by KamalenduGhosh 09CE3112 Under the Guidance of Prof. Arghya Deb Department of C!l Engneerng "ndan "nsttute of #e$hnology Kharagpur 2012%13 Department of C!l Engneerng "ndan "nsttute of #e$hnology Kharagpur%&21302 CERTIFICATE #hs s to $ertfy that the pro'e$t ttled (Numerical Analysis of Tuned Liquid Dampers for structural Control) s a bona*de re$ord of the +or, $arred out by amalendu !"os"#$%CE&''()* under my super!son and gudan$e for the partal ful*llment of the re-urements for the degree of .a$helor of #e$hnology n C!l Engneerng/ "ndan "nsttute of #e$hnology/ Kharagpur. Dr. Arghya Deb Department of C!l Engneerng "ndan "nsttute of #e$hnology Kharagpur Date0 2 Ac+no,led-ement " +ould l,e to e1press my sn$ere grattude to Prof. Arghya Deb for gudng me on my +or, n mdst of ther e1tremely he$t$ s$hedule he had. 2s $onstant mot!aton and support en$ouraged me to ,eep gong. 2s tmely nter!entons helped me o!er$ome glt$hes n my +or,. " than, hm on$e agan for hs o!er+helmng support. Date0 KamalenduGhosh
09CE3112 3 TA.LE /F C/NTENTS C"apter '0 Introduction 1 C"apter (0 Literature Sur2ey 3 C"apter &0 4et"odolo-y &5' T"e Approac" 6 &5( Some 4at"ematical formulations 6 &5( Finite Element 4odel &5&5' TLD % &5&5( Liquid % &5&5& Tan+ '$ &5&51 Interactions '$ &5&53 !ra2ity Load '$ &5&56 4odi7cations '$ &5&58 4odelin- for a -i2en structure '( C"apter 90 Results '3 3 C"apter 10 Conclusion '% C"apter 30 Future :or+ ($ References (' 4 '5 Introduction #he $urrent trend of buldng stru$tures of e!er n$reasng heght and the use of lght+eght/ hgh strength materal s often a$$ompaned by n$reasng sus$eptblty to e1$tatons su$h as +nd/ o$ean +a!es and earth-ua,es. #o redu$e the rs, of stru$tural falures/ part$ularly durng a $atastroph$ e!ent/ t has be$ome mportant to sear$h for pra$t$al and e5e$t!e de!$es for suppresson of these !bratons. #he de!$es used for suppressng the stru$tural !bratons $an be $ategor6ed a$$ordng to ther energy $onsumpton0 Pass!e Control A$t!e and 7em%a$t!e $ontrol 2ybrd system $ontrol system. Pass!e $ontrol de!$es are systems that do not re-ure e1ternal energy supply. 7u$h systems are relable as they are una5e$ted by po+er outage/ +h$h s $ommon durng earth-ua,es. #hese de!$es dsspate energy usng the stru$ture8s o+n moton to produ$e relat!e moton +thn the $ontrol de!$e. An ad!antage of su$h systems s ther lo+ $ost of mantenan$e be$ause no a$t!aton me$hansm s re-ured. E1amples of su$h systems n$lude !s$ous dampers/ tuned mass dampers/ tuned liquid dampers/ et$. "n #9Ds the sloshng moton of the :ud that results from the !braton of the stru$ture dsspates a porton of the energy released by the dynam$ loadng and therefore n$reases the e-u!alent dampng of the stru$ture. #uned sloshng dampers $an be $lass*ed nto t+o $ategores/ shallo+ and deep +ater dampers based on the rato of +ater depth to tan, length n the dre$ton of moton. #he #9D system reles on the sloshng +a!e de!eloped at the free surfa$e of the l-ud to dsspate a porton the dynam$ energy. #he ; gro+ng nterest n #9Ds s due to ther lo+ $aptal and mantenan$e $ost and ther ease of nstallaton n stru$tures. (5 Literature Sur2ey Pass!e dampng usng #9Ds s a $hallengng area of resear$h and nno!aton and as su$h a host of lterature +h$h $o!ers ths d!erse area e1sts. 7ome referen$es are summar6ed here +th a gst of ther $ontent. <n%Kyu et al.