COMMERCIAL-IN CONFIDENCE 7 VIBRATION CALCULATIONS When a fluid flows across or past a supported tube it can excite vibration of the tube. In soine cases this vibration can be severe and result in damage of the exchanger. It is therefore important that the possibility of tube vibration is assessed during the design of shell and tube heat exchangers. 7.1 Mechanism of Vibration ‘The cause of flow induced vibration in tube bundles remains an area not totally understood. Many mechanisms of vibration can be identified, but three major mechanisms are relevant to shell and tube heat exchangers, These are turbulent buffeting, vortex shedding (including acoustic resonance) and fluidelastic instability. These mechanisms are described here. 7.1.1 Turbulent busfeting ‘Turbulent buffeting occurs when a body is placed in a turbulent flow field and the forces on the surface of the body are changed rapidly with time and position. ‘This is because the magnitude and direction of the velocity vector are magnitude and direction continually changing, If the variation in force is well correlated, that is the changes infor along a significant length of tube are in phase, and the frequency of the variation is close to the natural frequency of the tube, tube vibration may result. A phenomenon which can cause these forces to be well correlated over a tube length, isthe occurrence of standing acoustic waves. This may occur when the shellside fluid is a gas or vapour. Inorder to examine the possibilty of turbulent buffeting excited tube vibration, the program estimates the tube natural frequency (f), the turbulent buffeting frequencies (f) and the acoustic resonance frequency (f). The results are presented as ratios of the turbulent buffeting frequencies to the natural frequencies and if the shellside fluid is a ‘gas/vapour the ratios of the turbulent buffeting frequencies to acoustic resonance frequency are also presented. 7.1.2 Vortex shedding and acoustic resonance Vortex shedding is caused by the periodic shedding of vortices from the tubes and can lead to damaging tube vibrations if t occurs atthe tube natural frequency. This can cause substantial vibration in liquids, but only negligible vibration in gases (unless the ges has a high density and velocity). However, vortex shedding may become important in gases if it is reinforced by another tube vibration mechanism which happens to occur at the same flow condition, Such reinforcement may come from acoustic resonance. ‘Acoustic resonance is important only in gases. It occurs when the frequency of an acoustic wave in the heat exchanger coincides with the tube natural frequency. It should be noted that even if the acoustic wave does not cause tube vibration, it can lead to intolerable noise levels. Itisnot very likely that damage from acoustic resonance along will occurin the cylindrical shell and tube heat exchanger geometry. In some cases it may lead to damage to batile plates and to ducts containing the tube bundles. Acoustic effects may also be important in two phase flows where the sonic velocity can be small. VIBRATION CALCULATIONS cat © 1993 HTFSCOMMERCIAL-IN CONFIDENCE In onder to check for the possi frequencies (f,) and outputs the ratios of these to the natural frequencies (,). Ifthe shellside fluid is a gas or vapour the program also outputs the ratio of vortex shedding frequencies to the acoustic resonance frequency (f). ty of vibration due to vortex shedding, the program calculates the vortex shedding 7.1.3 Fluidelastic instability Fluidelastic instability is important for both gases and liquids. It occurs whenever a fluid flows through an exchanger and some motion is imparted to the tubes. The displacement of a tube from its normal position alters the flow field through the bank which in turn alters the force balance on other tubes. The fluid imparts energy to the tubes. This cnergy is dissipated by the damping of the tube. As the velocity of the fluid is increased the amount of energy imparted tothe tubes increases. If the energy imparted to the tubes exceeds that dissipated by damping a vibration of increasing amplitude is set up. A critical flow velocity can be identified below which no vibration problems is envisaged and above which significant damage may occur. ‘The fluidelastic instability process has been analysed by Connors (1970) who has shown that the critical velocity is dependent upon the natural frequency of the tube, the properties ofthe fluid, the effective mass and diameter of the ‘tube and the damping of the tube. This analysis has been extended by Goyder (1982) to the two phase situation. The program calculates the critical velocity using three values of log decrement; 0.01, 0.03 and 0.1. This allows the User to judge the severity associated with the fluid elastic instability. The recommended value of log decrement is 0.03. Note that this is not a resonance