti。n. [Key words]salinity polymer solution viscosity electrodilaysis sewage ≯ (上接第5629页) 23 Clough R W,Penzien J.七’yI1踟ics of structures.New Y0rk:McGr 卜 calculation of their correlation functions.Supervision Test and C0st 0f Hill,1993 Construction,2009;2(6):I5—18 24 Chen X F,Xu R.Common mode1s of earthquak ound m0tinn and Application of K—L Expansion for the Analysis of Non—stationary Seismic Response XU Rui .(Scho0l 0f ciViI Ensineering and Transportati0n,South China University of Techn0l。gy,GuI衄g,h。u 510640,P R.Chin丑) [Abstract] rI’he。bjec iVe 0f th。present w0rk is introduced to a new procedure fo¨and。m response analysis of s ruc ure subjec ed t0 nOil。stationary seismic excitation. According to Karhunen.Loeve(K.L) expansion the0rem. wh ch al10ws t0 represent random seismic random excitation as a series terms involving a complete se£of determinis— t c iUnct ons with corresponding random coefficients. Clearly,for linear structures,each detem1injslic te瑚of K—L expans 0n 0i the se s蛐c excitation corresponds to a response.Therefore,any available dete邢inistic dynamic jnte. grat on scheme can be applied to calculate structural response associated with each term of K.L expansi0n of excita— t 0n pr0cess・Th s m tuITI opens the avenue for the use of commercial deterministic finite element paekages in seis— mlc rand0m anaJys s,which may greatly enhance the acceptance and use of random analysis in ensincering prac一 ‘ ces・ A zl st0ry frame structure subjected to uniformly modulated non.stati0nary seismic excitation was used t0 demonstrate the procedure. [Key words] seismic rand0m vibI_ation K—L expansion non—stationary random excitation 

因篇幅问题不能全部显示,请点此查看更多更全内容

Copyright © 2019- 517ttc.cn 版权所有 赣ICP备2024042791号-8

违法及侵权请联系:TEL:199 18 7713 E-MAIL:2724546146@qq.com

本站由北京市万商天勤律师事务所王兴未律师提供法律服务