=1999> ds$usses the numer$al modelng of a tuned l-ud damper=#9D> as an e-u!alent tuned mass damper +th non%lnear st5ness and dampng. #hs non%lnear st5ness and dampng =?7D> model $aptures the beha!or of the #9D system ade-uately under a !arety of loadng $ondton. "n part$ular/ ?7D model n$orporates the st5ness hardenng property of the #9D under a large ampltude e1$taton. @eed et al. =199A> $on$luded that the dsspaton of !braton energy due to a #9D n$reases as the e1$taton ampltude n$reases and the #9D beha!es as hardenng sprng system. #he authors also sad that to a$he!e a more robust system the desgn fre-uen$y for the damper/ f t s $omputed by the lnear6ed +a!e theory/ should be set at a !alue lo+er than that of the stru$ture response fre-uen$y/ sn$e ths may result n the a$tual nonlnear fre-uen$y of the dampng mat$hng the stru$tural response. #he authors also found that e!en f there damper fre-uen$y s mstuned slghtly/ the #9D & al+ays performs +ell. #hey obser!ed no ad!erse e5e$ts due to slght mstunng. Kareem et al. =2009> elaborated on the nherent non%lnear beha!or of the #9Ds. #he non%lnear beha!or of the #9Ds s $hara$ter6ed by an ampltude dependent fre-uen$y response fun$ton/ +h$h translates nto $hanges n fre-uen$y/ as +ell as $hanges n the dampng due to sloshng/ +th ampltude. "n ths paper #9Ds are modeled usng a sloshng%slammng =7 2 > analogy/ +h$h $ombnes the dynam$ e5e$t of l-ud sloshng and slammngBmpa$t. Kareem et al. =1994> ds$ussed the dynam$ $hara$terst$s and e5e$t!eness of multple mass dampers =CCDs>. A $olle$ton of se!eral mass dampers +th dstrbuted natural fre-uen$es under random loadng +ere n!estgated n ths paper. #he CCD atta$hed n parallel $on*guraton mod*es the transfer fun$ton of the damper buldng +th a sngle #uned Cass Damper. #he CCD parameters $onsdered here n$lude the fre-uen$y range of the CCDs/ dampng rato of nd!dual dampers/ and the number of dampers. A parametr$ study +as $arred out to $ompare the results +th a sngle tuned mass damper. "t +as demonstrated that he CCD $on*guraton s more e5e$t!e n $ontrollng the moton of the prmary system. &5 4et"odolo-y 3.1 The Approach In a Lagrangian finite element formulation the mesh motion is the same as the motion of the material. In TLDs, the motion of the liquid, especially at the free surface, can result in extreme deformations. This will result in severe mesh deformation as well. This will cause the Jacobian to become very small or even negative. negative determinant of the Jacobian matrix would cause A the analysis to terminate. In order to prevent this, and to permit simulation of large amplitude sloshing motions, a coupled Lagrangian!"ulerian approach was adopted. In contrast to a Lagrangian analysis, where the Lagrangian elements comprise a single material only, in an "ulerian analysis the "ulerian elements may not always comprise of a single material# many may be partially or filled with air i.e. void. The "ulerian material boundary must, therefore, be computed during each time increment and generally does not correspond to an element boundary. The recomputed material boundaries must ensure conservation of mass, momentum and energy. In addition, the nodes in an "ulerian finite element mesh are fixed in space, unli$e in the Lagrangian case where the nodes move with the material. The "ulerian mesh is typically a simple rectangular grid of elements constructed to extend well beyond the "ulerian material boundaries, giving the material space in which to move and deform. If any "ulerian material moves outside the "ulerian mesh, it results in loss of mass from the simulation, and in the present application at least, in accurate results. %nli$e the liquid in the tan$, the tan$ and its walls undergo limited. &e thus adopt Lagrangian finite element analysis for these parts. 'ut at the liquid structure interface, it is necessary to reconcile the two different formulations through "ulerian!Lagrangian (ontact. In this approach, at the interface the liquid motion results in traction forces at the boundary of the solid, while the displacements of the solid result in displacement or velocity boundary conditions for the liquid. 3.2 Some Mathematical formulations The governing equation in the fluid domain is based on potential flow theory. If the fluid is assumed to be inviscid and incompressible, and the flow irrotational, the governing equation is given by Laplace)s equation, * + ,-. 