You are on page 1of 7

JOURNAL OF RESEARCH of the National Bureau of Standards-B .

Ma thematics and Mathematical Physics
Vol. 64B, No. 2, April- June 1960

Upper and Lower Bounds for the Center of Flexure
Lawrence E. Fayne
(Febru ary 1, 1960)
There is d isagreement in t he li terature as to how t he center of flex ure of an isotropic
elastic beam should be d efined . In either of t he two most widely accepted definition s, upper
a nd lower bounds for t he coordin ates of t he ce nte r of fl ex ure ar e obtained .

1. Introduction
We consider a cantilcver beam of uniform cross sec tion and choose the z-axis to coincide
with the line of cross-section centroid s. The plane end z = O is assumed fixed and the other
end z= l is loaded by some distribution of forces whi ch is sta tically eq uivalen t to a single force
with components (vVx , W y , 0) acting at a lo ad point (XL , YL, l) . In the treatment of the flexure
problem [9]1 (a similar tr eatment appears in [3]) we seek a load point (XfJ YfJ l) which has the
property that any load (Wx, W11' 0) applied through th is point produ ces no local twist at the
centroid of the section, i. e., the m ean value of the lo cal twist over the cross section vanishes
(see [9] p. 200) . This point (XfJ YfJ l) is r eferred to as t he center of flexure [9 ]. The determination of (XfJ YI, t) p ermi ts us to break up the general flex ure problem into two separate problemsone of pure torsion and one of pure fl exure. This point is som etimes referred to as the cen ter
of sh ear (see, for instance, [3]), and it is well known that there is difference of opinion in th e
literature as to the way in which the point should be defined (see, for instance, [1 , 2,4,10, 11].
A disc ussion of these differences has been given in a r ecent paper by P earson [7], who demonstrates that in general th e two expressions for the center of shear yield resul ts which are very
near to one another.
It is not our poin t h er e to en ter into the di sc ussion of the relative m eri ts of the two definitions. We merely point ou t tha t the expression of Trefftz [10] is simpler and ind epend en t of
Poisson's ratio . The center of flexure of [9], on the other hand, p ermits an easy decomposition
of th e general flexure problem.
In order to avoid confusion we shall her eafter refer to th e point defined in [9] as th e cen ter
of flexure and that given by Treff tz [10] as the center of sh ear .
In this p ap er th en we derive upp er and lower bounds for t he coordinates of t he cen ter of
flex ure (XfJ YfJl) and the center of shear (x s, Yo, l ). In the first case we use the formulation of the
flexure problem derived in [9] and thu s concern ourselves wi th the problem of pure flex ure.
It is of course well known (see, for instance, [7]) that if we have the solu tion to the pure
torsion problem for the b eam in question then we can determine both the center of sh ear and
the center of flexure without knowledge of the solution to the flexure problem . As would be
expected it is possible in each case to obtain upp er and lower bounds for th e coordinates of the
point by approximating the solution to the torsion problem alone.

2 . Bounds for the Center of Flexure
We denote by 9} the cross section of the beam and let C be its boundar.v. (The r egion
m ay be simply or multiply connected. ) The origin of a rectangular coordinate system is
taken at the centroid of ~. For simplicity the x and Y axes ar e chosen to be principal axes of
inertia, i.e.,


xyclA= O.


Figures in brackets indicate t he literature references at thc end of this paper.


(2. 1)

) The coordinates of the center of flexure (XI> Yr) are defined by: xr 2(1~0")Ix ff ~~2_y ~. We note now that the coefficient of '1'2 in the boundary integral in (2.4) 2(1~0")Iy ff[ x ~~l_y ~:I+(1 +0")x2y.8) Using the divergence theorem we have 2(1 + 0" )Ixxr=~'1'2 (xny-ynx)ds.0 0<. ox" oy2 ~= [(1 + 0" )x2-O"y2]nx (2.-(1 +0")xy2+ O"x3JdA.7) Thus (2.02+ 0'1' O<'o2J ox ox oy oy dA. (2.11) Thus if we let (2. (The normal is assumed directed outward from 9). (2.9) is proportional to the boundary value of the normal derivative of the warping function 'I' in the pure torsion problem.14) ~ The determination of bounds for Xr is thus equivalent to the determination of bounds for 106 x.13) where D( '1'.8) where 0" is Poisson's ratio and nx and ny are the x and y components of the unit normal on C.J c (2.9) J = ff[(1 + 0")x y2 _ O"x 3JdA.O"y31dA (2.According to [9J the pure flexure problem is solved once the solutions to the following boundary value problems have been determined: i)2<'o1 ()2'1'1 ()Z":. .12) we find (2.6) (2. (2 .2) ox 2+ oy 2 =0. In fact on C.10) where ~ Since J is a geometric quantity it can be computed explicitly.2+ 02'1'2=0 in ~ (2. (2.5) [x 9) Yr ~ where (2. '1'2 )=55[°<.