but a true instability limited only by the damping due to tube/tube, tube/battle and tube/shell colisions, 7.2 Application of Vibration Assessments in TASC3 7.2.1 Shell types considered ‘The assessments of vibration problems by TASC3 are not applied to all the geometric configurations for which the program can perform heat transfer calculations. The following identifies the geometries for which assessments are or are not made. Vibration calculations re performed for B, FJ, Gnverted J) and X shell types [No vibration analysis is available for G and K shell types [Exceptions to the above type classification are that no vibration analysis is made for unbaffled or rod-baffled units. Special consideration is made for X shells. This shell type may be unbaffled, but the fluid on the shellside is in ccrossflow. Vibration assessment is therefore important, 7.2.2 Tube rows examined ‘A major problem in applying vibration tests to a shell and tube geometry is the complexity of the flows, The fluid on the shellside ean flow in many dircetions, mainly governed by the baffle arrangement, and with different magnitudes. ‘VIBRATION CALCULATIONS 72 © 1993 HTFSCOMMERCIAL-IN CONFIDENCE However, experience has shown that there are certain tube rows within an exchanger that are most likely to suffer vibration damage, These are basically the uppermost and lowermost tube rows with respect to the flow and the tube rows just inside or just above the edge of the baffle cut, These are the tube rows examined in TASC3. They are summarised in Figure 7.1 for the various configurations of bafile allowed in TASC3. ‘The main effects considered by these defined tube positions are the amount of support and the velocity impacting the tube. Clearly a tube in the window region of an exchanger is supported by every other baffle. In the overlap region a tube is potentially supported by every baffle. This has an important effect on the calculation of, for instance, tube natural frequency and hence the likelihood of vibration. For each tube row three positions along the exchanger are considered. These essentially correspond to the inlet and, defined by the region between the tubesheet and first baffle; the outlet end, defined by the region between the last baffle and the other tubesheet; and a mid-space, defined by a typical region between two baffles. For I and J type shells assumptions have to be made because of the inlet and outlet geometries. These are described in Section 7.2.3. For X type shells certain assumptions must also be made since the shell is potentially unbaffled. The vibration methods available in TASC3 prior to the inclusion of X shells assumed that a shell contained a minimum of 2 baffles. To simplify inclusion of the methods for X shells this assumption has been retained and satisfied by the use of fictitious baffles located very close to the tubesheets. This means that essentially all of the exchanger is mid-space and the ‘inlet’ and ‘outlet’ regions of the output are not reported. An X shell can of course have bafiles in the form of intermediate supports and these are considered as for other shell types. 7.2.3 Entry conditions ‘The velocity distribution on the first (top) tube row at theexchangerinletis very much dependent on the entry conditions. ‘The flow is dictated by the nozzle type and presence or absence of an impingement plate. Such variation is considered by TASC3. Furthermore, in determining the likelihood of tube vibration due to fluidelastic instability the variation of velocity along the entire length of a tube row and the mode shape of tubes in that row are very important. For this reason ‘TEMA E and J shells are treated differently as are shells with or without impingement plates. Allowance is also made for shells with end-spaces whose lengths are different and/or not equal to the baffle pitch. In general, for shells with no impingement plate the total shellside flow (or half this for TEMA I shells) is assumed to impinge as a jet (of cross-sectional area equal to the nozzle cross-sectional area) on the first tube row encountered, For shells with an impingement plate the total flow (or half of this for TEMA I shells) is assumed to be spread over the whole area of the top tube row minus the area taken up by a square impingement plate of side equal to the inlet nozzle diameter. In every case for tube rows other than the top tube row the crossflow velocity is calculated for the ‘mean shellside properties and quality taking account of bundle bypass and baffle leakage flows, For TEMA J or I shells the calculations assume a "zero" cut baffle on the inlet (Gutlet for I shells) nozzle centre-line. VIBRATION CALCULATIONS 73 © 1993 HTFSCOMMERCIAL-IN CONFIDENCE 7.3 Interpretation of Results ‘The following discussion of tube results produced by the vibration assessments of TASC3 follow closely the data presented in the table of vibration predictions of TASC3’s output. An example of this output can be found in Table 7.1 for reference. 