¢ sipeii Beaters 1.0 T ti HAL T 7 08 | + 3. | | | ost! ! ! IX 2/H Ht TET i os j | \ 02 | 0.2 0.4 6 0.8 1.0 Go (21/9 (0) PIG. C2.18 : DIMENSIONLESS IMPULSIVE PRESSURES ON RECTANGULAR TANK WALL PERPENDICULAR TO DIRECTION OP EARTHQUAKE TANK WALL PERPENDICULAR TO DIRECTION OF EARTHQUAKE 1.0 ~ i I ] | | 08 7 —t | | | I 0.6 7 7 qol0) . 0.6 | 02 ] | HAL PIG. C2.19 : PEAK VALUE Of DIMENSIONLESS IMPULSIVE PRESSURES ON RECTANGULAR WALL PERPENDICUEAR TO DIRECTION OF FARTAa piz,.t) = ie ‘) z.t) + p(zetl Generally the walls of rectangular tanks will be suffic flexibl to significantly increase the period of the impulsive vibration. For this case, reference should be made to the modifications given in the section on flexible tanks below. Impulsive Pressure Component on Walls The impulsive pressure component on the walls is given oy: pyl2) = agl2) ey(Tg) vy The shape and magnitude of the dimensionless pressure funct varies with the geometric ratio H/L as shown in Figures C2. 23 c2.19. Convective Pressure Component on Walls The maximum convective pressures on the walls in th mode of vibration are given by: The dimensionless convective pressure fu sloshing modes are shown in Figure C2. OBI 0 02 06 06 0.8 gy (2) (a) First Mode (b) Second Mode RECTANGULAR EARTIQUAR. FIG. C2.20 : DIM "ANK SIONLESS CONVECTIVE PRESSURES ON ALL PERPENDICULAR To DIRECTION ©Mass Analogy The spring mass analogy, defined previously for the vertical circular cylindrical tank, can also be used for rectangular tanks. For design purposes, the values of the equivalent masses, mg, s+ and thelr heights above the base ho, hy ... and hg, Ai ay be assumed to be the same as for the circular cylindrical tank with the half length, L, of the rectangular tank used instead of the circula radius, R. In most cases this approximation is expected to give ba shears and overturning moments within 18% of values from more exact theory. (o) HORIZONTAL ¢ In the analysis of horizontal cylindrical tanks it is convenient to consider motion either along the direction of the axis or transverse to the axis as shown in Figure (2.21. Approximate values for earthquake induced hydrodynamic pressures and forces in either the longitudinal or transverse direction can be obtained from solutions for the rectangular tank of equai dimension at the liquid level and in the direction of motion, and of a depth required to give equal liquid volume. This approximation gives sufficiently accurate solutions for design purposes over the range H/R between 0.5 and 1.5. Where H/R exceeds 1.6, the tank should be assumed to behave as if it were completely full of liquid. That is, the total mass of the contents acts as an impulsive mass at mid height of the tank. A more exact analysis is available {2.35}, for a nartially full motion transverse to the axis. The maximum impulsive pressure distribution for motion in this direction can be expressed as: For the case of the tank half full of liquid (H = R), the divensionless pressure function is given by: ag?) = “ye nel This pressure function is plotted in Figures C2.22. By integrating the pressure distribution the impulsive mass is evaluated to be: sin 2ne (2.18) (2n)Po = 19) 0.4m, 19) Because the pressures are in the radial direction, the forces acting on the cylinder pass through the centre of the circular section, ang both the impulsive and convective masses should be assumed to act at this point. Solutions for the convective pressures are not available in conve form for design. When the tank is approximately half full, (H = 8),Ku) Transverse FIG. C2.21 : NOMENCLATURE FOR HORIZONTAL AXIS CYLINDRICAL TANK Pressure antisymmetric about FIG. ON HORIZONTAL TRANSVERSEthe first sloshing mode mass, m,, can be evaluated by: = 0.6m, (£2.20) Expressions (C2.19) and (C2.20) are expected to be reasonable approximations for liquid levels given by H/R ranging from 0.8 to fact solutions are not available for other than special cases (e.5 spheres} in this category, It is believed that a reasonably € approximation to the seismic forces of the comparatively cormon class slevated tanks whose shape approximates part of an inverted cone ~ay obtained From the analysis of an equivalent cylindrical tank contsining ' the same volume of liquid as the real tank, and whose diameter same as the real tank at the free water level. Convective and Shears and heights of action for this equivalent cylinder should 5 ed to the real structural shape, when calculating stre: ' (e) FL IBLE VERTICAL CIRCULAR CYL i The flexibility of the tank wall increases the impulsive mode period, Because this period, for most tanks, is shorter than the value corresponding to the peak on the acceleration response spectrum, the ‘ result is generally an increase in the earthquake induced hydro stresses. e Although there are an infinite number of tank-liguid modes of vibration related to the wall flexibility, it is generally only : lowest of these modes that contributes significantly to the earthauake Stresses in the shell. In the lowest horizontal mode of vibrat tall tanks deform in an analogous manner to a deep vertical 5 beam, deforming in bending and shearing under the action of ' loading. Radial deformations are small with the cross section generally remaining circular. The lowest tank-liqui¢ mode for broad tanks is dominated by the radial deformation of the shell resulting in significant out-of-roundness of the cross section. Tank wall flexibility also increases the period of the vertical nodes of vibration. The lowest vertical mode is often referred to as 2 breathing mode and has a deformation pattern dominated by axially symmetric radial displacements. The results given below for pressures, spring-mass analog and periods of vibration are based on the assumption that th tank wall is unstiffened by ring girders or a roof structure assumption is expected to give satisfactory approximations practical cases where the top of the wall is st stiffened Convective Pressure Components In the analysis of flexible tanks, it is found that the tank Flexibility has only a small effect on the convective components of the hydrodynamic pressures, and for design purposes assumed to be the same as far rinid tanksImpulsive Pressure Componen The impulsive components mey de significantly increased by tark wall € or foundat'on flexibility end it is important to allow for these effects in design. For flexible tanks, with H/R ratios in the range of 0.25 to 1.5, the ¢ naximum impulsive component of hydrodynamic pressure on the tank wall can be approximated by p,lz08) = dglz) ¢,(Te) v2 %, cose (c2.2h) where qo(z) = dimensionless pressure fun o rigid tank cy(T~) = seismic horizontal force coefficient for the f" mace of the tank-liquid system : periad of vibration of first tank-Iiouid moce including the effects af soil structure interaction = 8, = mean radius of tank There is no simple expression for impulsive pressures where H/R ; exceedes about 1.5. For this case, expression (C2.21) gives é conservative values for the pressures below mid-height and uncer fstimates the pressures above mid-heicht. Typicel impulsive pressure distributions for the first mode of a flexible tank are compered with rigid tank impulsive pressures in Figure C2.23 for a range ratios. be /Re 0901 ty/Re 001 10 os 06 " z RIGID TANK. L ae cl 02 ° tay wi oa (after Veletsas ond Yang 19761 fatter Horaun cad Kousner, 19821 ty = wall thickness FIG. €2.23 : FIRST 1Pressures arising from the higher medes of the flexible tank will reduce the difference between the first mode pressures and the rigid tank impulsive pressures. Thus, for most practical purposes, the rigid tank impulsive pressure distribution can be used to provide a conservative approximation for the flexible tank case. The spring-mass analogy deserized belox provides a good approximation for the practical range of H/R ratios (including values exceeding 1.5} and generally provides sufficient information for design purposes. Spring-Mass Analogy The influence of the tank flexibility cn the base shears and overturning moments is conveniently analysed by using the sprirg-mass analogy (Fig. 2.2) proposed by Haroun and Housner (2.9). The mass ratios, me/mg and mo/2, may be obtained from the results of Haroun and Housner (2.9) and Haroun anc Ellaithy (2.8). These ratios ere plotted in Figure C2.24 as functions of the geometric ratio R./ The impulsive mass components are dependent on the retio of the wall thickness to radius ratio and the ratic of the wall and liquid unit weights. However, these parameters are of secondary importance and for practical design purposes their influence may be neglected. ‘ote that the impulsive mass component, m, shown in Figure C2.24 identical to the rigid tank impulsivé mass shown in Figure C2. The effective heights at which the two masses act, for the case excluding base pressure, are plotted in Figure C2.25. Heights for the case including base pressure are given in Figure (2.26. As for the case of the mass components, the effective heights are functians of other parameters that may be neglected for practical design purposes. sonents are given by ion 2.6. The base shears from the thres expressions (2.3), (2.4) and Pressures from Vertical Accelerations cal acceleration component ‘s The pressure, py(z}, from the vert ferential direction and is given uniformly distributed in the c approximately by: (1 -2/k) o(T,) 1, (€2.22) = seismic vertical force coefficient for the first vertical breathing mode of vibration. 5 period of vibration of first vertical breathing mode of vibration including soil-structure interaction. TANGULAR TANK ive Pressure Compor The tank flexibility has no significant influence on the conv: pressure component and thus tne rigid tank oressures can be use. design. for& R Mass Ratio Shown | ail 02 04 06 o8 10 20 30 40 Height to Radius Ratio, H/Rm PIG. C2.24 : IMPULSIVE MASS COMPONENTS FOR FLEXIBLE TANKS 260 phe 4 30 cy 7 HEIGHT-RADIUS RATIO IR FOR WALLS OF FIG. C2.25°@¢ Dimensionless Ratio Shown Height to Radius Ratio, H/Rm FIG. C2.26 : HEIGHTS OF IMPULSIVE MASSES FoR FLEXIBLE TANKS CINCLUDING PRESSURE ON BASE N(2-@} Impulsive Pressure Component [he impulsive pressure component is influenced by the wall flexii but because there are no analytical solutions for this case if is recommended that the rigid tank distrisutions be used. The order of error in this approximation is unknown but it is thought that it is a Teasonable approximation for relatively rigid walls where H/L is lese than 2.0, Spring-mass Analogy The spring-mass analogy described previously for the flexible cylingrical tank may be used for the rectangular case. onever, because of the lack of analytical solutions it is not possible to accurately define the tho inpulsive messes m, and m. To obtain an approximate solution it is recommended nat 4, be eSken os nes) co Mp. giving m= 0. Approximate values of thefequivalent masses ana their neight® above the base can be found from gee iene values for the rigid vertical cylindrical tank, with the half length, Lo? the rectangular tank used instead of tne ¢ reular radius R.(2.4.7 Granular Contents The assumption of the total mass af the contents being rigidly connected to the tank shell gives a reasonably good aoproximation to the total horizontal earthquake force when the H/R ratio is greater than 1.0, It gives conservative results for 4/R less than 1.0. A more exact approximation for the force, and details sures from granular materials are given by Trahair et 2 2.6 BASE _SHEARS (a) Rigid Tanks Expressions (C2.8) to (C2.13) give the base shears and moments from the two mass components (impulsive and convective) of the spring-mass © analogy. ¢ For design purposes, an estimate of the maximum base shear and moments from the combined impulsive and convective pressures can be obtained by using the square root of the sum of the squares rule as given in Section 3. (b) Flexible Tanks e Expressions (2.3), (2.4), (2.5), and (2.6), » (2.8), give the base shears and moments respectively from the three mass components the spring-mass analogy for a flexible tank. Only the first mode impulsive mass components are used in thi tank spring-mess analogy described in Section (2.8 above. Ti lexib contributions of higher modes to the base moments are cenerally small for most practical cases. For relatively sierder tanks, the higher ‘ modes may add 2 significant contribution to the base sear. ‘owever for H/R ratios up to 3.0 the higher impulsive node contributions to the total base shear are unlikely to exceed 19%. Because of the neglect of higher mode effects, a conservative approach should be adopted for combining the components to estimate the maxim total base shear and moments. It is reconmended that the impulsive mass components be combined by taking the absolute sum and that the square root of the sum of the squares rule be used for adding the total impulsive component to the convective component (see Section 3). 