Sei sulla pagina 1di 10

Journal of Sound and Vibration(1978) 60(l), 1I-20

A FINITE ELEMENT

FOR

THE VIBRATION BEAMS

ANALYSIS OF

TIMOSHENKO

D. J. DAWE
Department of Civil Engineering, The University of Birmingham, Birmingham B15 2TT, England (Received 11 February 1978)

A Timoshenko beam finite element is presented which has three nodes and two degrees of freedom per node, namely the values of the lateral deflection and the cross-sectional rotation. The element properties are based on a coupled displacement field; the lateral deflection is interpolated as a quintic polynomial function and the cross-sectional rotation is linked to the deflection by specifying satisfaction of the governing differential equation of moment equilibrium in the absence of the rotary inertia term. Numerical results con6t-m that this procedure does not preclude convergence to true Timoshenko theory solutions since rotary inertia is included in lumped form at element ends. The new Timoshenko beam element has good convergence characteristics and where comparison can be made in numerical studies it is shown to be generally more etkient than previous elements.

1. INTRODUCTION

The classical Bernouilli-Euler theory predicts the frequencies of flexural vibration of the lower modes of slender beams with adequate precision. However, because in this theory the effects of transverse shear deformation and rotary inertia are neglected the errors associated with it become increasingly large as the beam depth increases and as the wavelength of vibration decreases. The effect of rotary inertia on beam frequencies was first considered by Rayleigh [l] and later Timoshenko [2,3] extended this to include the effects of transverse shear deformation. In Timoshenko theory the rotation of a cross-section, 6, is the sum of the shearing angle, JI, and the rotation of the neutral axis, dw/dx (where w is the transverse displacement and x is the longitudinal axis). The problem is thus governed by two variables, w and 6 say, rather than by w alone as in Bemouilli-Euler theory. Exact solutions of Timoshenko s equations are available for simply supported beams [4,5] and for beams with other boundary conditions [6,7]. Results obtained through use of Timoshenko equations have been shown to be in good agreement with results obtained by using the classical equations of theoretical elasticity [3, 81. A variety of finite elements has been presented [9-201 for the dynamic analysis of beams in which shear and rotary inertia effects are significant. The basic Timoshenko beam element has a total of four degrees of freedom (Wand 8 at both ends) and can be based on a coupled displacement field which exactly satisfies the governing equations of a beam which is unloaded between its ends. Higher order elements are based on independent polynomial assumptions for the bending and shear deformation; these various models have more than two degrees of freedom per node and/or more than two nodes. The existing Timoshenko beam elements are discussed in the next section. In the present work a Timoshenko beam element having a total of six degrees of freedom is presented. The element properties are based on the representations of w and 8 by quintic
11 0022460X/78/0908-001 1 SOZ.OO/O 0 1978 Academic Press Inc. (London) Limited

12 and quartic polynomial through consideration freedom are the values Numerical studies are element compares with

D. J. DAWE

functions respectively, but the of the differential equations of of w and 0 at the two end points presented to show the way in that of existing elements.

polynomial coefficients are coupled equilibrium. The element degrees of and the centre point of the element. which the performance of the new