9 &here ,/x,y,t0 is the velocity potential function. 1owever in a moving tan$ the total potential function can be split into two parts, the potential function ,t due to the moving tan$ and the disturbed potential function ,s due to fluid motion. ,- ,t 2 ,s "xpressing the Laplace equation in terms of ,s * + ,s-. Hydrodynamic sloshing force 1ydrodynamic sloshing force is obtained by integrating the hydrodynamic pressure over the pro3ected area of the tan$, which in turn acts on the structure as a base shear force. If the base shear force is out of phase with the motion of the structure, it is li$ely to damp structural motion. The base shear is defined as# 4i-b 5 !
6 where 78 and 7+ are the free surface sloshing elevation on the two side walls of the tan$, p/!L9+, y,t0 are the liquid pressures at x-/2:0L9+, and b is the width of the tan$, the dimension in the other orthogonal direction. 1ydrodynamic pressure is obtained as# ;-!p5..<= + 2 / + > 2 2 g.y 2 st 2 ?,s 6 In the above equation st is the acceleration included in the tan$ due to structure response, 10 ?,s is the damping term in liquid sloshing. Shallow water and deep water Tuned Liquid Damper Tuned sloshing dampers can be broadly classified into shallow and deep water dampers based on the water depth to tan$ dimensions in the direction of motion. In shallow water dampers structural vibration energy is mainly dissipated through the sloshing and wave brea$ing. The sloshing of the fluid generates hydrodynamic force , primarily in the form of base shear that acts to control structural response. 'ased on the linear wave theory /Lamb 8@A+0, fundamental natural frequency of the liquid sloshing fw, is fw= 1ere L is the length of tan$ measured in the direction of excitation and h is the stationary water depth. In case of shallow water dampers the h/L ratio is limited to ..8<, thus water depths required for short period structures from the above equation are too small for practical application. In addition, the above equation for sloshing frequency is very sensitive to the water depth to length ratio /h/l0. Thus any fluctuation in the water level will alter the sloshing frequency. It is found that shallow water dampers do not perform well for broad range frequency excitations. These shortcomings of shallow water dampers can be overcome by adopting deep water dampers. In deep water dampers optimal tuning frequency can be ensured by adopting water depth ratios of more than 8 /since sloshing frequency depends on the length of water tan$0. 4undamental frequency of deep water dampers /with h/L B80 is given by# 11 fw = It is clear from the above equation that for deep water dampers the sloshing frequency is insensitive to water depth ratio. 1ence deep water dampers can be tuned for short period structures and hence there is a possibility that they can perform well for earthqua$e or wind excitation. 3.3The Finite Element Model The details of the finite element model are discussed in the next section. 3.3.1 TLD The TLD was modeled using 'C%D9"xplicit. Dince liquid sloshing is essentially a +!D phenomenon, a strip model of the tan$ was chosen. The dimensions of the tan$ and the "ulerian domain are as given below# Part Dimension( in metres) Tan$ A.<<x+.8x...E "ulerian domain 8.x<x...E water reference part was created for computation of volume fraction of the liquid in the tan$. The bloc$ was meshed with bric$ shaped finite elements of the type "(ADFG with siHe ...Ex...Ex...E m A and the initial volume fraction of liquid in each element was computed using a partial filling strategy. 3.3.2 Liquid The liquid was modeled with a linear %s!%p equation of state material. 