2.7 NATURAL PERIODS (a) Rigid Vertical Circular Cylindrical Tank The period of vibration of the jth convective or sloshing mode is given by: 2: RG (€2.23) Dy tank 0; WR ‘iec The dimensionless geriod, T ¥g/R, is plotted for the first two modes in Figure C2.27. Period Dimensionless 2 3 Height to Radius Ratio Hip FIG. C2.27 : DIMENSIONLESS PERIODS FOR CONVECTIVE MODES Ee (b) Rigid Rectangular Tank The period of vibration of the jth convective or sloshing mode is given by: 2 22 ATG (02.24) 3 fa tanh (a5 RAL J= x/2, 32/2, 52/2, dimens yoni is plotted in period, T G/L, for the gure C2.27. rst two modes of vibration Ce) Other Rigid Tanks Earthquake induced hydrodynamic pressure distributions for tanks, shapes other than vertical circular cylindrical and rectangu not been published in suitable form for design. Approxina for tanks of other shapes can be undertaken using the results gf previously for the vertical circular cylindrical and rectangular tanks. It is helpful in these analyses to have an approxima estimate of the first sloshing mode frequency or period. In Figure (2.28 the first sloshing mode dizensionless frequency, f, /R/g or ¥7g, is compared for the following shzpes . Vertical circular cylinder - Vertical circular cylinder with spherical bottom Horizontal circular cylinder, notion along axis Horizontal circular cylinder, notion transverse to axis Rectangular - Spherical Most of the information plotted in Figure 2.28 has been obtained fron Abramson’, Results of analytical and experimental studies for more complex shapes are also given in this reference. (d) Flexible Vertical Circul The period of vibration mode is given by: nit weight of contained I icute Young's modulus for tank material pression (2.25 gives the first tank-liguid horizontal mode period for a roofless tank with uniform weil thickness, ty, Poisson's ratio 0.3 and full with liquid. The expression was developed for the case be used for other tank tank of a steel tank filled with water but m materials and liquids when the mass of comparison to the mass of the liquid. approximation for most practical cases. 5 relatively smail in ely to give 2 goodFIG. PIG. 62.28 oat 2.29 tei cylioaer, | Vertical eytinger spherical bo Vertical eylinger a Redtangular (length = 24) Horizontal eyuineer, Longifusine_( tengtn Sai HR =05 BIR on Hit FOR RIGID TANKS OF VARIOUS SHAPES ACR RIGID TANKS OF VARIOUS SHAPES 4 7 18 ve MOGHT TO RAOIUS RATIO, H/Ry + PERIOD COEFFICIENT ky FOR FIRST. HORIZONTAL TANK - EIQUID MODE (2.9) + BIMENSIONLESS FIRST CONVECTIVE MODE FREQUENCY74 The first node period of tanks of non-uniform wall ess may be determined by using the average thickness. This will give @ reasonably good approximation up to H/R ratios of about 4.0. this value the flexural deformation dominates tne response and exact period may be obtained by using an average of the wall thickness aver the lower half of the tank. The influence of the roof mass on the perid Ounkeriey's method as outlined by Veletsos | applications the roof mass may be neglected. Yertical Mode The period of vibration of the first tank-liquid vertical ( mode of vibration is given by: é where k, * period coefficient defined in Figure 2.30. | Expression C2.26 gives the first tank-liquid vertical mode o o | a roofless tank with uniform wall thickness, Poisson's ratio of 0.3 and full with liquid. As for the case of expression C2.25, far the horizontal period, it has been developed for a steel tank filled with water but gives good approximations for other materials and liquids when the mass of the tank is relatively small in comparison to the e mass of the liq hy Period Cot HR PIG. C230 : PERIOD COBPFICIENT ky FOR FIRST VERTICAL TANK —EIQUID MODE (2.30)( ' | 75 fe} Flexible Rectangular Tank Horizontal Mode The period of vibration of the first impulsive tank-liquid horizontal mode is given approximately by: 1, = 2s GTS (2.27) deflection of the tank wall on the vertical centre-line and at the height of the impulsive mass, m,, when loaded by @ uniformly distributed load in the direction of ground motion and of magnitude m,o/4BH where 28 = tank width perpendicular to direction of loading, The impulsive mass, mp, can be obtained from the equivalent cylindrical tank results and should include the wall mass. For tanks without roofs the deflection d, may be calculated assuming the wall to be free at the top and fixed’on the other three sides. Table C2.1, reproduced from Timoshenko and Woinowsky-Krieger (2.27), can be used to calculate d, for this case. TABLE C2.1 OEFLECTIONS, BENDING MOMENTS, AND REACTIONS OF UNIFORMLY QOADED RECTANGULAR PLATE WITH THREE EDGES BUILT IN Ai FOURTH EDGE FREE (FROM TIMOSHENKO AND WOINOWSKY-KRIEGER) {_zrar-t | saayenre 7 face tee iat | pele gee et | vans | oma | ous, | exer To. caw | Touts | vas va | Gos | our } ovis | ona | oy | 22% ans | Zouss | ote os | vu | vous | oon | aust | uo | Wound van | S003 | ole vo | d.ousss Dob | O.wvsy | O.usI7 | 0.0134 | —o oMsy, 0.4us | 0.0510 | 0.401 Us | ean | gone | ota son | To ter ois | Zo omy | os qe w= plate deflection vertical Mode There is no simple method for calculating the periods of the vertical modes of vibration for a flexible rectangular tank. For this case, use should be made of the results for the equivalent cylindrical tank.76 (f) Soi] Structure Interaction Because of soil-structure interaction effects, a stiff tank resting on a soft sot] may have an earthquake response significantly different from the case of a tank on a rigid foundation. Soil-structure interaction may lengthen the period of vibration of the impulsive modes and also increase the damping. Generally an increase in the jmpulsive mode periods will result in an increase in response but the additional damping will tend to counteract this effect. Strictly speaking, soil-structure interac impulsive and convective periods but the modes are small and can be neglected. on will change both the fects on the convective Jennings and Bielak (2.12) showed that for single-storey buildings, soil-structure interaction under earthquake Toads could be conveniently considered by defining an equivalent single degree-of-freedom system. The method can be readily extended to the case of tanks directly supported on the ground. Although the original work was for the case of a structure on a rigid foundation base, Testing on elastic soil, the method has more recently deen extended to include the effects of foundation embedment and material camping the soil (Bielak (2.3,2.4)). The theory formulated by Jennings and Bielak has been simplified by Veletsos (2.30) and presented with a notation more suitable for design applications. A summary of this simplified approach is given below. Horizontal