Scribd Logo
Upload

Welcome to Scribd!

  • Instability Threshold and Stability Boundaries of Rotor Bearing Systems W. J. Chen 1994.

    Uploaded by

    jjavorova
    20 views8 pages
    scientific paper

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    scientific paper

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    20 views8 pages

    Instability Threshold and Stability Boundaries of Rotor Bearing Systems W. J. Chen 1994.

    Uploaded by

    jjavorova
    scientific paper

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    of 8
    PD-Vol. 64-7, Engineering Systems Design and Analysis Volume? ASME 194 INSTABILITY THRESHOLD AND STABILITY BOUNDARIES OF ROTOR-BEARING SYSTEMS W. J. Chen Centsitugal Compressor Division IngersoltRand Company Mayfield, Kentucky pestract M_—msvceria matic ‘A. direct numerical method for the determination of Mw ‘state space matrix defined by equation (9) inode ental saby touatc of fens rior Ny Torapee re a ee pe natiy bony of ann degra com a Ctithedy examining he vanoin np opening parte eter Fee ee eacia tt Gs ening acer fo ee Sick cat dote eed wimser Gf gaged cr a ae ee dodgy weables erepening Sata, ten model nom SE cS Te aa ote Scie Sooo Patan ctu ceca many ie mele algorithm is based on the nonlinear optimization techniques. x’ state space vector defined by equation (9) A ee mociel ts lege be fein, ny apie Tesi ac lier poped poi. Acme Waxing lind Fant al tas prance nha Gr ut pecans aie cess mci | Wee a ae ociea Goel eit ta pascal piety Orage He ere remprem eon ners perm a ta pesman vil open es roe eens Be epee teal hit teeny NOMENCLATURE i hea iy cn aamping © damping mai G —beangraialleartce wrroouerion Wert carare Tre doug af rotor axing systems are econing more tern ameter comlek Soto the tooe tte rr rola en ye orm Sma te the aay casllenin, an covens fev fncon SONS cde Tp des sommes om & Genopi nee cr eal the pti a oping a the lad L Saruglewh ee ete catia amyl Kile epten nd pee rtrsvespad, Vn ft, Be Kies matic SER ced. adm uring al “ad K ‘state space matrix defined by equation (9) ‘aerodynamic forces are the only restrictions to the wide range of 5h tae EEE A vey ge and ote Win au ‘component, typically ranging from 20% to 80% of the rotor speed, can be observed by dynamic signal analyzers or FFT emulators if the system is operated beyond the instability threshold or the stability boundaries. This excessive vibration usually results in machine damage. Therefore, the determination ofthe instability threshold and the stability boundaries of the ‘operating parameters is critical fr the machine’s safe operation. ‘When a rotor is operated beyond a certain rotational speed, a very high and unstable vibration component with a fractional fiequency ofthe rior speed will be developed. This rotor speed is referred to as the instability threshold, At a constant rolational speed, the rotor can experience the same unstable state when an operating parameter is beyond a certain valve This limit is refered to as the stability boundary of that ‘operating. parameter. The instability threshold and sisbility boundary are usually determined from a stability map. A map of logarithmic decrements or damping coefficients versus the ‘operating. parameter is generated by repeated calelation of igenvalves for a range of an operating parameter under stu Lind (1974) utilized the transfer matrix method for the ‘calculation of damped critical speeds and instability threshold by repeatedly solving the eigenvalue polynomial equation. Since then, various roots searching techniques based on the transfer ‘matrix polynomial method were proposed by Murphy and Vance (1983), Kim and David (1990). ‘The instability threshold is then determined graphically by a stability map. A disect numerical ‘procedure forthe determination of instability threshold based on the transfer matrix method was presented by Zhou and Rieger (1985). A quadratic interpolation and iteration schemes were ‘sed in their algorithm to determine the instability threshold and the corresponding whirl frequency. No stability boundaries on the operating parameters were discussed. The mumerical difficulties involved with the transfer matrix method as the ‘model size increases are well known. The possibilty of missing some vibration modes sill remains. A closed form solution of the stability parameter for a rigid and symmetric rotor system ‘supported by two identical uid film bearings was presented by Lund and Thomsen (1978). Roo (1983) extended the work to 3 flexible and symmetric roior supported by two identical bearings. However, for the practical applications, « stability aly fe sateen omdaion aes required to determine the system sabi "The development offi element formulations fr wse in the rotor-bearing foundation systems has received considerable ‘attention within the past few years (Nelson and MeVaugh, 1976, Rouch and Kao, 1979; Nelson, 1980), The capability of ‘modeling complex systems and the stability of the numerical ‘algorithms make the finite element method a popular and indispensable too! in the eld of rotordynamies. The effects of rotatory inertia, gyroscopic moments, shear deformation, axial loads, flexible disks and housing are included inthe work. The linearized bearing. and seal coefficients, aerodynamic. cross ‘couplings, and other sources of interaction can be easily incorporated into the mathematical model. Rajan etal. (1986) presented a direct numerical algorithm for the calculation of ‘damped critical speeds using eigenvalue sensitivity based on the finite clement formulation. The parameter sensitivity in the ‘dynamics of rotor systems was discussed in their poper, These 22 sensitivity coeficients ar valid and canbe used only fra smal design change. The instability threshold still has 10. be determined by the graphical method. Very often, the effets of the operating parameters on the sjstem stability ae essential in the design stage. The rotor speed dictates the aerodynamic performance. The lubricant viscosity determines the ol type and operating temperature range. The bearing loads determine the range of throng. Therefore, iis extemely usefil to have an fvomated procedure to direlly calculate the insabiliy Aweshold and. the stability boundaries of the operating parameters ‘A direct numerical procedure proposed in tis paper allows the designers to determine the instability threshold and stability boundaries ofthe operating parameters if the boundaries exist The governing system equations of motion are derived fom the finite element method. The solution procedure is based on optimization techniques. The optimization techniques are very effective and suitable for this type of algorithm. The operating parameters are subject to constraints as dictated by practical Timitations. Examples ar presented to illustrate the proce GOVERNING EQUATIONS OF MOTION “The finite element formulation ofthe rotor assembly is well documented by Nelson (1980) and Ehrich (1992). The equations ‘of motion for the rotating assembly and foundation are of the form M, 0 ofa) fe, 0 oq o mM, Oat} 0 G, ohat+ o o mila} Lo o olla} a K, Ky Ky][a} (@ K, Kz 0 lat=1a k, 0 K,|la,} a, where the vectors q,, q,, dy fepresent the displacements of shaft elements (and rigid disks), flexible disks, and flexible foundation, respectively. The massinertia matrix M is a real symmetric matrix which is @ function of structural properties ‘The gyroscopic matrix G isa real skew symmetric matrix which isa fiction of structural properties and the shaft rotational speed, and the stiffness matrix K. can be an asbitrary real mtzix ‘due to the nonconservative axial loads. The motion of the rot is described by two translational and two rotational coordinate displacements at each finite clement station. The motions of the flexible disks and foundation can be described by six degrees of| freedom at each finite element station. The matrices of flexible disks and foundation can be obiained from any existing structural finite element computer programs, such as NASTRAN snd ANSYS. The three primary structural components (shaft and rigid disks, flexible disks, and foundation) are coupled by the interconnecting components Fluid film bearings are commonly used in the rotting machinery due tothe bearing damping, sifiness characteristics and long life of operation. The use of linearized bearing dynamic coefficients inthe analysis of rotor dynamics has been ieyecepted (Lund and Thomsen, 1978; Land, 1990), The Tearaed governing equation o mation fo a Od fi ering eSecting shat atin and founationj (or station jo nother shaft) is given by G Ga {& Kf] [2] @ » Hal LK, ese, ‘where the 2x2 matrices C, and Ky, are the bearing damping and Sess cosets witch are” detained. by sling the Reynolds equation and its perturbed equations (Klit and Lund, {Bee Tost the elit of bearing preter on the ster aii, te bang Symi counts are expressed ine ondimensional formats and represented as functions of the Sommerfeld number. The bearing Sommerfeld number is R defined by 5, = 7s (2) Wale and the non-dimensional stiffness and damping coefficients are e me ZC, -q G0, Rye Gy deny seay) @ ‘The Sommerfeld number and non-dimensional coefficients ‘ae sometimes normalized with respect to the lobe radial clearance. However, the bearing radial clearance is more often ‘used in the hearing and equipment manufacturers for the neasurability and machinability. Hence, the bearing radial clearance is employed in this paper. The relationship between the bearing and lobe radial clearance is related by the bearing woh Bia o For 2 given operating condition, # Sommerfeld number is | determined and the calculation of corresponding bearing stiffness and damping coefficients becomes. straightforward. “Thus, for a given type of bearing, the dimensional bearing ‘oeficents in equation (2) are functions of rotor speed, cil