1owever the speed of sound in liquid c. was not ta$en the actual value of 8EI.m9s, but a magnitude two orders small. This had the advantage of reducing the bul$ modulus of the liquid, increasing the stable time 12 increment and thereby allowing the analysis to be completed in a realistic time frame. viscosity of magnitude of 8." !< was also added. 3.3.3Tank The tan$ was modeled using (ADFG solid elements in view of the fact that the deformation of the tan$ does not significantly affect the sloshing motion of the liquid. The dimensions /internal0 are as given above. The tan$ was also made a rigid body. 3.3.4 Interactions The sloshing motion of the liquid will result in impact between the liquid and the tan$ walls. 1ence, contact between the "ulerian liquid material and the solid parts were enforced using general contact in 'C%D9"xplicit. Dpecifying automatic contact for the entire model allowed interactions between all Lagrangian structures and the liquid. The contact penalties were default penalties calculated by 'C%D, which were the maximum possible penalty values that can be used without adversely affecting the stable time increment. 3.3.5 Gravity Load Jravity load was applied using smooth step and initial condition stress conditions were specified corresponding to a hydrostatic state of stress were prescribed.. The model was submitted for analysis. The Kinetic energy of the system was found to increase to a maximum point and then decrease. 1owever the K" again increased by small magnitudes indicating lea$ages. 3.3.6 Modifications 4illets were added to the sharp corners of the tan$ walls and additional layers of tan$ walls were introduced. 4illets helped remove the singularity in the surface normal at sharp corners in the tan$ walls, while additional layers of tan$ walls provided the extra contacting surfaces that trapped any lea$ed material and limited its escape. &ith these modifications it was possible to prevent lea$age significantly of the liquid and allowed reasonable numerical solutions. 13 4ollowing figures depict the final TLD model# 4ig. 8 The TLD model set up 13 4ig. + The TLD model with "ulerian domain. 4ig. A The TLD model and "L4 at the end of Dtep!8 /Jravity Dtep0 Initially a liquid with density-8... $g9m A was ta$en. 1owever the dampers with this liquid did not show much damping. This was because the mass ratio 8 was too small. &e thus increased the density to +E8. $g9m A to obtain a reasonable mass ratio of ..8 for the shallow water damped tuned to the frequency of the structure. 3.3.7 Modein! for a !iven structure structure was modeled using +!D beam elementsM so that it is first "igen frequency would be small enough to allow us to model a shallow water tan$. The following properties and dimensions were chosen# 'eam (olumn 1 Cass ratoDCass of damperBCass of stru$ture 14 "- A8.IJ;a "-A8.IJ;a Density- +E..$g9m A Density-+E.. $g9m A Length- 8. m. Length- +<m. (ross!section- ..Ax..A m + (ross!section-..Fx..Fm + The natural frequency of the structure was found out using 'C%D. 4or the above natural frequency, the dimensions of the tan$ and the height of water were calculated using the sloshing frequency formula# fw= The tan$ was placed on the top beam of the structure and analyHed for ground excitation. The structure without the damper was also analyHed for the same excitation. 