Periods of Tank-Foundation The periods of vibration for the first two impulsive modes of the tank-foundation system are given by: ay, the horizontal translational stiffness of the foundation the rocking stiffness of the foundationv, * Poisson's ratio for the soil &, = shear modulus for the soil Ry = radius of the foundation a,+ @, * dimensionless factors that convert the static stiffness values to dynamic values h h, = height of impulsive mass, r r my = mass of base Expression (C2,28) gives the period of the first tank-liquid soil-structure interaction mode of the spring-mass analoay. Expression (C2.29) gives an approximation to the second mode period for the flexible tank on a flexible foundation. In most practical cases m, and m, are small and Ty can be taken as zero. For a rigid tank T, = 0 and Kr becomes infinite. For this case, expression (C2.29) may be used to give the first node period by replacing m, with mg and h, with ho The dimensionless function of the frequency parameter, a, defined by ae (ca. Y&oJe, the shear wave velocity of the soil oe, = the soil mass density T, = T, or Tg for expressions (C2.28) and (C2.29) 1 réspect¥vety. For most design applications the factors o and as may be taken as 1,0, that is the foundation stiffnesses may be taken as the static values. However, for larga tanks on soft soils a, may be Significantly less than 1.0 for the impulsive modés of vioration. Earthquake Response The earthquake response of the equivalent single degree-of-freedon oscillators, with periods of vibration Ty and Tp, representing the impulsive modes of tank-feundation system can be obtained from the previously given acceleration response spectra. Vertical Periods of Tank-Fourdation System The vertical period of vibration of a flexible tank including soil-structure interaction is given by: ¢ “height of impulsive mass, me, above the tank foundation » above the tank foundation ctors e, and ¢, are plotted in Figure (2.31 as a 30)78 FIG. C2.31 : PERIOD COEFFICIENTS FOR RESPONSE OF TANK - FOUNDATION SYSTEM@ 79 where dimensionless factor to convert the static stiffness value to the dynamic value The vertical period of vibration of a rigid tank including soil-structure interaction is given by: 1, = on IR (c2.32) > Wy where om, =m, +m, = total mass of tank including base, support system and structural foundation The dimensionless factor a, is plotted in Figure ¢2.31. The expression for the period of the flexible tank, (C2.31), has been simplified by neglecting the mass of the tank base and any structural foundation. To include these masses requires a two mass springonass analogy which adds complexity that is unnecessary for many design applications. If the base and foundation are unusually heavy a more rigorous analysis may be justified. 2.8 DAMPING Unless @ more rigorous analysis is undertaken the damping values given in Table (2.2 may be used for design. TABLE ¢: DAMPING FACTOR, < INK DESCRIPTION | % DAMPING % DAMPING SOFT SOIL | FIRM SOIL AND Rock jeg) ce ey (ces ei | Horizontal | Vertical T | Direction [irestion Direction | Direction i | | | | heats tanks and si 7.5 flexible tanks | lanchorea to prevent | luplift | Unanchored flexible tanks where significant| luplift is expected for horizontal earthquake loads 2 5 is 15 10 | 5 | I60 05 08 os bh 4 FIG. C2.32 : DAMPING COEFFICIENTS FOR RESPONSE OP TANK - FOUNDATION SYSTEMat The damping values given are estimates for typical tank systems, and are total values including material damping, in both the tank and soil foundation, as well as radiation damping from the foundation. uhere more exact estimates of damping are required for major tank installations the following approximate expressions from work of Bielak (2.4) and Veletsos (2.30) can be used. The effective damping ratio of the tank-foundation system in the Flexible mass impulsive mode is given by: +n (£2.33) (15/7) = foundation damping including soil material and radiation damping where material damping in the tank system above the foundation for the impulsive mode The foundation dam ing factor is given by: where 6, dimensionless factors for translation and rocking damping respectively Values of 8, and 8, are plotted as a function of the frequency ratio, a, in Figuré c2.32° These damping factors only include the effects of radiation damping and not the soil material damping. Oamping factors that include both radiation and soil material damping can be obtained fron Veletsos and verbic (2.32) and Veletsos and Nair (2.31). The soi] damping for the case of a rigid rank can also be obtained from expression (C2.34) by putting mp = my and he = hg. 2.10 CONVECTIVE WAVE EIGHT (a) Vertical Circular Cylindrical Tanks The maximum vertical displacement of the convective sloshing wave from the at rest level of the liquid is given by: 4 R Vio.88 oy(Ty)1? + [0.07 oA(TA)}2 +... (c2 ‘max 2 The above expression is basea on the assumption that the vertical Liquid displacement can be calculated approximately by adding the first two sloshing mode components using the square root of the sum of tne squares rule.gular The maximum vertical displacement of the convective sloshing wave from the at-rest level of the liquid for rectangular tanks is given 6: grax 7b oy(T) (62.38) This expression only includes the first mode displacement but generally higher mode effects can be neglected. {c) Roof Pressures If the freeboard is less than the height of the convective (sloshing) waves, then hydrodynamic pressures will be generated on the roof structure, The estimation of these pressures requires the solut a complex non-linear wave problem that involves geometric par: Such as the slope of the roof and the height of the roof abov Viguid surface. Wave surface when undisturbed by roof (Cosine curve approximately) Static liquid surface A awe meet aeteesh [estnesnigsnnessaas| Run up height from Ay = Ap Pressure on roof at radius r Peo = Yy be FIG. C2.33 : APPROXIMATE METHOD FOR ESTIMATING BUOYANCY PRESSURE ON TANK ROOF83 There is little published information that provides useful design pformation on roof pressure loading. Some guidance can be obtained from information in Kobayashi (2.17) and Leonard ef ai (2.18) a The kote! pressure on the roof will be the result of both wave impact ! loading ana a varying buoyancy loading that results from the wave peak Turning up the roof slope. The total pressure is given bys ep te 2 (c2.37) Pots ua where py = buoyancy pressure = wave impact coefficient wave particle velocity B e a ph ePProxinate graphical method of evaluating p, is shown in Figure 2.33. An upper bound to ¢, is apparenti? asot 20 and the approximate maximum value of & caf be obtained fron i # dae (€2,38} 2.11 ENBEDDED TANKS Soilostructure interaction effects for embedded tanks are generally complex and require special studies. Earthquake induced sort Ore hee re anger found from rigid wall solutions given by wood (2.38). The cartnquake Induced pressure distribution shown in Figure C> a4 wit Provide 2 conservative estimate for most embedded tanks, the tn the coi p en” 0° Mot include pressures arising fron gravity forces in the soil. Ideatised pressure distribution mbedment depth, He Tank base Ls 0,5 Ch 101 Ys Hy FIG. C2.34 : COMPONENT OF EARTHQUAKE SOTL PRESSURE ON EMBEDDED TANK WALLIn Fig. €2.34, OPog is given apg = where ys and of = x a4 the increment in at-rest earth force due to earthquake, by CylOdrgHye (02.33) unit weight of soil embeddnent depth increment in at-rest soil pressure due to earthquake. 