2. REVIEW OF EXISTING ELEMENTS

The four degree of freedom basic Timoshenko beam element has appeared in a variety of guises [9, 10, 12, 14-17, 191 and, as noted by D. L. Thomas, Wilson and Wilson [18], some confusion has arisen both due to errors of detail in published work and in some instances, to misconceptions as to which rotational degrees of freedom should be used at the nodes. The correct nodal degrees of freedom are values of the transverse displacement, w, and the cross-sectional rotation, 8, but in some cases [14, 171 dw/dx has been used in place of 8; this does not allow the correct representation of the clamped boundary condition or the correct coupling of adjacent elements in general situations. Davis, Henshell and Warburton [16] and Narayanaswami and Adelman [19] have shown that the stiffness matrix of the basic element can be based upon the exact solution of the differential equations of static equilibrium of a beam unloaded along its length except at its ends. Under these conditions the shear force is constant along the length of the element and is equal to the first derivative of bending moment. Hence bending and shear deformations can be coupled and no extra degrees of freedom are involved in the beam analysis; the lateral deflection is a cubic polynomial function of the longitudinal co-ordinate and the crosssectional rotation is a (dependent) quadratic function. The philosophy has something in common with that adopted by Egle [21] in deriving a simplified version of the Timoshenko theory for the beam as a whole. The general use of the basic element stiffness matrix and associated consistent mass matrix invibrationapplications has beencriticized by J. Thomas and Abbas [20] whoargue that it is invalid to use this Timoshenko beam element in vibration work where the shear force due to dynamic loading is clearly of variable intensity. However, although the governing equilibrium equations cannot be satisfied at the microscopic level within each element this does not mean that equilibrium conditions are not satisfied at the macroscopic level (at the nodes) for an assembly of elements, and that convergence to true Timoshenko theory energy levels will not occur with use of an increasing number of elements. The situation is not unlike that which exists in finite element vibration analysis based on Bernouilli-Euler theory where the well-proven basic element [22], with a cubic displacement function, is also implicitly based on the exact satisfaction of the governing differential equation of the beam element carrying no load between its ends. The earliest higher order element was presented by Kapur [l 1 ] and has a total of eight degrees of freedom. The element is based on considering the total transverse displacement as the summation of parts due to bending and due to shear. The variation of each of these displacement parts along the element is represented by a cubic polynomial function and the degrees of freedom at each end node are the values of the bending and shear displacements and their first derivatives. Nickel and Secor [17] describe a seven degree of freedom element based on a cubic function for w and a quartic for 8; the degrees of freedom are w, dw/dx and 0 at the two end nodes plus 0 at a mid-point node. Carnegie, J. Thomas and Dokumaci [13] base an eight degree of freedom element on cubic functions for both w and 0, with values of these quantities as degrees of freedom at four node points located at each end at the one-third points. An element due to D. L. Thomas, Wilson and Wilson [18] has six degrees of freedom with values of w, 0 and $ as degrees of freedom at the two end points;

FINITE ELEMENT FOR TIMOSHENKO BEAMS

13

these degrees of freedom are associated with an assumed cubic variation of w and a linear variation of $. Finally J. Thomas and Abbas [20] have described an element having eight degrees of freedom, comprising values of w, dw/dx, 8 and d0/dx at the element ends; these degrees of freedom are selected so that all standard displacement and force boundary conditions can be imposed at beam ends. Since this element is based on assumed independent cubic functions for w and 8 it is, as noted by D. L. Thomas [23], simply a new form of the element of Carnegie, J. Thomas and Dokumaci [13], obtained by a suitable transformation of nodal displacements. However, because of the different conditions applied at element junctions and beam boundaries the numerical results obtained by using the two forms can be quite different [23, 241; on an element basis the earlier version is superior but on a degree-offreedom basis the latter version is more accurate. Of the existing higher order elements all but that of Carnegie, Thomas and Dokumaci [13] involve use of quantities at the end nodes of the element other than those which are strictly required by the variational procedure. For such elements difficulties are introduced in analysing assemblies of beam elements which are not unidirectional or in analysing complex structures which combine beam and other types of finite element. Some comparisons of the numerical performance of the available Timoshenko beam elements in vibration analyses of single beams are recorded elsewhere [18,20,23,24] and are also provided in section 4 of this paper. 3. TIMOSHENKO BEAM ELEMENT PROPERTIES An infinitesimal length 6x of a uniform beam element vibrating with circular frequency
p is shown in Figure 1 in dynamic equilibrium under the action of the shear forces and bending

moments and the inertia force and couple. Timoshenko beam theory is based on the following set of equations (see the Appendix for a list of notation) : dF/dx = pAwp2, M = EZdB/dx, dM/dx - F = -pZl.lZ?, 8 = dw/dx + $. (132) (3,495)

F= AGKt,b,

Of these equations the first two are the dynamic equilibrium equations, the next two are the elasticity or stress/strain relations, and the last is a compatibility relation linking the geometric deformations. Normally, in applying the finite element displacement method only the elasticity and compatibility relations are explicitly satisfied within an element, of

Iv

Figure 1. Equilibrium of an infinitesimal length of beam.