and v is any function continuous on ° a/os denotes the and satisfying the condition tangential derivative on ~ vds= O. ] 5) y c If cp is not known we in trodu ce two arbitrary harmonic fun ctions u and U2 and defin e al = -D( .23) to the boundary integral of the square of the normal derivative of cp-u.CP2) = ~U[ (1+ cr )y2_ crx2]n yds.a2+ a3=.U2)= -~U2(ynx-xnIJ dS.20) in terms of the boundary integral of the square of its normal derivative.22 ) where L denotes the length of the perimeter of 0. 107 . (2 .21 ) u)ds= O. the differ ential equation which arises as the Euler equation for the minimum of the quotient pV ds. c It follows then that c (2. a2. 19) c An application of Schwarz's inequali ty yield s (X-al-a2+a3)2::.D( cp. (2 . 17) c (2.U) { [(1+ cr )y2. To this end we note that (2. 0. 2 (~(ov/ os)2ds). By adding an appropriate constan t (if necessary) it is possible to choose cp---"U in such a a way that ~ (cp- (2. (2.23) We seek now an inequality which relates th e integral on the right-hand side of (2 .u 2) = -~(cp. ~ (cp-u)2dS~ { c [(1 + cr )y2. - (2.Note that if th e warping fun ction cp is known the value of mce X= XI can be determin ed explicitly -~cp[( 1 + cr )y2_ crx 2]n ds.crx2]n y - ~:2 }ds. and a3 explicitly.20) in terms of known quan tities. (2. We introduce therefore.22) is just th e eigenvalu e of the vibrating string equation .18) With U and U2 prescribed we can determine ai.u. We now form the expression x .D (U.20) c It remains now to approximate the first boundary integral on the rig h t of (2 . The quantity c on the right of (2.1 6) c a2= .cpz. c With cp-----1L so chosen we shall estimate the first integral on tb e righ t of (2.crx2]n y - ~~: } 2ds.

ow ::.W .ow d D( ) ::. 8]).27 ) and (2.30) 2 c c We now insert (2. (2. make use of the Schwarz inequality and obtain where Pmin is the minimum value of P on C.2Jf ::.Igr.29) c In view of (2. (2 . we can find a constant l' such that (see [6. 27r j-h(OW)2 as dsJ' [rt. p.l= 1.26). eq (2..30 ) back into (2.(OW)2 j on ds JJo '.fkOXkow ow ds+Jf ofkIgrad wl2dA-2Jfo fk ow ow dA .26) Since the functions i k are prescribed.2. +[ -47r'lL" 2--on dsJ!)2 +p(OW)2 -on ds pOlin c 108 c ) (2. The i k are now assumed to be so chosen that (2.25 ) A decomposition of the derivatives of u in the boundary integrals into normal and tangential components yields (see [8]) thp (OW)2ds=rhp (OW)2ds +2rt. l' W. This inequality yields P p(OW)2 'I~!.29) yields L [r D(w. ( arbitrary continuous vector function with components R ellich identity (see l6. c (2.32) .28) which yields by Schwarz's inequality D(w.27) 1 !!) (» We make use also of the ordinary Green's identity (2. Y as j on j os as on OXk oXz OXk oXz c C ::» c !!) (2. 555]) ofk::. jw 2ds )t(rh(OW)2 j on ds )!''. Letting w=cp-u.a d W2dA .23). we have i k and consider the generalized where Ll denotes the Laplace operator (Ll= (02 joX2) + (02joy2» and summation I S to be carried out over the repeated indices k .w):::.w):::.JJ·Ofk UXk U XI U Xk uXz a:::.