2,12 ELEVATED _TANKS A satisfactory method of analysing most elevated tanks i two mass model suggested by Housner (2.10) and shown in to use the igure (2.3 Pann mass brnw—iny br Impulsive "o. mass FST (a) Hydrodynamic Masses of Elevated Tank m1 | impulsive + img "| Structure! masses k : s ine FIG. €2.35 (d) Two Mass Mode! + TWO. ASS MODEL FOR ELEVATED5 Convective end impulsive masses representing the hydrodynamic pressures may te evaluated using the riethods given previausly for rigid tanks. One of the masses of the model is equal to the convective mass and the other is equal to the sum of the fluid impulsive mass plus the mass of the tank and supporting structure The Tocation of this second mass is assuned to be at the centre of gravity Of its components including the fluid impulsive mass. The stiffness of the convective mass spring may be estimated from: 2,2 Kyra ty ay (2.40) The ‘spring connecting the second (impulsive) mass to the ground is assioned a stiffness, Kz, equal to that of the support scructure horizontal force applied at the same height as the mess. The earthquake response of the two mass model can be calculated using Standard structural dyanmics and response spectra methods. Camping for the mode involving predominantly liquid sloshing should be taken as 0.5%, In the other vibration mode, the damping should be evaluated from Table (2.2, expressions in Section 2.8 and the expressions given below for equivalent damping in a ductile supporting structure. the Square root of the sum of the squares rule should be used for combining the modal responses. A simplified analysis procedure has been suggested by Housner (2.10), this apyroach the two masses are assumed to be uncoupled and the earthquake forces on the supports are estimated by considering two Separate single degree-of-freedom systems: one representing the sloshing of the convective mass in a tank assumed to be rigidly connected to the ground, and the other representing the impulsive plus Structure mass behaving as an inverted pendulum with a stiffness equal to that of the support structure. If the periods of these two systems differ by more than a factor of 2.5, then it is likely that the mecnod will give a satisfactory approximation for design purposes. If the supporting structure is detailed to have significant ductility then ¥t may be designed to respond beyond the yield level under the design earthquake. In this case, the acceleration response of the impulsive mass is reduced by the absorption of energy in the yielding system. A convenient method of allowing for ductility effects is to use the following expressions given by Iwan (2.11) for defining an equivalent elastic system. TefT = 1+ 0.12 (u = 170-98 (c2.41) 25+ 5.9 ty - 13997 (€2.42) where period of vibration for equivalent linear system T = period of vibration based on stiffness at yield viscous damping for equivalent linear system, percent of critical 5 + viscous damping estimated for response up to yield, percent of critical In86 ce = displacement ductility factor Numerical studies reported by Iwan showed that the above expression for the equivalent linear system, over at least the mid period range of 0.4 to 4.0 sec, had smaller errors than other methods generally used to allow for ductility effects, The equivalent linear system is easy to apply as direct use can be made of response spectra published for elastic structures. REFERENCES 2.1 Abramson, H.N., ed (1966), "Dynamic behaviour of liquids in moving containers", NASA SP-106. 2.2 Berrill, J.B., Priestley, M.J.N. and Peek, R. (1981), "Further comments on seismic design loads for bridges", Sull of NZ Nat Soc for Earthquake Eng, vol 14, No. 1, March. 2.3 Bielak, J. (1975), “Dynamic behaviour of structures with embedded foundations", Earth Eng and Struct Dynamics. 2.4 Bielak, J. (1976), "Modal analysis for building-soil interacton", Journal of the Engineering Mechanics Division, ASCE, Yol 102, No. EMS. 2.5 Clark, R.H., Dibble, R.R., Fyfe, H.E., Lensen, G.J. and 6 Suggate, R.P. (1965), “Tectonic and earthquake risk zoning", Trans Royal Soc of NZ, vol 1, No. 10, April. 2.6 Fujino, Y. and Ang, A.0. (1982), “Prediction of seismic response of long-period structures", J of the Struct Civ, ASCE, vol 108, No. ST7. 27 Gazetas, G. (1983), "Analysis of machine foundation vibrations: state of the art", J Soil Oynamics and Earthquake Eng, Yol 2, No. 1. 2.8 Haroun, M.A. and Ellaithy, H.M. (1985), "Model for flexible tanks undergoing rocking", J of Eng Mech, ASCE, Yo! 111, No. 2, February. 2.9 Haroun, M.A. and Housner, G.W. (1981), “Seismic design of < liquid’ storage tanks", J'of the Tech Counc of ASCE, vol 107, No. Ch. 2.10 Housner, G.W. (1963), "The dynamic behaviour of water tanks", Bull of Seis Soc of America, Vol 53, No. 2, February. 2.11 Iwan, W.D. (1980), Estimating inelastic response spectra from elastic spectra”, Earthquake Eng and Struct Oyn, vol 8. 