7.3.1 Tube natural frequency ‘The basic equation used to determine the tube natural frequency, fy is hoe se &) a) where: 8% isa dimensionless quantity based on geometry (refer to DR48 or HTFS Handbook Sheet VM7) L__ isthe tube length E__ isthe Young's modulus Tis the second moment of area M, is the effective mass, is calculated from the relationship I= gdi-d) 12) ‘where d, isthe tube inside diameter and d, isthe tube outside diameter for a plain tube. For a low fin tube d, is, in terms of the tube inside diameter, d, and the root diameter, d,, as d, = d, + 1.08 (did). ‘The effective mass, m,, for a plain tube is given by + pli + pads ~ di) (73) where: cc is the added mass coefficient (refer to DR48 or HTFS Handbook Sheet VM7) pis the average fluid density outside the tubes is the fluid density inside the tubes pq is the fluid material density. For low fin tubes the equation becomes: a + pdt + oe(s-4)- era | aa) VIBRATION CALCULATIONS 14 © 1993 HTFS(COMMERCTAL-IN CONFIDENCE “The above calculations are for straight tubes. When exchangers have U-bends the derivation ofthe natural frequency is more complicated and based on the ‘dominant span method’ outlined in HTFS Handbook Sheet VMI1, Essentially this means that the natural frequency fora span is assumed to be independent of adjacent spans. The natural frequency is then taken tobe the lowest frequency calculate. According tothe normal tube postion considered TASC3 assesses the natural frequency for tubes atthe baffle edge and in the window region. ‘The frequency is given by 0.04025 C, (Ey — lin (73) ee M, where: C, is constant dependent upon the relative unsupported lengths of the U-bend at the top and bottom of the bundle and ris the radius of the U-bend at the point of calculation. 7.3.2 Acoustic resonance “The acoustic resonance frequency is calculated differently according to the phase of the flow, the difference being due to the calculation of the velocity of sound. The acoustic resonance frequency, fy is given by u, f, = 05861. > 7) where: Dis the shell diameter and U, is the velocity of sound. Fora liquid U, is set to on and for a vapour or two phase flow P 14 = a. 5 (78) where P is the system pressure. Jn the above calculations the density, p, is taken to be the lowest of shellside inlet and shellside outlet homogeneous densities. Itshould be noted that the constants in equations (7.7) and (7.8) are fora and water. The velocity of sound calculations for other fluids are therefore only approximate. ‘VIBRATION CALCULATIONS 18 © 1993 HTESCOMMERCIAL-IN CONFIDENCE, 7.3.3 Stroukal number TThe characteristic frequency for periodic wake shedding is obtained from the Strouhal number defined as a a9) where f;is the frequency, d, is the tube outside diameter and wis the flow velocity. ‘The Strouhal number fora single tube is 0.2 over a wide range of Reynolds number. However, for bundles the Strouhal number may vary over a considerable range. 7.34 Vortex shedding "The damage due to vortex shedding alone is only considered important when there are high density fluids (e.g. liquids) ‘on the shellside, For two phase calculations the mean shellside homogeneous density is used. In low density fluids (e.g. gases) vortex shedding is only considered important when there is coherent shedding of the vortices along a tube length with the same frequency as the tube natural frequency. ‘This coherent shedding of vortices ccan occur when the shell acoustic frequency is within 20% of the vortex shedding frequency. In this case the vortex shedding frequency "locks-in" to the shell acoustic frequency and large amplitude tube vibration can occur. "The vortex shedding or periodic wake shedding frequency, f, is derived from the Strouhal number described in Section 7.3.3. The velocity used in the equation is different according to the tube row selected and whether the inlet, mid or ‘outlet region is being considered, since the flow area changes. The velocity used is the crossflow velocity (i.e. bypass and baffle leakage flows are accounted for). ‘The vibration analysis is presented in the form of the ratio of frequencies, Fitzhugh (1973) made the following recommendations for the frequency ratios which may present a problem, fy he i¢ 2<07 or > 13, no vibration likel 7.10) E i ly (7.10) it 07 <4.