14

D. J. DAWE

course, and the microscopic equilibrium equations are approximated by the variational procedure. The derivations of the basic, four degree of freedom element by Davis, Henshell and Warburton [16] and Narayanaswani and Adelman [ 191 are based on explicitly satisfying equations (l)-(5) with the right-hand sides of equations (11 and (2) set to zero. This results in an element with constant shear force along its length, linear variation of moment, quadratic variation of cross-sectional rotation and cubic variation of transverse displacement. The element model derived here is based on explicit satisfaction of the homogeneous form of equation (2), that is dM/dx - F= 0, so that on using equations form (3), (4) and (5) a relationship is obtained linking (6) w and 0 in the (7)

EId2 9/dx2 = GKA(tl - dwldx).

This procedure means that the rotary inertia term is ignored in the moment equilibrium equation within the element but the effect of rotary inertia will nevertheless be included in lumped form at the nodes. During the vibration the spatial variation of w and 0 along the beam element is initially assumed to be w=A,+A,x+A2x2+A,~3+A4x4+As~s, 8=Bo+B,x+B2x2+B,x3+B4x4. (8)

By using equation (7) the coefficients B, . . . B, can be expressed in terms of the coefficients A o... As by equating coefficients of powers of x. This procedure yields B, = 120e2A, + 6&A3 + A,, B, = 60&A, + 3A,, B, = 4A,, B, = 24&A, + 2A,, B4 = 5A,, (9)

where E = EI/GKA. It is convenient to express the full list of coefficients A o... As alone in matrix form as f=Rb,

A. . . . A,, B, . . . B, in terms

of (10)

. . As,B,,. . . B4}, b={A,,. . . As} and R is an 11 x 6 rectangular matrix. wheref=(A,. The nodal displacements of the element are the values of w and 0 at the ends 1 and 2 of the element and at the element centre point 3. Applying the boundary conditions gives d = Cb, where d = {w,, e,, The strain energy,
~2902, ~3963).

b= C- d,

(11)

U, and kinetic energy, T, of the uniform

Timoshenko

beam element are

.=ii,:[($~dx+KGA

_si-3 dx),

Tw2dx+I
-I/2

(12)

FINITE ELEMENT FOR TIMOSHENKO BEAMS

15

By using equations (8)-(11) it is a simple matter to express these energies in the usual quadratic forms U=+d kd where
l/2 I/2

and

T=$p2dTmd,

k = (C-l) RT EZ

j- DTD1dx+KGA
-II2 l/2

j D;DJD2dx
-f/2 l/2

RC- 1
RC-

(13)

i and
m=(C-l)TRTA

DfD,dx+Z

I D:D,dx
-l/2

(14)

i -l/2
are the

required element stiffness and mass matrices. The matrices D, to D, are defined as D, = [0, 0, 0, 0, 0, 0, 0, 1, 2x, 3x2, 491,

D, = [l,

x4,x5,0,0,0, 0,01, x, 9, x3,


(15)

0,0,0,0,0, 1, x, x2, x3, a+] Ds = [0,


4. NUMERICAL RESULTS

Results of the application of the Timoshenko beam element to the calculation of the first four natural frequencies of three simply supported beams are given in Table 1. If the beams are assumed to have a solid rectangular cross-section the three quoted values of r/L of 0.008, 0.04 and 0.08 correspond to beam depth/length ratios of 0*0277128,0.138564 and 0.277128, respectively. In each case the effect of shear deformation and rotary inertia on any frequency
TABLE 1