2.12 Jennings, P.C. and Bielak, J. (1973), "Dynamics of building-soil interaction", Bulletin of the Seismological Society of America, Vol 63.2.14 2.15 2.17 2.18 2.20 2.21 2.22 2.23 2.24 2.28 2.26 Joyner, W.8. and Boore, D.M. (1981), “Peak horizontal acceleration and velocity from the strong motion records including records fron the 1979 Imperial Valley, California, earthquake", Bull of Seis Soc of America, vol Ti, No. 6, December. Voyner, W.B. and Boore, D.M. (1982a), “Estimation of response ~ spectral values as functions of magnitude, distance and site conditions", Open File Rep 82-881, US Geol Svy, Menlo Park. Joyner, W.8. and Boore, D.M. (1982b), “Prediction of earthquake response spectra", Open File Rep 62-977, US Geol Svy, Menlo Park. Kanamori, H. (1979), "A semi-empirical approach to prediction of long-period ground motions from great earthquakes", Sul! of Seis Soc of America, vol 69, No, 6, p 1645-1570. Kobayashi, N. (1980), "Impulsive pressure acting on the tank roofs caused by sloshing liquid", Proc 7th world Conf Zerth Eng, Istanbul. Leonard, J.W., Garrison, C.J. and Hudspeth, R.T. (1981), "Determination fluid forces on structures: a review", Proc ASCE, Vol 107, No. ST6, June. Matuschka, T. (1980), "Assessment of seismic hazards in New Zealand", Rep 222, Dept of Civ Eng, Univ of Auckland. Mulholland, W.M. (1982), "Estimation of design earthquake motions for New Zealand", Rep 82/9, Dept of Civ Eng, Univ of Canterbury. Norton, J.A., Gillies, A.G. and Edmonds, F.D. (1982), "Recommendations for the seismic design of petrochemical plants", Bull of the NZ Soc for Earthquake Eng, val 15, No. 3, September, Peek, R. (1980), “Estimation of seismic risk for New Zealand", Rep 80/21, Dept of Civ Eng, Univ of Auckland. Priestley, M.J.N, and Park, R. (1984), "Strength and ductility of bridge substructures”, RRU Bull 71, National Roads Board, Hell ington. Smith, W.D. and Berryman, K.R. (1983), "Revised estimates of earthquake hazard in New Zealand", Bull of NZ Natl Soc for Earthquake Eng, vol 16, No. 4, December, Smith, W.0. and Lensen, G. (1981), "Sesimicity model for New Zealand", Unpublished working paper, Seis Risk Subcommittee, Standards Assoc of NZ. Standards Association of NZ (1986), 02 4203 "General structural design and design loadings for buildings".2.27 2.28 2.29 2.30 2.31 2.32 2.33 2.34 2.35 88 Timoshenko, S. and Woinowsky-krieger, 5. (1959), "Theory of plates and shells", McGraw Hill, Secon Edition. Trahair N.S., Abel A., Ansourian, P., Irvine, H.M. and Rotter, J.M. (1983), "Structural design of steel bins for bulk solids, Aust Ins of Steel Const. Trifunac, M.0. and Anderson, J.G. (1977), "Preliminary empirical models for scaling absolute acceleration spectra", Rep CE 77-03, Dept of Civ Eng, Univ of Southern California. Veletsos, A.S. (1984), "Sesimic response and design of liquid storage tanks", published in "Guidelines for the seismic design of oi] and gas pipeline systems", ASCE, New York. Veletsos, A.S. and Nair, ¥.V.D, (1975), “Sesimic interaction of structures on hysteretic foundations", Journal of the Structural Division, ASCE, Yol 101, No. STI. Veletsos, A-E. and Verbic, 8. (1973), ‘Vibrations of viscoelastic foundations", Earthquake Eng and Structural Dynamics, Vol 2. Westermo, 8.0., Anderson, J.G., Trifunac, M.D. and Dravinski, M, (1980), " Sesimic risk tables for pseudo relative velocity spectra in regions with shallow seimicity” Rep CE 80-01, Dept of Civ Eng, Univ of Southern California. Wood, J.H. (1973), “Earthquake-induced soi] pressui structures", Rep EERL~73-05, Earth Eng Res Lab, Cali Technology, Pasadena. Werner, P.M., and Sundquist, K.J. (1949), "On hydrody earthquake effects", Trans Anerican Geophys Union, ¥ No 5, OctoberoC COMMENTARY ON SECTION THREE Design Actions 3.1 scope The methods of analysis suggested in this section are approximate and conservative. The designer may use more sophisticated methods of analysis if in his judgement they are required. For a vertical cylindrical tank the stresses required for design are {hose resulting from: hoop forces and vertical bending monents (described in Section 3.3); axial rembrane stress and base Plate tension (described in Section 3.4) and shear stress ia°the walls (described in Section 3.5). For some tanks other than vertices cylinders, different resisting nechanisms may predominate and co 214 fbove stress resultants may not act. For example, hoop forces would be minor in a rectangular tank but horizontal bending ronents nay be significant. Figure 3.1 illustrates for a vertical cylindrical: tenk and a rectangular tank the action of the significant stress resultancs C3.2 METHOD OF COMBINING ACTIONS. The horizontal impulsive components {rigid and flexible) should be added to provide the maximum horizontal impulsive response. However, because of the low probability of coincident response of maximum horizontal impulsive, convective, and vertical impulsive aecions, t SRSS method of combining these actions 1s recommended. Thus. tor example, the resultant hoop force N, for a flexible vertical axis cylindrical tank would be + Nog * maximum horizontal impulsive hoop fol where Nei = Nop and Ng, © maximum horizontal rigid-impulsive hoop force Ngp = maximum horizontal flexible-impulsive hoop force SF (see C1. 2.4.2) Ng) = maximum Ist-mode convective hoop force Ngy = maximum vertical impulsive hoop force Stresses may be found in similar fashion. for example, the components of vertical surface stresses induced at height z by seismic sction will be90 Axial membrene Hoop force = Vertical bending tomes lin Shear fores (ol CYLINORICAL TANK : y Nz + Axia! membrane Paes My = Horizontal beng eae Ee = fr foree fin LI St {x Nay= Shear force fir eas My & Sy, 6 yA wt Ne |b] RECTANGULAR TANK FIC. C3.1 : STRESS RESULTANTS FoR DESIGN OF 7A (impulsive ri (impulsive flexibie) (ist mode convective) (vertical impulsive)a1 and the resultant seismic stress would be 3.3 €3.3.1 Rigorous Analysis - axisymmetric Tanks with 2 Constant Wall Thickness Section 2.4 defines ti @ pressure distributions for impulsive a Myective modes of response. ihe pressure distributions vary up che rtical axis and around the circumference. However, Priestley has shown that where the pressure varies only slowly in the circu: ential direction, as is the case for seismic pressure distributic the stress distribution at the cylinder generator of saxinum ore: (9 = 0) can be calculated without significant loss of accuracy by assuming that the entire tank is subjected to an axisymmetric dr Gistribution equal to the pressure distribution at g = 0. Consequent axisymmetric structural simulations are possible, which greatly reduce the coroutational effort in calculating critical design forces example, the equations governing axisymmetric shel! behaviour re. that of a beam on an elastic foundation (GEF) (3.2). Priestley describes a simple frane analogy of the SEF method cagable of ca Stresses and deflections for both cylindrical and non-cylindrical with uniform or varying wall thickness. This method ras been u Onjunccion with the ectual pressure distribution of Section 2.4, produce dimensionless design charts for hoop forces and vertical ending moments, which are presented in an Appendix to this section Walues of hoop force and bending moment are given for hydrostatic and nence vertical acceleration), 1st mode convective, and impulsive rvgic pressure distributions, for a range of H:R and R:t ratios. 