<13 tube vibration may occur an h if 08 <2 <12 tube vibration will occur 7.12) where; fis the excitation frequency for vortex shedding, and ‘f, is the natural frequency of the tube. ‘The TASC3 output is highlighted by asterisks when the ratio of vortex shedding frequency to tube natural frequency falls within the last of the above categories, namely between 0.8 and 1.2. The ratio of vortex shedding frequency to acoustic resonance frequency is treated similarly due tothe “lock in” effect. ‘VIBRATION CALCULATIONS 16 © 1993 HTFCOMMERCIAL-IN CONFIDENCE <05 no coupling of processes expected (7.13) > 20 1no coupling ofprovesses expected (7.14) El <20 congress 5) hab it os <(£ oc £)<12, coupling ike 71.16 (f i soupling likely (7.16) For rotated square tube layouts large amplitude acoustic standing waves ean be generated when only vapour is present pn the shellside. ‘This condition can usually be avoided quite easily by using detuning plates which are intended to ‘change the resonant frequency of the heat exchanger. 7.3.5 Turbulent buffeting ‘Therecommended ratios of turbulent buffeting frequency. to natural tube frequency, f,, Which delincate the likelihood of excessive vibration are the same as those for vortex shedding. Turbulent buffeting is not thought to be such an important mechanism of vibration damage unless coupled with vortex shedding and/or acoustic resonance. Ithas only been measured in water. ‘The turbulent buffeting frequency, f, is calculated as 4 a7) t where: dy _ is the tube diameter is the velocity , _ isthe appropriate longitudinal pitch allowing forthe flow direction P, isthe appropriate transverse pitch allowing forthe flow direction 7.3.6 Thorngren damage assessment criteria If vibration occurs then tube damage may result, This could be in the form of damage at the bales or possible by direct collision of tubes. Thomgren (1970) identified criteria for baffle damage and collision damage as follows. Nyo < 1, tube damage atthe baffles unlikely Ngo 2 1, tube damage at the baffles likely Neo < 1, tube collision damage unlikely Nep ® 1, tube collision damage likely where: Ngo is the baffle damage number, and VIBRATION CALCULATIONS 1 © 1993 HTFSCOMMERCIAL-IN CONFIDENCE Neo is the collision damage number as defined by Thorngren (1970), ‘The baffle damage and collision damage numbers are defined below. Baffle damage number, dy 0 Neo = Fy Sy Aw B, Collision damage number, No = ee ee FR Ay @ + d) GB where: p, = _shellside fluid density (kg/m*) F, = tbe/baffle clearance factor (-) ie. tube/baffle dismetral clearance 0.79 mm Fy tube/baffle diametral clearance 0.40 mm Fy Sy = maximum permissible tube material fatigue stress (N/m’) ie, carbon and low alloy steels Sy = 6/2400 high alloy steel, stainless steel, cupro metal Sy, = (E= Young's modulus for tube material) Ay = cross-sectional area of tube material at baffles (m*) bafile thickness (m) C= minimum clearance between adjacent tubes (i) In order to make better allowance for the effect of physical properties, a modified form of the baffle damaged number, ‘Nac s also defined. ‘This number is dimensional, TASC outputting a value always in PURE SI UNITS. No = (J Nao 7.20) ° where: w shellside fluid bulk viscosity (for liquid ortwo phase this liquid bulk viscosity) ° homogeneous shellside density ( avin ° inlet density (P,) te = inlet viscosity ny, at outlet ° outlet density (?,.,) 1 = outlet viscosity (aay) formidsspaces Tovest of Pe Par ™ = comesponding viscosity ‘VIBRATION CALCULATIONS 18 © 1993 HTFSCOMMERCIAL-IN CONFIDENCE 7.3.7 Log decrement ‘The tube vibrations are damped because energy is dissipated by the impaction and sliding at tube supports, the damping characteristics ofthe tube material and the effect of true fluid outside the tube, These damping effects are very complex and individual tubes can have very different characteristics. For luidelastic instability, in particular, the damping mustbe considered. The overall level of damping can be indicated by aterm called the log decrement, 8,. TASC3 considers three levels of damping defined by log decrements of 0.01, 0.03 and 0.1. The highest value, 0.1, reflects heavy damping; the value of 0.03, is taken to be a typical value repre- sentative of a heat exchanger. See DR48, Part 4 (1993), for a discussion of damping values 7.3.8 Fluidelastic instability ‘Although described last, this is perhaps the most important of the modes of vibration that the program considers. Unlike the other types of vibration described this is not a resonant vibration and ultimately tube movement is only limited by contact with other tubes or mechanical constraints. For the resonant type vibrations a reduction or increase in velocity for a given geometry may remove the vibration. For fluidclastic instability the phenomenon occurs above actitical velocity. ‘TASCS assesses fluidelastc instability by comparing the actual mass flow into the exchanger with the critical mass flow for onset of instability. Since the calculations are dependent on damping, results are expressed at three levels of Jog decrement (see Section 7.3.7) which are equivalent to light, typical and heavy damping. When the ratio of actual to critical mass flow is greater than 1,0, TASC3 identifies this with asterisks in the output Connors (1970) identified a means of calculating the critical velocity according to the equation K vpa? fg, Vm, 8 21) W, is the critical