Finite element frequency results for three simply supported beams (K = 0.85, v = 0.3)
values of 1

Values of % error in finite element frequencies for the


rP 0.008 Mode 7 1 2 : 0.04 1 2 3 4 1 2 3 4 following number of elements (in complete beam) A ---2 3 4 6 8 16 32 0.03 0.47 80.88 50.62 0.08 1.19 9445 166.26 0.19 2.20 108.10 66.5 o*Oo 0.04 0.58 0.37 0.00 0.19 140 2.19 o*OO 0.46 246 3.08 o+IO 0.00 0.20 0.05 o*OO 0.02 0.33 1.08 0.00 0.13 062 1.72 o+IO 0.00 0.06 0.00 OGO 0.01 0.09 0.45 OGO 0.08 0.29 0.69 0.00 o*Oo 0.00 o*Oo 003 OGI 0.04 0.13 0.00 0.04 0.14 0.27 0.00 o$IO 0.02 0.08 0.00 0.03 009 0.16 0.00 o*OO 0.01 0.02 0.00 0.01 0.02 0.05 0.00 0.00 0.00 O-01 0.00 0.00 o*OO 0.01 -

-b Exact Timoshenko theory 9.8750 39.278 87.823 154.79 9.5710 35.359 71.657 113.85 8.8397 28461 51.498 75,365

Exact Bernouilli -Euler theory 9.8696 39.478 88.826 157.91 9.8696 39.478 88.826 157.91 9.8696 39.478 88.826 157.91

0.08

16

D. J. DAWE

I2

16

I2

I6

Number of degrees of freedom

Figure 2. Percentage errors in the calculated frequencies of a simply supported beam with K = 045, = 28,461; (c) mode 3, I. = 51.498; (d) mode 4, r/L = 0.08. (a) Mode 1, I.* = 8.8397; (b) mode 2, ,I 1t * = 73.365. -o--, Present element; --;--, basic element [16]; -a--, element of Thomas and Abbas

WI. is clear on comparison of the quoted values based on the Timoshenko and the BernouilliEuler theories. The results of Table 1 confirm that the finite element answers always converge, from above, onto the exact results as given by the Timoshenko theory. To demonstrate this conclusively very many more elements have been used than are needed to yield results of sufficient accuracy

Number of degrees of freedom

Figure 3. Percentage errors in the calculated frequencies of a cantilever beam with K = 2/3, r/L = 0.02. Present element; --O--, (a) Mode I, ,l1 2= 3.500; (b) mode 2,1 = 21.35; (c) mode 3, ,I 1/Z = 57.47. -0-, basic element [16]; -z--, element of J. Thomas and Abbas [20]; -.-A-.-, element of D. L. Thomas ef al. [18];----0------,elementofCarnegieelnl. [13].

FINITE ELEMENT FOR TIMOSHENKO BEAMS

17

for all practical purposes. Convergence to good practical values of accuracy is in fact rapid for all the beams; even for the very deep beam the use of four elements (with 16 degrees of freedom in the whole beam) is suflicient to predict the first four frequencies to within O-7%. As the exact solution is closely approximated, within half a per cent say, a reduction in the rate of convergence becomes evident for some modes. This behaviour is presumably linked with the constraint imposed by the specific exclusion of the effect of rotary inertia in the equilibrium equations at the microscopic level within each element and its inclusion only at at the macroscopic level as a lumped quantity. Be that as it may the overall effect on convergence rate is very small.

Numbar

of degrws

of freedom

Figure 4. Percentage errors in the calculated frequencies of a cantilever beam with K = 0.65, 2= 44.62. Key as for Figure 3. (a) Mode 1, A1/z = 3.419; (b) mode 2, A112 = 18.61; (c) mode 3, A1

r/L = @05.