3.2 Approximate Ani Aickness ysis - Cylindrical Tanks with a Variable A reasonable approximation to behaviour where wall thickness t does no vary excessively is to use the design charts presented in the Agpendix, basing stresses on the local R/t ratio. where greater accuracy is required, the wall should be analysed under the appropriate pressure cistribution for impulsive, and convective pressure, in Section C2.4 Clause 3.3.1 permits the use of an equivalent linear pressure ai Qution with height, rather than the curved distributions shawn in Section 2.6. In calculating the equivalent linear pressure districutiPE Pe > H 4 q . 7 nh = + ‘ ip iad € ° a t : ACTUAL EQUIVALENT UNIFORM = HYDROST: FIG. C3.2 : EQUIVALENT LINEAR DISTRIBUTION OF PRESSURE ON ‘tans Diz)» it is convenient to work directly actlon h defined fran the Sering-ass an cylindrical tank, the maximum force/unit ci occurs at 6 = 0 is given by aes 03.3 The linear equivalent pressure distribution is then calcul, statics. Thus in Fig. (3.2, en< where gy and gn are equivalent uniform and hydrostatic pressure distr! butions. Taking moments about the base of the wall ans, wet cng ee Simultaneous solution of Eqns. C3.4 and hence the uniform and hydrostatic pres noted that computational effort can be considersdly reduced resultant shears Qp» Qe, Qj. and the resultant moments Hr, are each combined according to the requirements of Section 3. a resultant pressure distribution can be defined thus requiring only 2 2 structural analysis. Non-Cylindrical Axisymetric Tanks Section C2.4(f) suggests an approximate method for calculati forces for axisymmetric tanks whose shape zoproximates an i Equations ¢3.3 and C3.4 can be used to calculate the equiva pressure distribution on the walls rylinder R. However, the actual horizontal pressure to be &; tank walls should be modified to@¢ 93 where ®(z) is the tank radius at height z. This modification is needed to ensure’ the resultant force on the tank, and its moment about the dase, remain correct. Structural analysis of these tanks can conveniently be carried out by the method given in Ref. 3 Rectangular Tanks roun (3.3) has calculated an exact solution for bending ronments by an impulsive pressure distribution in the walls of vertica gular tanks. “An acceptable approximation to behaviour, including convective components, is to use a linear equivalent pressure distribution as snown in Fig. C3.2 defined by Eons. C3.4 and €3.5. In this case, the shear per unit length, q, is given by ote (63.7) induc axis rec ig the tank width perpendicular to the ground acceleration. Analyses of rectangular plates with appropriate boundary conditions, d to uniform or hydrostatic pressure distributions are available in standard texts on plates (e.g. 3.4) ca TURNING E! (3.4.1 Non-Uplifting Tanks - Axial Stresses A tank that is not securely fastened to the foundation must be initially checked to see whether it will uplift under the action of the design forces. The mechanical analogue described in Section 2 is one possible means of proceeding with this check. This gives for the overturning monent, the expression: aoy(T p91 n(Te)o The symbols in Eqn. C3.8 are defined in Section 2. Note that in Eqn. 03.8 the impulsive-rigidand_impul Flexible moments are added directly. for a rigid tank, Tp=To» and the impulsive term reduces to Po Sh Because the interaction of the vertical component of the earthquake motion with the uplifting mechanism of a tank is not understood, the usual gravitational constant g has been chosen for the calculation tank wefoht. Thus, for an unanchored circular tank, the restoring moment Np is given by im, + MAR (c3.9) ‘or non-uplifting tanks, the additional stresses and displacements resulting from the overturning forces may de calculated by considering the tank to be a vertical cantilever. Thus, at the base4 aa where {R+t) is the mean radius of the tank and wall, and t is the wall thickness at the dase. For a rectangular tank where 2. is the tank length in the direction of ground 28 is the tank width. Displacements for a cylindrical tank, the maximum d edge, resulting from overturning effe: action displacément component}, is given by 2 ; ieee | aaa Mina. CylT els xg} (C3.13 where 8, = modal particioation factor of the 7 = associated with the flexibility of the shell ( CylT~) = seismic horizontal for the fundamental period 8 coefficient corresponding 20 (Section 2.2) H! 05 10 415 20 25 30 35 HEIGHT-RADIUS RATIO (H/R) PIG. 3.3: PART. CTOR, By, (3.12)3.3.2 g Tanks If Mor calculated by Eqn. C3.8 exceeds Mp calculated by Eon. (3.9, and fe tank is unanchored, uplift will occur ne primary effect of uplift increase the compressive axial stress in the shell. addition, distortion of ¢i be considered (3.5 J] and tension Forces in the tank base may need to 7). fa Yertical cylindrical tanks nderstood the up- the mechanism of tank uplift is complex and not completely To describe it fully, the e of; large displacements lifting base, yielding of the base-shell joint, membrane forces in Ving of the + imperfections of the shell geometry, and foundation flexibility need to be included. The calculation of design forces on a rocking tank hav tricted to a quasi-stazic rather than dynamic methods (3.5-3.3) Vowing descr modified version of the method presented by Clough (3.5). js recommended es it best describes a believable rocking mechanism for a cylindrical tank. As part o base uplifts, the weight of liquid above the up} portion assists in stabili ing the tank. The higher Mor fs, the greater the extent of uplift that will be required to stabilise the tank Thus, in Fig. C3.4, the overturning moment, Myr is resisted by the action of three forces; Ne, Wp, and W forning Mp =r) Mp = Wy KR + We where W = total weight of the fluid weight of the fluid supported directly by the foundation over the area that does not uplift (radius r} weight of shell and roof Ws = W + My ~ We = compression reaction at shell base RY is the radius of the tank, 9* is the half angle which 1 bese in contact wit kR = distance from c tank centreline, ea 8 Uplifted crescent rn i wd (R-r} EVATION FIG. C3.4 : RESTORING Fon NAN UPLIFTING CIRCULAR TANKMy the resisting moment 2 (ij guessing u # r/R wy ting 6* = arctan (y25) calculating k ener Alternatively or as a useful s Fig. C3.5 may be used. tarting point in the iterative oroce Note that the method suggested assumes that the overturn ren js not affected by the degree of uolift. In fact, uplift can be expected to modify the effective flexibility of the impulsive moce, fnereasing the period, More precise information is available in recent research report by Peek et al (3.9). Peek et al also rote that Stability against overturaing of unanchored tanks can de greatly raved by 'pre-uplifting’ the tank. This involves supporting of the tank on a ring such that the base is uplifted for of the diameter, right round the tank. The fluid pressure on uplifted portion of the base provides an additiona! compressive reaction on the ring. This célays the onset of uplift, aad reduces Stresses, relative to @ tank without ‘pre-uplift’, upliftin occurs. More details are eveilabie in Ref, 3.10. 