Chemical Engineering ELSEVIER ‘Chemical Engineering Journal 72 (1999) 91-95 Journal Vibrocentrifugal fluid pump Reh-lin Chen”, Steven L. Garrett Gradate Progam it Acoutet, PO. Bor 20, State Cole, PA 16802, USA Recived 25 Mach 1H recive in vse form 12 Septem 1988; scp 8 Otcer 1998 Abstract Anexperimental study ofthe ful pumping effect ofa clamped:-tee bar driven at its fist resonance frequency i presented. A flexible tube is attached to the bar undergoing osllatory excitation. The centrifugal Tore created by the vibration will generate a pressure gradient casing ‘steady, un-digeesional Guid Now. This process is extremely atractve forthe sel-pumping of heat-transport liquids Uueph the heat exchangers :torsionally resonant torial shermoacoustie refrigerator In that application, a single motor can be wsed to provide the resanant exertion of toe acoustlestanding wave within te hermoacoustic refigeratorand eliminates the requirement fortwo addition motors to pump the hot and cl heat-transport Mui, Measutements ae foun to agree with simple theoedcal predictions. € 1999 Elsevier Science S.A. Al rights reserved, Keywords: Ceuntuge foros Ocilatny excitation: Vibetion L. Introduction Centrifugal force has been used to create a large variety of pumps with pressure differentials which vary from inches of ‘water (blowers) to more than 10.000 atmospheres (compres- sors), Forexample, when water in 8 cup is sted vigorously, the water constrained by the curvature of the cup will rise ‘due to centrifugal force [1], The resulting pressure head depends upon centrifugal acceleration which is directly proportional to the square of the tangential velocity and inversely proportional to the rotational radius. A flexible, corrugated pipe open at both ends can emit sound by holding ‘one end and swinging the free end at some angular velocity. This serves as another demonstration of the use of centri- {gal force to provide centrifugal pumping {2.3}. The tones generated by the tube are harmonics of its acoustic reso- nance frequencies. The emission of sound is determined by periodicity of corrugation and flow velocity which induce turbulence in the tube. I€is aot surprising in the acoustical demonstration that swinging the tube in an oscillatory manner will also ereate sound. A typical centrifugal pump rotates in a single direction, Flow passes through the impeller, picks up rotational energy from it, and develops a pressure head. As the dependence of pressure is quadratic in the rotational velocity, vibratory excitation will also fead to steady, uni-directional fui flows as in the case of uni-directional fluid pump. Experimental *Conesponding author ‘measurements of a vibrocentrifugal pump which is driven by a ‘clamped-fiee" (clamped and driven at one end and free at the other) bar undergoing resonant flexural vibration will be reported. Fluid flow is restricted to a flexible tube attached to, or internal to, the oscillating bar. This process is extre- _mely attractive forthe self-pumping of heat-transport liquids through the heat exchangers in a torsionally resonant for- ooidal thermoacoustic refrigerator {4,5]. In that application, a single motor can he used to provide both the resonant excitation of the acoustie standing wave within the thermo- acoustic refrigerator [6,7] and the heat-transport fluid cis- culation. This eliminates the requirement for two additional -motors to pump the hot and cold heat-transport finds. 2. Theory ‘The governing equation for thin rod undergoing trans- verse vibration based on Bernoulli~Buler theory of beams is well known as [8,9] F Py ay 2 (e122) +00 a For a homogeneous beam with constant cross section, assuming harmonic motion, the above equation can be reduced ta ty 3 Q 1388.9041001 - 80 font mater) 1989 Elsevier Scence S.A, All ighns reserved PIT: 81385-8947(98)001 38~2 Rl Chen SL Garen/Chemical Engineering Journal 72 (1998) 91-95 Displacenent t | Fig. 1, ie thee taneverse os of clamped ba. where yu"EI and jis linear density of the material ‘Tae general solution of Bq, (2) can be expressed as dis placement yix)=A cosh Se+B sinh B-+C eos xD sin 3x. ‘The clamped-free boundary conditions preseibe both dis- placement and its slope to be zero (9=0 and By/x=0) a the clamped end. The moment and shear force ate both zera (8 yO" =0 and 9°y/84°=0) atthe fre end. This results in the characteristic equation, cosh cos +-1=0, which has roots of discrete value. The displacement of nth made at any aiven location has the form {10} {cosh3,x — e0s/,x) Bal + COs yl Sibi, sing? (sinh, — sing. a “The natural frequency of the fist mode is hence given by 52/8 VET. ‘The aluminum bar used io these experiments is 36 em Jong. Gem wide, and 4:mm thick. The resonance frequen- cies of the fist three modes ofthe bar are computed to be 5.7, fr=160.9, and fi=450.5 (Hz) (Fig. 1). Here only the first mo sof interest From the basie centrifugal force formula, ve have df(x) = dinj(x)/x, where dis the mass within dr at position x. The amplitude of the transverse velocity at any given location forthe fist mode is uy (2) ‘Therefore, dp/ de = p(x)" /x, and the pressure differential between