mass flowrate is the Connors constant (typically between 2.5 and 3.5) where: = is density is the flow area is the tube natural frequency is the effective mass per unit length oe mR Dm and is the log decrement. In order to allow for different geometries Goyder (1983) proposed a modified form of the equation _ ee ae Wo = KN aata ee fmE C2) where the terms x, to 2, are representative of the kinetic energy content of the vibration from different parts of the exchanger. VIBRATION CALCULATIONS 19 © 1993 HTFSCOMMERCIAL-IN CONFIDENCE, Neither equation takes account ofthe differing flow directions ofthe fluid, such as when vertical cut baffles are present, which affects the value of K, Furthermore, a comparison of flowrates assumes a knowledge of the flow of fluid over ‘particular tube. At the inlet a tube will se the full flow, whereas in the mid-space only a fraction of the flow will ‘be important because of bypass and leakage flows. | ‘The varying flow directions are taken account of by incorporating K with the {}' term and multiplying x; to x5 by appropriate values according to the local flow direction, Differing mass flowrates are aeounted fr by directly calculating the ratio %, using flow factions to reflect the local variation in low. Hence, equation (7.22) is modified as l= | _¥, | s| We 023) | + | where: Wi, are flow fractions K, are equivalent to the Connors constant and Wy is the total mass flowrate into the exchanger. ‘The value of x, is very dependent on the exact exchanger geometry (i.e. shell type, inlet conditions, baffle orientation etc), For the geometries considered by TASCS the values are defined as follows a= 0) In 24 Ta (724) _ +1) SF Dak wet 8 Wet — nO, a) % 0.26) 5 a2) | % = T0508 a? p, © +b) ao SE 0.28) by = tN tN“ DE 729) exceptions to value of 2 are for the following situations ‘VIBRATION CALCULATIONS 7.10 (© 1993 HTFSCOMMERCIAL-IN CONFIDENCE J shell with horizontal baffles: WB = DOAN ANI tr Tshelt: be w= (NG + My + NOOSE ‘The term x has a large number of possibilities because of inlet conditions. No impingement protection (circular ‘jet’ from nozzle) we 21.0: (=i) - 35.0 P-if) + 1580 WB) * * 1058 AT Bs c No impingement protection, J shell, horizontal cut baffles 2 in (42 320 Fi {a ra (| Ay, [2 ~ ae No impingement protection, J shell, vertical cut baffles - a fy salt} “ae [2 i ‘Square impingement plate ae Hf = 21.0 RUP= Ih) + 35.0 Il 105 AF p, ” F ha)) — 15.007 =I) Square impingement plate, J shell, horizontal cut baffle R 2 (by AP cS) ‘Square impingement plate, J shell, vertical cut bajjle ‘VIBRATION CALCULATIONS 7 (7.30) 31) (732) (733) (734) (7.35) (7.36) (731) © 1993 HTFSCOMMERCIAL-IN CONFIDENCE Vapour belt (or full length impingement plate) ee erie 38) 1050 BAP, ae Vapour belt (or full length impingement plate), J shel, horizontal cut baffles . oD) £2 ly AP Pr : Vapour belt (or full length impingement plate), J shell, vertical cut baffes f : (7.40) DAF ‘where in all above equations: 2, = inlet density 2, = outlet density Ny = number of spans with density p, = number of spans with density between p, and p, ‘= number of spans with density p, let span length inlet span length minus inlet nozzle diameter ‘= mid-space span length (normally baffle piteh) baffle pitch outlet span length inlet span cross-sectional area for flow rmid-space (ie, baile to baffle) cross-sectional area for flow A, = outlet span cross-sectional area for flow d, = nozzle diameter ‘The geometry for an X shell is somewhat different, necessitating a slightly different set of solutions. As described in Section 7.7.2, there is an assumption of 2 baffles for the vibration methods, but all flow is uniformly across the exchanger. The modified form of equation (7.22) is. Dit D; = Vegan ome ty where ee 2 ee 2) = TS Earp 742) ‘VIBRATION CALCULATIONS 742 © 1993 NTFSCOMMERCIAL-IN CONFIDENCE D = 1.43 TAN 2 Bon oa) wp aap 105 EAP p, PRE aay 2O+D p, = EH+D oo 105 In D, = w-9 746) A isflow area p— isdensity Nis number of spans K is fluidelastic constant 1 isspan tengths and subscript, | is for left-hand end of exchanger is for right-hand end of exchanger ‘m i for mid region (in practice the whole length of the exchanger). The terms D; and D, are treated as always having a uniform flow, and are only intended to satisfy the necessary equations. The term D; is for the main part of the exchanger and depends on the inlet geometry. The value above is for the case of a uniform flow which might be expected if there are a large number of inlet nozzles. In practice one of the following conditions is assumed, No impingement plate (circular ‘jt’ from nozzle) 1 Dy = spas ho - 49) 2 Fae ON KE b oan where: Ais the nozzle area andy Ae define the position of a nozzle as shown in Figure 7.1 Square impingement plate D, z fa, ay =a + (sin 2 sin 22%) 7.48) eos = + [sin SF -sin = Be o7Az panies ade ede eee maa Cae) Ch) VIBRATION CALCULATIONS 73 © 1993 HTFS