2 rin part Axial_membrane sti un sins) The maximum axial stress in tre shell is computed as Re where © i on stiffness factor el. oundation 20. le foundati = ht hy) oe he?) Me eM + My Wp We cr 5) calibration factor (suggested value The method described by Clough (3.5) is known to be non-conservative (3.11). The suggested calibration factor has been deriv: results of Clough with those obtained from experinents ( Radial membrane stress in base An estimate of the menbrane stress fpp in the base plate as a result oF uplift is derived by Cambra (3. €)Half-Angle of Contact Arc 6", («1 Radions} r/R OS Region, Radius Ratio of Bottom-Contact 0 97 0 0.5 1.0 fensionless Overturning Tendency, M/RW 0 0.5 1.0 Dims FIG. C3.5 : where ensioniess Overturning Tendency, M/RW DESIGN AID FOR OVERTURNING TENDENCY FOR UNANCHORED CIRCULAR TANKS (c3.1Half-Angle of Contact Arc Radius Ratio of Bottom-Contact FIG. _ 7 a 0.4 0.3 0.2 ©", («Tt Radians} Out a5 1.0 Dimensionless Overturning Tendency, M/RW 0 0.5 1.0 Dimensionless Overturning Tendency, M/RW ©3.5 : DESIGN AID FOR OVERTURNING TENDENCY FOR UNANCHORED CIRCULAR TANKS = thickness of base plate in uplift region = E/(1 - vy“) = Youngs Modulus93 estimate de of the s lift may be ob ig a modified version of a formula derived by Canbra uplift is, has be hall j {o) Res ‘tangular tanks There are no recorded ‘studies of tank up! than that for vertical cylindrical tanks. Given in the f Suggested procedure for the analysis of uslifting rectangular tanks which is based uson the principles underly ing the previous formulae for cylindrical tanks. In F t mechanism is y which has owin: 3. Hustrated with the length of the base, lifted off the foundation. en, 2L _-Uplifted 1 region ELEVATION FIG. C3.6 : RESTOR FORCES UPLIFTING RECTANGULAR TA!99 With this mechanism, taking moments about the con the resisting moment is given by eee pz WL twL (1 - 2°) Thus for a rectangular shaped tank u can be solved for dire 1 ik 4 ue fi-(Ge- me ae (4) Axial membrane stress in shell Mat -d) 2h (eae ereceeae ce (11) Longitudinal membrane stress in base a? (a. yy?) 43 fy Po | 63.22) Sk - (iii) Uplift of sheli This is given by Eqn. C3.17. In this case however, the upli is given by sb (l= x) b AN] terms used in Eqns. €3.20 and (3.21 are defi Eqns. €3.15 anid C3.16. 3.5 SHEAR TRAN (3.5.1 Roof to wall The total shear force to be, designed for will be: (a) for a rigid tank (cf. fqn. 2.5, with Tp = Q* 6, (T.) mo (3.2 (b) for a flexible tank (cf. Eqn. 2.5 and Eqn. €3.3) 05, c, Gl ms (c3.28) where m, is the mass of the roof. 3.5.2 Wall to Foundation Sections 2.5 and 2.6 provide expressions for the seismic base shea components, and direction for the inclusion of shall and roof ine| 100 he shear stresses that result from these forces can be calculated for in the following manner: (a) vertical cylindrical tanks it may be conservatively assumed that the total seismic component of base shear, Q, is taken by menbrane action. In this case the cires ential shear flow to be designed for is given by sind R 4 2nd (Ont 9: {For broad thick-walled tanks the shear stresses resulting om vertical bending may be significant. The maximum shear force/unt circumferential length is maximum at 8 = 0 and is given by Re (03.26) zn a, where V, is the slope calculated at z’ = 0 from the charts in ¢ in the Appendix. The total base shear taken by vertical bending is given by For the cases where Jy is not insignificant, the circumferent flow may be calculated by using Eqn. ¢3.25, where = OO £3.28) () Rectangular In the central region of walls perpendicular to the direction of ground acceleration it may be conservatively assured that the pressures are carried entirely through vertical bending action. For this case ear force/unit length, perpendicular to the wail, can be calculated directly from Eqn. C3.7. Towards the edges of the same wall, pressures will be carried by transverse bending back into the walls parailel to the direction of ground acceleration. The reactions from this action will be transmitted to the foundation by in-plane shears. The magnitude of the in-plane shear stress is unlikely to be critical.101 REFERENCES é 3.1 Priestley, M.J.t., "Analysis and design of circular prestressed € storage tanks", Journal Prestressed Concrete Institute, Vol 30, No 4, July/August 1985, pp.64-85. ‘ € 3.2 Hetenyi, 4., "Beams on elastic foundations", The University of Michigan Press, 1961. 3.3. Haroun, M.A., “Stress analysis of rectangular walls under é seismically induced hydrodynamic toads", ulletin of S. Soc. : of Am., Vol 7%, No 3, pp.1031-i041, June 1984 3.4 Timoshenko, S.?. and Woinowsku-Krieger, S., "Theory of plates and shells", McGraw-Hill, 1959. 25 Clough, 0.P., "Experimental evaluation of seismic design methods for broad cylinder tanks", EERC 77/10. 3.6 Cambra, F.J., “Earthquake response considerations of broad Viguid’ storage tanks", EERC 82/25 3.7 Choon-Foo and Shih, “Failure of liquid storage tanks due to ¢ earthquake excitation", EERL 81/04. 3.8 Wozniak and Mitchell, "Basis of seismic design provisions for welded steel oi] storage tanks", API Refining 43rd Mid-Year Meeting, vay 9 1978. egc 3.9 Peek, R., Jennings, P.C. and Babcock, C.0., ‘The preuplift method for "Anchoring" fluid storage tanks', Proceedings 3rd US National Conference on Earthquake Engineering, Charleston, August 1986. 3.10 Peek, 2. “Analysis of unanchored liquid storage tanks under seismic loads", Report No EERL &6/01, California Institute Technology, Pasadena. 3.11 Haroun, M.A., “Behaviour of unanchored oi] storage tanks : Imperial Valley Earthquake", Journal of Technical Topics in Civil Engineering, Vol 109, No 1, April 1983, ASCE. 3.12 Clough, R.M. and Niwa, A., "Static tilt tests of tall cylindrical liquid storage tank", EERC 79/06.