the 10 ends is fob a ‘To simplify calculation of the centrifugal pressure, the displacement is expressed as a fourth-order polynomial in Dispiacemens 2 as 1 0.5 0.05 (OLB terms of position along the bar as shown in Fig. 2 where itis ‘compared to the exact solution, Rayleigh’s method [11] was. ‘used to compute resonance frequency of the first mode. By assuming yix)—x'bartthytechd, coefficients in the polynomial sre determined by the four boundary conditions. It follows that »,@)=s*—4i°+607, The first resonance frequency can then be calculated by equating potential energy with kinetic eneray of the transversely vibrating ‘beam, This approximation enables us to easily calculate the centrifugal pressure head, Ap accurately (wy = 3.53/? JET jj from this approximation). The calculated pressure differential at both ends is 0.68( 1/205"), where § = wy (1) is tip velocity, Ap has been normalized by kinetic energy density 1/2, The exact solution using Eq, (3) Mathema- tica™ gives Ap = 0.65(1/2p5"), ‘The model developed here takes into account dissipation ‘due to several loss mechanisms and incorporates them into the Bernoulli equation. For flow in a pipe, the major losses are shock loss due fo sbrupt changes in eross-sectional area and viscosity of the fluid. There are also some minor loss mechanisms, for example, roughness and bending. The Bernoulli equation inthis application can then be expressed a a sum of pressure heads as 6) where the first term on the right is due tothe eneray imparted from transverse oscillation, the second term is shock oss ‘with the constant & determined by the particular experi ‘mental geometry, and the thind term is fritional loss. The fiction factor /=64/Ren for laminar flow. When flow ‘becomes turbulent, f depends on Rep and relative roughness etd (12) and can be found on a Moody chart [13]. 0.2 0.25 0.3 0.35 Fig, 2. Displasemsat dition along the bax The sli fine represents exact solution ad she Hine ian approximation wing fourth-order polynomial.Rol Chen, 8. Garet/Okemical Engineering Jaumal 72 (1999) 81-95 * 3. Experiment In order to test the theoretical predictions, a simple experimental apparatus was designed and instrumented, "The core of the vibrocentrifugal pump is an aluminum bar of 36em long, 6 cm wide, and 4mm thick. The bar is clamped and driven at one end by a Briel and Kjer Type 4813 shaker and free at the other, Itis driven at its first ‘resonant inode as shown in Fig. 5. In addition to preliminary calibration of sensors, the experiment has swo parts. The first deals with flow rate as a dynamie measurement of flow velocity with respect t0 pressure head. The second relates bartip velocity to static pressure built up inthe tube due to centrifugal acceleration, 3d. Calibration of sensors In order to measure the vibrating velocity of the bar, an aceareleromieter is placed at its tip. Four struin gauges, connected as a Wheatstone bridge, were then used 10 measure displacement in order to eross-calibrate andl eplace the accelerometer. State loading measurements were per formed to confi that the bas behavior conformed with expectation. A linear variable ferential wansformer ALYDT) was used w measure the end-loaded deflection caused by a known weight W (Fi. 3). Since surface strain at the clamped end is given [14] as cmo/E6WHbrE) and deflection as y=—WF/(3E, one can relate end deflection y direely to strain The Young's modulas was found to be 7,06x10'°40.8% (Pa) and agrees with typical values for aluminum, The mol of a materia orsional longitudinal, and flextral) can also be ineasired accurately using « resonance technique [5} 3.2. Flow-rate measurement ‘To check the applicability of Eg. (5) prior to use in the pump, two flexible plastic tubes of 3.25 and 6.5 mm inside wor diameters and lengths of 725 and 95.0 0m, respectively: were connected to & water tank, Fluid flowing out ofthe tank was measured for known pressure heads. The Reynolds number ranged from ca. 1000 to 4000. Data show that pressure head has a quadratic dependence on flow velocity. ‘hile the ratio of flow rate to pressure bead varied somewhat, from one measurement to another ‘The results from typical measurements are shown in Fig. 4 und are fited using Eq. (6) as dh = 0.06 + 0.450" o here vis the averaged flow velocity based on net mass flow and sir = pv [kg/s). The flow is found to be only slightly laminar and dominated mostly by turbulence at flow rates typical of our experiment. Difficulties in making accurate comparisons to our theory caused by neglecting turbulence and entrance effects, were overcome by measuring the centrifugically induced static pressure head. 3.3. Static pressure head measurement A flexible be was affixed along the top ofthe aluminum bar with its free end vented to atmosphere and its driven end connected to a water tank, with the water level lower than the free end by 5 to 10 cm. This pumping device was able 10 draw water out of the tank and discharge at the tip of free end, Tip displacement amplitude for the onset of discharge al the first mode of vibration is around 1em, Higher accuracy was obtained using a static approach in. which the