iscosity, bearing load, and bearing clearance, | Other sources of self-excitation include the aerodynamic effects of impellers and seals. This type of de-stabilizing forces ‘a be expressed inthe form (Weiser and Nordmann, 1989) ee aE elt} ‘assembled equations of motion which describe the characteristics of the system are ofthe form. +IG+C,]q+1K +K,]q=Q o is the vector of the operating parameters which inelude ‘olational speed, lubricant viscosity, bearing loads, bearing ‘and axial loads, etc. The matrices C, and Ky are 213 quite sparse and can be rearranged for minimum storage requirement. SYSTEM STABILITY For a given operating condition, the stability of the dynamic stem can be determined by the eigenvalue equation of the homogeneous state space form MX(1)+Kx(Q=0 where wf], ® o4 xr n[OtG, K+K, “1 o and fo HOS {eo} “The solution has the form x()= $e" ‘The associated eigenvalue problem becomes (AM* +K")>=0 with 2n eigenvalues 2; and the associated eigenvectors Since Mis a positive definite real symmeticmatix, equation (11) canbe tedced tothe fom Aga ae and A=(MY'K™ O} (19) ay i) M'G+C,) -M(K+K,)] (3) 1 ° is an arbitrary real matrix. The solution of the eigenvalue problem is complex: A,=0,+jo, (=1,2,...2n) ay ‘The stability boundary can be established by examining the variation in the real part of 2, as a function of an operating parameter. Its very common that logarithmic decrement is sed fo express the degree of the stability instead of using the damping exponent: ~2a0 o ‘When logarithmic decrement becomes negative or the damping ‘exponent becomes postive, the system becomes unstable andthe linear theory is no longer valid. Nonlinear theory should be applied when the system operating parameter exceeds the stability boundary. However, experience shows that in most practical applications, the unstable vibration component grows very rapidly and violently such that the machine cannot be continuously operated without damage in one form or another. b= as) NUMERICAL ALGORITHM At the ueshold of instability, the Cis zero such that Aasraott = JP rraott 6) ‘To determine the stability boundary of an operating parameter, the problem can be stated as one of choosing an operating parameter subject to practical limitation of the design such that ‘one of the system eigenvalues has a zero real part. The ‘corresponding imaginary part is referred to as unstable whirl frequency. In most applications ifthe system becomes unstable, the rotor whils in its first mode lowest whi frequency with, forward precession. It should be noted that the fist unstable rode can be either a bending mode or a rigid body mode, depending upon the potential energy distbution. The numerical algorithm can be writen mathematically as follows: To find an operating parameter, p, which an andissubjectto p! < p< p! (as) The operating parameters considered in this paper can be the ‘rotor rotational speed, lubricant viscosity, bearing loads, bearing clearance, and axial loads, ec, The optimization procedure is terminated at the ih iteration step if the objective function, ‘F(p), satisfies the following conditions: S(p) < Epsilon a9) IF (PY ~ f(py" I< Epsiton (20) ‘where Epsilon is a small number defined by the users. If the convergence equation (19) is satisfied, the instability boundary is then established. If the procedure stops due to convergence equation (20), no instability boundary will be established and the system is either in a stable or an unstable state throughout the limits of the operating parameter. The feasible direction method (Vanderplaats, 1984) is employed as the optimization solver. The gradients required in the iterative process can be obtained analytically in the following expression (Rajan et al., 1987) 1 OM" | aK" aoa, By y, saint: Airey tape ey 2 Fi where §, @, are the corresponding right and left eigenvectors, respectively, The right eigenvector is defined in equation (12) ‘and the lft eigenvector is defined in the following equation A= 2g" (22) “The associated modal nom is defined as R= Ad, @3) ‘The matrices in equation (23) are quite sparse, Computational procedures may be adopted which allow for the ealeulation of Cigenvalue sensitivities with minimum computational expense (Rajan et al, 1987). To determine the instability boundary, only the real pat of equation (21) is used. 218 PARAMETRIC STUDIES. AA single journal bearing system (Kirk and Gunter, 1970) is used for the parametric studies. The physical parameter of the system are summarized in Table 1 Joumal Mass = 22.766 Kg (0.13 Lbfsee fin) Bearing Length = 12.7 mm (0.5 in) ‘Bearing Diameter = 50.8 mm (2.0 in) Bearing Clearance = 0.0635 mim (0.0025 in) Bearing Load = 222.4 N (50 Lb) oil ‘89478 Centipoise (1.006 Reyns) Table! System Parameters or 8 single joural bearing system, the rotor is assumed to be rigid and symmetric which is usually not prectical, The stability of @ rotor-bearing. system depends on the bearing