1; 4ig. < Nodeling the structure 4ig. I 4inding the fundamental frequency of the structure 1& 4ig. O The TLD water height is modified and fixed at the top beam 4ig. F 1A Dloshing of the liquid in damper due to applied base excitation 4ig. @ Dloshing of the liquid in the tan$ . Two dampers, a shallow water damper and a deep water damper both tuned to the frequency of the structure were modeled. Shallo ater tan! Deep ater tan! Length /l0 - Am Length /l0 - I.FEm 1eight of water /h0 - ..E<m 1eight of water /h0 - Om h9l - ..8< h9l B8 Nass ratio -..8 Nass ratio - A.E 19 95 Results 95' Sub;ectin- t"e structure to Koyna base e<citation #he undamped stru$ture and the damped stru$tures =+th shallo+ +ater tan, and deep%+ater damper> +ere sub'e$ted to the Koyna base e1$taton and analy6ed. #he follo+ng results +ere obtaned0 20 Eg. 10 Dspla$ement =of the topmost node> !s #me Eor the stru$ture +th shallo+ +ater damper/ dampng s $learly !sble. 2o+e!er/ due to phase shft the dampng e5e$t of the deep +ater damper s not $on$lus!e from ths *gure. #he nternal energy of the stru$ture +as therefore e1amned to ma,e some de*nt!e $on$lusons regardng dampng e5e$ts =Eg. 11>. 21 Eg. 11 #me 2stores of the "nternal Energy of the stru$ture Agan nothng $on$lus!e $an be dra+n from ths *gure. #he nternal energy +as therefore ntegrated to obtan Eg. 120 Eg. 12 #otal "nternal Energy #hus t s $lear that o!er the duraton of the analyss/ the total nternal energy =stran energy> of the stru$ture for a shallo+ +ater damper s 22 lesser than that n the $ase of an undamped stru$ture. 7n$e the total nternal energy of the stru$ture s a measure of the deformatons undergone by the stru$ture o!er the duraton of the loadng/ t s $lear that the shallo+ +ater damper results n sgn*$ant dampng. 2o+e!er/ the deep +ater does not $ause dampng. @ather the stru$tural stran energy s hgher at the end of e1$taton and the stru$ture s n dstress. #o understand the reason for ths the KE of the stru$ture and the +ater for the shallo+ and deep l-ud $ases +ere plotted0 Eg. 13 Knet$ Energy of the stru$tures +th shallo+ +ater damper and deep%+ater damper 23 Eg. 13 Knet$ energy of the l-ud n the tan, n shallo+ +ater and deep%+ater damper #he KE of the stru$ture/ the +ater and the stran energy are all larger for the deep +ater damper $ase. 2o+e!er the total energy of the e1$taton s the same for both the shallo+ and the deep +ater damper are the same. #hus t $an be $on$luded that the response of the deep +ater damper s unstable. #o $on*rm ths a EE# 2 analyss of Koyna e1$taton +as $arred out. 2 EE# D East Eourer #ransform 23 Eg. 14 Koyna a$$eleraton data =tme hstory> Eg. 1; Ere-uen$y Doman Erom Eg. 1; t s $lear that the Koyna e1$tatons $ontan sgn*$ant spe$tral densty n lo+ fre-uen$y bands $lose to the natural fre-uen$y 24 of the stru$ture. Fhen the shallo+ +ater damper s repla$ed by the deep +ater damper/ the addtonal mass of the +ater results n a redu$ton n the natural fre-uen$y of the $omposte stru$ture%damper system. "n the present $ase/ +hat appears to be happenng s that the redu$ton n fre-uen$y of the stru$ture%damper system results n near% resonan$e $ondtons/ sn$e no+ sgn*$ant spe$tral densty s present at or near the natural fre-uen$y of the stru$ture. #hs results n energy gro+th or Pnegat!e dampngQ that results n unstable response. #he e5e$t of n$reasng the number of #9Ds/ +th fre-uen$es spannng a range of fre-uen$es/ +as studed by atta$hng t+o #9D s to the stru$ture. 95( Sub;ectin- t"e structure to El-Centro base e<citation #he undamped stru$ture and the damped stru$tures =+th shallo+ +ater tan, and deep%+ater damper> +ere sub'e$ted to El-Centro base e1$taton and analy6ed. #he follo+ng results +ere obtaned0 Eg5 1& Dspla$ement =of the topmost node> !s #me 2; .oth n the $ases of shallo+ and deep +ater dampers dampng +as obser!ed. Agan the nternal energy of the stru$ture +as e1amned to get a more a$$urate estmate of the amount of dampng. Eg 1A #me 2stores of the "nternal Energy of the stru$ture #he tme hstory of the nternal energy sho+s that the nternal energy of the stru$ture n the damped stru$ture =both shallo+ and deep%+ater damper> s $onsderably smaller than the undamped stru$ture. #he total nternal energy presents a $learer p$ture0 2& Eg. 19 #otal "nternal Energy #hus both the deep +ater damper and the shallo+ +ater damper/ tuned to the stru$tural fre-uen$y/ damp the stru$ture $onsderably. EE# +as performed on the El-Centro e1$taton +th the am of desgnng a damper tuned to the domnant e1$taton fre-uen$y and studyng ts e5e$ts on stru$tural !bratonal mtgaton. Eg. 20 El%Centro a$$eleraton data 2A Eg. 21 Ere-uen$y Doman #he deep +ater damper thus modeled tuned to the most domnant e1$taton fre-uen$y ho+e!er dd not e1hbt any sgn*$ant dampng. Eg. 22 Dspla$ement of topmost node !s. tme 29 Eg. 23 "nternal Energy of the stru$ture 95( Sub;ectin- t"e structure to periodic #sinusoidal) base e<citation #he undamped and damped stru$tures +ere sub'e$ted to perod$ e1$tatons +th fre-uen$es e-ual to the domnant fre-uen$y of the Koyna and El%Centro e1$tatons. #o study the n:uen$e of nonlnear =e1$taton ampltude dependent> e5e$ts/ three d5erent ampltudes +ere used for the snusodal e1$tatons. #he ampltude A +as ta,en as 0.02 L 3 / 0.03L 3 and 0.04L 3 respe$t!ely/ for these smulatons +th !aryng ampltudes. 95(5'Sinusoidal e<citation ,it" frequency equal to t"e most dominant frequency of Koyna e<citation # = > 8 ?@) 95(5'5' AAL >$5$( 3 AD Ampltude of e1$taton 9D9ength of the tan, 30 Eg. 24 Dspla$ement of the topmost node !s. tme Eg. 2; "nternal Energy tme hstory 31 Eg. 2& #otal "nternal Energy 95(5'5( AAL >$5$&& Eg. 2A Dspla$ement of the topmost node !s. tme 32 Eg. 29 "nternal Energy tme hstory Eg. 30 #otal "nternal Energy 95(5'5& AAL > $5$1 33 Eg. 31 Dspla$ement of the topmost node !s. tme Eg. 32 "nternal Energy tme hstory 33 Eg. 33 #otal "nternal Energy 95(5(95(5( Sinusoidal e<citation ,it" frequency equal to t"e most dominant frequency of El-Centro e<citation # = > '51 ?@) 95(5(5' AAL >$5$( 34 Eg. 33 Dspla$ement of topmost node !s. tme Eg. 34 "nternal Energy tme hstory Eg. 3; #otal "nternal Energy 3; 95(5(5( AAL >$5$&& Eg. 3& Dspla$ement of topmost node !s./ tme Eg. 3A "nternal Energy tme hstory 3& Eg. 39 #otal "nternal Energy 95(5(5& AAL>$5$1 Eg. 30 Dspla$ement of topmost node !s. tme 3A Eg. 31 "nternal Energy tme hstory Eg. 32 #otal "nternal Energy 95(5& Bercenta-e dampin- for all t"e cases discussed abo2e 39 Type of e<citation 4a< Internal Ener-y C dampin- 9 Dndamped #') Damped #S"allo, ,ater) #() Damped #deepE ,ater) #&) S"allo, :ater Tan+ DeepE ,ater Tan+ Koyna E<citation 3.24110 3 2.4110 3 3.24110 3 23.0& %30.&& Sinusoid al #=>8 ?@) AAL $5$( 3;.194110 3 29.;;110 3 3;.32110 3 1A.04 %0.;2 AAL $5$&& 1A110 3 13.&4110 3 14110 3 23.;1 1;.;& AAL $5$1 34.3110 3 29.9110 3 30.9110 3 14.43 12.&1 El-Centro E<citation 0.2110 ; 0.03A110 ; 0.0394110 ; &; &4.24 Sinusoid al #=>'51 ?@) AAL $5$( 4.3110 3 2110 3 3110 3 ;2.9; 33.33 AAL $5$&& 13110 3 3.9110 3 ;.4110 3 ;4 43.4& AAL $5$1 2&.4110 3 10110 3 13110 3 ;3.;3 42.&3 15 Conclusions Cass rato has a sgn*$ant e5e$t on dampng. A small !alue of l-ud mass resulted n !ery lttle dampng. 7gn*$ant n$rease n dampng +as not$ed on n$reasng the mass of the l-ud. Erom the present results t +as found that n all the $ases a shallo+ +ater tan, resulted n more dampng than a deep%+ater damper. #hus n$reasng mass rato does not result n n$reased dampng beyond a lmt. A shallo+ damper/ +th the sloshng moton of the l-ud domnant/ $learly has more dampng e5e$t than a deep +ater damper. E5e$t!eness of the damper n$reased as the rato +as n$reased from 0.02 to 0.03. .ut further n$rease n ampltude of e1$taton dd 3 G dampngD 1100 30 not bear out ths trend0 the per$entage dampng de$reased as +as n$reased to 0.04. #hus the n:uen$e of the ampltude of e1$taton does not appear !ary monoton$ally. 31 References 1> A.AHI7 ;.9 Jnlne Do$umentaton. 2> Dorothy @eed/ 2arry Keh/ <n,yu (An "n!estgaton of #uned 9-ud Damper for 7tru$tural Control). 3> <n%Kyu%Ku/ #oshhro Fa,ahara and Dorothy A. @eed =1999> (A ?on%9near ?umer$al Codel of the #uned 9-ud Damper). 3> Ahsan Kareem/ 7+aroop Kalla/ and Cegan C$Cullough/ (7loshng%7lammng Dynam$s%7 2 % Analogy for #uned 9-ud Dampers). 32