For the very deep simply supported beam, with r/L = O-08, results for the Timoshenko beam elements of Davis, Henshell and Warburton [16] and Thomas and Abbas [20] are available and these are compared graphically with the results for the present element in Figure 2. For a given number of degrees of freedom the present element clearly compares very well in this problem with the earlier elements. (In Figures 2-5 the quoted number of degrees of freedom are those in the complete beam after application of the boundary conditions.) Results are available in the literature of the application of a number of Timoshenko beam element models to the solution of cantilever beam dynamic problems. Three cantilevers have been considered in references [23] and [24]; these are a slender beam A with r/L = 0.02 and K = 213, a beam B of intermediate thickness with r/L = 0.05 and K = O-65, and a deep beam C with r/L = 0.08 and K = 2/3. The models for which results are documented are the basic four degree of freedom model, the eight degree of freedom models of Carnegie, J. Thomas and Dokumaci [ 131 and J. Thomas and Abbas [20] and the six degree of freedom

18

D. J. DAWE

Number of degrees of freedom

Figure 5. Percentage errors in the calculated frequencies of a cantilever beam with K = 2/3, r/L = 0.08. /2 = 3.284; (b) mode 2,1 = 15.49; (c) mode 3,J. liz = 34.30. Key as for Figure 3. (a) Mode 1, l.2

model of D. L. Thomas, Wilson and Wilson [18]. These results are shown together with results based on the present model in Figures 3-5. The present model appears to be the most efficient for all three cantilever beams.

5. DISCUSSION

AND

CONCLUSIONS

The new six degree of freedom Timoshenko beam element presented here is based on a displacement field which couples the primary variables w and 6 by specifically satisfying within the element the differential equation of moment equilibrium with the rotary inertia term ignored. Numerical results demonstrate that convergence to exact Timoshenko theory solutions does take place since the effect of rotary inertia is included at the end points of elements. The new finite element model is based on the assumption of a quintic variation of w along the element length with, correspondingly, a coupled quartic variation of 8. The new model is clearly related in its philosophy to the four degree of freedom basic model as presented by Davis, Henshell and Warburton [16] and Narayanaswami and Adelman [19]. As with the basic model the new model has the considerable advantage of using only values of the primary variables as nodal degrees of freedom. This facilitates the use of this element in the analysis of framed structures and of general structures involving other types of finite element. Comparison of the performance of the new model, where possible, with that of the basic model and of previous higher order models shows that for practical levels of accuracy the new model is generally superior on a total degree of freedom basis, whatever the relative depth of the beam. Furthermore, with the exception of the model of Carnegie, Thomas and Dokumaci [13], previous higher order models have the disadvantage of using values of

FINIlZEL.EMENTFOR TIMOSHENKO BEAMS

19

other than the primary variables as nodal degrees of freedom. The higher order models are, of course, based on independent assumptions of the spatial variation of the primary variables and consequently considerably more element degrees of freedom are involved for given orders of interpolation. If independent interpolation is to be used then it might be mentioned that recent work [25] on the vibration of plates using Mindlin theory (the equivalent theory in plate analysis to that of Timoshenko in beam analysis) suggests that an increase in efficiency would result if orders of interpolation higher than those used in previous beam models were employed.

REFERENCES 1. LORD RA~HGH 1877 Theory of Sound. London: Macmillan and Co. 2. S. P. TIMOSHENKO 1921 Philosophical Magazine 41, 744-746. On the correction for shear of the differential equation for transverse vibrations of prismatic bars. S. P. TIMOSI-&NKO 1922 Philosophical Magazine~43, 125-131. On the transverse vibration of bars of uniform cross-section. R. A. ANDERSON 1953 Journal of Applied Mechanics 20,504-510. Flexural vibrations in uniform beams according to the Timoshenko theory. C. DOLPH 1954 Quarterly of Applied Mathematics 12, 175-187. On the Timoshenko theory of transverse beam vibrations. T. C. HUNG 1961 Journal of Applied Mechanics 28,579584. The effect of rotary inertia and of shear deformation on the frequency and normal mode equations of uniform beams with simple end conditions. 7. T. C. HUANG and C. S. KUNG 1962 Developments in Theoretical and Applied Mechanics 1, 59-71. New tables of eigenfunctions representing normal modes of vibration of Timoshenko
beams.