characteristics, bearing locations, rotor stillness, and operating speed, ec. Therefore, fora given rotating system it is nevessary to determine the stability toundaries by using a complete system ‘model inchuding rotor, bearing properties, and all other effet. ‘This simple two degrees of fteedom system is only employed to demonstate the validity ofthe numerical procedures proposed in this paper. The results obtained fom the algoritm presented here cam be easily verified by using the conventional graphical method Using. the analytical procedures described. catler, the instability threshold can be found to be around 14025 rpm by considering the rotational speed as the operating parameter. For the purpose of comparison, the whit speed and stability maps sown in Figues 1 and 2 are generated by calculating sytem cipeavalues over a range of rotational speed. Figure | shows thatthe backward mode is areal non-vibratory mode with zero ‘hi frequency from zero to 4615 gpm and becomes a vibratory mode after 4615 rpm. The forward mode is always a vibratory mode throughout the speed range. Figure 2 shows that he forward mode becomes unstable asthe spend exceeds 14025 ‘pm. The instability threshold determined by using the proposed ‘rovedure isin good agreement with the result obtained frm the Stability map. A parametric study is then carried out to understand the influence of the individual parameter on the instability dueshold. Figure 3. shows that the instability threshold decreases as the oil viscosity increases and itis very sensitive atthe low viscosity range (i. high temperature range), Figure 4 shows that decreasing the bearing load monotonously lowers the instability threshold. Figure $ shows that the lowest instability teshold occurs when the bearing clearance is around 0.0575 mm and the system stability can be improved by either decreasing or increasing the clearance fom that point ‘At 12000 rpm, the stability boundaries of the bearing lad and ol viscosity were calculated to be around 185 N and 9.1 Centipose, respectively. The logaritumic decrements vers bearing load and cil viscosity at 12000 rpm are plotted in Figures 6 and 7 forthe comparison purpose. Figure 6 shows that the logarithmic decrement ofthe forward mode decreases asthe bearing load decreases. On the contrary, the logarithmic decrement of the backward mode slightly increases as the bearing load decreases. The forward mode becomes unstable so00 A TaN te & & i F Fowad Moie yo Pay B soo fF e e 2 sao 3 g 2 2 sow ° como 0 mo wee mw Rotation! Spee (pn) Bearing Loed 0) FIG. 1 WHIRL SPEED MAP FIG.4 INSTABILITY THRESHOLD VS. BEARING LOAD % sot i -& 25000 i ia Ene be seo 210 B Bacon Mode 2 3 owe 5} subte a Ronit so . eee ere ee ey 0 ve 10 pees Pailin elena Rotations! Spe pm) ‘Bearing Clearance (mm x 10°2) FIG. 2. STABILITY MAP FIG. 5 INSTABILITY THRESHOLD VS. BEARING CLEARANCE 20180 ‘0 qn0m oo i Eo ap] B-Band Mode B sc00 La é & tow & 1.00 2 ge 3 0m on fh Fo Mele ° a0 co? ter wn hw Hm» osm i Visor (Cece) Bearing Load) FIG. 2 INSTABILITY THRESHOLD VS. OIL VISCOSITY FIG. 6 LOGARITHMIC DECREMENTS VS. BEARING LOAD 218 Oil Viscosity (Centipoise) FIG. 7 LOGARITHMIC DECREMENTS VS. OIL, viscosity ‘when the bearing load is less than 185 N, Figure 7 shows that the forward mode becomes unstable when the oil viscosity is larger than 9.1 Centipoise. The logarithmic decrement of the backward mode increases and the logarithmic decrement of the forward mode decreases as the oil viscosity increases. The logarithmic decrements of the forward and backward modes ‘move in the opposite directions as the operating parameter varies. Under normal circumstances, the forward mode becomes ‘unstable. This effect can be mathematically proven by using the ‘energy approach (Lund 1990). APPLICATIONS TO AN INDUSTRIAL COMPRESSOR To illustrate the practical application of the algorithm, a high speed centrifugal compressor was employed as a test vehicle ‘The single overhung rotor system as a length of $50 mu and @ bearing diameter of 55 mm. The rotating assembly weighs about 17 Ke and is supported by two 3-lobe bearings as indicated in Figue 8. ISO VG 32 mineral oil was used in the test. The bearing loads are mainly due to the gear force which varies from 110 Kg to 450 kg depending on the loading condition. The oil temperature regulating valve was disabled during the test in ‘order to manually adjust the ol temperature. Two perpendicular vibration probes were located at station 4 between the impeller ‘and bearing, The effects of the bearing loads and oil temperature on the stability were studied. The first three forward synchronous rigid bearing critical speeds are 17730, 80720, and 177260 mpm. The rotor is operated at a constant speed of 29700 rpm which is above the first bending critical speed, In the analytical model, an aerodynamic cross-coupling of 350 Nim was included in the finite element station 2. The

    You might also like