ftee end is vented through a capillary tube and the driven end is terminated by an Endeveo Model 85308 pressure sensor, as shown in Fig. 5. Calibration was validated by standard laboratory methods. The vibration of the bar gen- erates centrifugal force and induces a pressure differential between the free end (higher pressure) and the driven end (lower pressure). The pressure head is then determined by the pressure sensor output voltage divided by its sensitivity ‘Water and mercury were used as the finid media in these swrain gauges 7 gauges 27 7» Fig. 3. Eaperinenal sup for mesusing deflection de to « own loud wsing LNDT and sai gauges.o RL Che, Sb. Garvett/ Chemical Engineering Journal 72 (1999) 91-95 Slow xato .is| 0.008} Co CC Fig 4. Row-rate measurement data wih tical as pote ps and horizontal ox ead in. The pois ndeate measured dat, ed the sli lines s gonratie carve ft measurements, Figs, 6 and 7 show typical results from this The measured induced pressure differential Ap = (1/20) static pressure head measurement with & varied from 0.47 to 0.57. ‘The driving frequencies were fixed at the first mode which tre 21.0 He for water and 19.5 Hz for mercury, respectively. veieetsy free end Geivon oret aluninnan bar Fig §, Soup for measuring prssu tfc bu up onthe ight co Sipe by wtmcentnigal force, Tw of four hosed resiie sain gaaes ae shown the vise Fig. 6, Tip velvity (avs) vers induced presse head for water. velocity aa] y a8 0.2 Fig 7. Tip velovty (mf) verses induced presse head for meres.Rel Chen, SL Garret / Chemical Engineering Journal 72 1999) 91-95 95 4. Results and discussions Although the flow-rate measurement has a very straight- Forward setup and dramatic results, these measurements were not well suited to detailed quantitative comparison with the simple theory. It was found in the flow measure- ‘ment that Reynolds number typically lies in the turbulent and transition regions. The lengths of the flexible tubes typical for implementation were not long enough in order to hhave fully developed flow. There is also a discontinuity at the connection of the water reservoir to the tube which causes loss as discussed previously, Nevertheless, the result of pressure head versus Velocity is always a parabolic curve ‘with a coefficient determined by a particular experimental geometry. Since centripetal acceleration is. a,—(a)", induced pressure head depends quadratically on the transverse velo- city of the bar and is inversely proportional to its length, A simple approximate theory, valid for small deflections gives the ratio of pressure head to kinetic energy density of 0.68 and agrees with our experimental measurements (© an average of 25%. The result is satisfying considering for measurements taken within quasi-turbulent region. We expect that & more robust attachment of the flexible uid fube to the vibrating bar might improve the agreement between theory and experiment Acknowledgements ‘We would like to thank the Office of Naval Research for sponsoring this work, Appendix 4.1. Nomenclature 4 centripetal aeceleration (an s~ ‘A cross-sectional area (mn?) b width of bar (m) d diameter of tubes E Young's modulus (Pa) § gravitational acceleration ~ 9.8 (ms pressure head (im) ‘moment of inertia for area (104) length of a material (m) mass flow rate (kg 8") acoustic pressure (Pa wert Ap pressure drop (Pa) Rey Reynolds nuuber peak vibration amplitude at aluminum bar tip (m) peak wansverse velocity amplitude at aluminum bar tip (ms) 1 thickness of bar (a) averaged fluid velocity (ms) W> weight of toad (W) transverse bar deflection (m) ¥ transverse bar velocity (ms) 42. Greek letters 8, roots of beam equation of mode © strain @ oscillatory angle (rad) p mass density of media (kg m™*) o stress (Pa) angular frequency (s~') linear mass density (kgm) References 11) HLH. Anderson. Ceaiugal Pumps and Alied Machinery, Bsever Scieae Pablbers Lid, th ad, 1994, [pI FrinkS. Croton, Singing soaps pipes, Am 5 Phys 42.0978) 275-283, [3} Lovis Fi Cadwell, Singing soragated pipes revisited, Amt J. Pays 21994 21-227 4) J, Osborne, SL. Garest, Trsonally Resonst Txoldat Thermo ator, 100, P. 2 (1986) 2816 [5] Stevo L. Gare, Torsiaally Resonant Teridal Thermoscoustic Resngeriten US, Navy patent case aeber 7521 (6) Gregory W. Swif, Thermoucoustc Engines 841988) 1148-1180, 17) Gregory W, Swit, Themoscoestic engines and reign, Pysios Today 497) (1908) 2-28. Is) Kart F Gra, Wave Maton in Ete Solids, Oxoud University Pass, 1975, hap. 3 19} Willan 7. Thomsee, Thoory of Won wth Application, th ‘ed, Chapesan and Hal, . 9, [10) Laveence B. Kinsler, Ausin R.Frey. Allan B. Coppeas, Janes V. Sander, Fundamentals of Acoostcs, ded, Jobn Wiles sa Sons, Dow York, 1982p. 75 U1] SP Den Histog, Mechanical Vibons, Dover, 1988, chp. 4. (02) Rober W. Fox, Alan T. MeDouald,Insoduction to Fd Mechanics, Sahin Wiley an Sons ine, 1992, 9p. 348-382. [05] LE Moods, Faction factors for pipe Now, Transetons of the ASME 648) (1848) 671-654 (14) Warren C. Young, Rohs Formls for Stress sl Sts, Meee Hl, 5h ed, 1989, p. 100, UIs) Steen 1, Gane, Resonant seas determination of laste moda A Avast, Soe Am. B81) (1980) 210-2