8. R. M. DAVIES 1948 Philosophical Transactions of the Royal Society A240, 375-457. A critical study of the Hopkinson pressure bar. 9. R. B. MCCALLEY 1963 General Electric Knolls Atomic Power Laboratory, Schenectady, New York, Report No. DIGISA 63-73. Rotary inertia correction for mass matrices. 10. J. S. ARCHER 1965 American Znstitute of Aeronautics and Astronautics Journal 3, 1910-1918. Consistent matrix formulations for structural analysis using finite element techniques. 11. K. K. KAPUR 1966 Journalof the Acoustical Society of America 40, 1058-1063. Vibrations of a Timoshenko beam using a finite element approach. 1968 Theory of Matrix Structural Analysis. New York: McGraw-Hill. 12. J. S. PRZEMIENLECKI 13. W. CARNEGIE, J. THOMAS and E. DOKUMACI 1969 Aeronautical Quarterly 20, 331-332. An improved method of matrix displacement analysis in vibration problems. 1970 Journal of Strain Analysis 5,239241. Inclusion of shear deformation in the 14. R. T. SEVERN stiffness matrix for a beam element. 15. R. ALI, J. L. HEDGES,B. MILLS, C. C. NORVILLE and 0. GURXIGAN 1971 Proceedings of the Institution of Mechanical Engineers, Automobile Division 185, 665690. The application of finite element techniques to the analysis of an automobile structure. and G. B. WARBURTON1972 Journal of Sound and Vibration 22, 16. R. DAVIS, R. D. HENSHELL 475-487. A Timoshenko beam element. 17. R. E. NICKEL and G. A. SECOR1972 International Journal of Numerical Methoa!s in Engineering 5,243-253. Convergence of consistently derived Timoshenko beam finite elements. J. M. WILSONand R. R. Wnso~ 1973 Journalof Soundand Vibration 31,315-330. 18. D. L. THOMAS, Timoshenko beam finite elements. and H. M. ADELMAN1974 American Institute of Aeronautics and Astro19. R. NARAYANASWAMI nautics Journal 12, 1613-1614. Inclusion of transverse shear deformation in finite element displacement formulations. 20. J. Tr-rohus and B. A. H. ABBAS1975 Journal of Sound and Vibration 41,291-299. Finite element model for dynamic analysis of Timoshenko beam. 21. D. M. EOLE 1969 NASA C&1317. An approximate theory for transverse shear deformation and rotary inertia effects in vibrating beams. 1963 Aeronautical Quarterly 14,224-240. The effect of lumped 22. F. A. LECKIEand G. M. LINDBERG parameters on beam frequencies.

20

D. J. DAWE

on Finite element model for dynamic analysis of Timoshenko beam. 24. J. THOMAS and B. A. H. ABBAS 1976 Journal of Sound and Vibration 46,288-290. Authors reply. 25. D. J. DAWE 1978 Journal Sound and Vibration 59, 441452. Finite strip models for vibration of Mindlin plates.

23. D. L. THOMAS 1976 Journal oj sound and Vibration 46,285-288. Comments

of

APPENDIX
A E F G I K L Ivl T u

: NOTATION

cross-sectional area of beam Young s modulus of elasticity shear force shear modulus second moment of area of beam cross-section shear coefficient beam length bending moment kinetic energy strain energy k element stiffness matrix element length m element mass matrix P radian frequency r radius of gyration of beam cross-section (r = Z/A) W lateral deflection co-ordinate measured along element axis, origin at centre X rotation of cross-section 8 = pAL4p2/EI, frequency parameter I p mass density v shear strain

Potrebbero piacerti anche