Yo) in ~.(L _.) W e now use the Rayleigh-Ritz technique to make the two integrals on the right of (2.32 ). and (2 . (2.:2 ~c [ : s (v-t1'2) ] 2ds ~ A{[(1 - cr)y2.32) it is thus always possible to oMain the inequality (2.36) at every point on C.38) 2 c c In the general case t hen . (2.+-. an inequality of the form is obtained. (2. 109 . . .41) we obtain (setting U2=0 ) Ix.Yo) is equivalent to the condition that (2..23) into (2. if we use instead of u a conjugate function v defined by2 (2.From (2. in this case. q.37) The left-hand side of (2 .31).crx2]n y }2ds.33 ) (see [5]) tllat ~ (OW)2 [ 10 1'1 ] 1.27 ) t hen vanishes identically. In fact . Note that it is possible to derive close upp er and lower bounds for XI without approxima ting the flexure function at all .39) and (2.35) 27rP~in max ]) 2 If the boundary C is star-shaped with respect to some point (xo. 1'2 (OW )2 Y Z)S ds '5. 20 ).42) c In this case the A term has been put in the second integral since this expression is a purely geometrical quantity (independent of v) and can be computed explicitly. wher e A is a completely determin ed positive function.33) ·where (2.34) and (3 = Ji)Xki} { -i os ]. The fact that the boundary is star-sh aped with respec t to (xo . the problem is simplified considerably. Thus. (2.22). and h ence we take "(=0 in (2. (2.~tQ 1+ os m ax 'Y q on ds. we find Inserting (2 .25 ) is satisfied if we let (2.35).40) small. (2 . 2 The fUllction v is not to be confused with tbat used in (2.a212'5.40) (It is apparent that the quantity A may be placed in the second integral rather than in the first if it is desirable. ' ''iTe find in this case instead of (2 .

-xn v]ds.6) and two harmonic functions hi and h2 • Let b1 =ph1 x 2n xds. j e 2 3 10] is G =6t P [3x2nxlO -x3(ynx-xny)] els. th en the same minimizing function U (or v) minimizes the error term in each case.. Note that if wc choose UJ.l) as defined by Trefftz [10] .2 (3.U2=O and merely approximate the warping function (or the conjugate function) on the right.5) We introduce the notation (3 . 110 (3..6I y rf.8) (3 .43 ) t.1 [3xnxlO-x 0on YS.y.2) Ys=* JJlOxdA.In an entirely analogous way we find that J .101)' (2. The coordinates Xs and Ys are given by (3.2(1 + u)I YYf + J[(1 + U) X2y . e c b3= ph1 e ~~ ds. (3.2) may be rewritten as (3.2 By introducing the two arbitrary harmonic functions u and Ul and proceeding as before.9) .1) Xs=-i JJlOydA t. (3. 3 . 9 Equation (3. Bounds for the Center of Shear We look now at the expression for the center of shear (x s.3) By Green's formula we have . we obtain upper and lower bounds for Yr.7) e b2=plO ~~ ds = ph2[yn.uy3]dA= D(IO.4) e The divergence theorem then yields (3.

Math. 272-280 (1944). WASH INGTON. [3] A. In a similar way i Lis possible Lo obtain upp er an d lower bound s for Lh e coordinate Xs .Y. App l. [2] J . Qu art. 551. F. '16. u. Cicala. Pacific J . Weinberger . Mech . 4. Sci.. Math . N . Note that jf Lhe bo undary valu e p roblem on C (3 . [8] F. Aeronaut. Sokolnikoff. Rellich. ~ ow York. 97. 36. Love. Phys.T. Math.b2 + b3=~ (II'. N.64 (1943).hl) (x 2n X.Combining (3. Goodier. 62. accad. it is t hen possible Lo determine Ys explicity wi t hout solving eiLhcr the torsion or t h e fl exure problem . [A] 10. I1l] A. F. Osgood. Mech . pp. Z. (P ap er 64B2-28) III . '1. [1 0] E. 291. Press. E. 1956) . 356-371 (1935).23) and (2. E.13) "'iVe inser t Lhe k nown boundary value for II' on th e righ t-hand sid e and usc t he R ayleigh-Ri tz techniq ue to obtain close upp er and lower bounds for Ys.11 ) Again it is desirable to choose (lI'. pp . 220. pp. Atti. 10) Then by Schwar z's inequali ty (3 . a ngew.9 ) we on tain <I> - b1 . Weinberger. Pearson.646 (1940) . 8.h2) ds. pp .225 (1935). we ob tain finally (3 . angew. pp. 94--96 (1956) . To ri no 70. J . In c. 635.307 (1955).3 9).99 (1947). Math . [4] W . E ngla nd . References [IJ P . Mathematical theo ry of elast icity. sci.12) Using (2 .C. [9] 1. Ap p!. H. pp.573 (1958). [7] C. u. [6] L.. [5J L. p p. 5. c (3. Math . E. Math .14) can be solved. R. pp.. . A treatise on the m at he mat ical theory of elasticity (Cambr idge Univ. J. E. 11. P ayne a nd H. Z. L ondon. S.h1 ) in s uch a way that (3 . D. 1927) . Weinstein. 2d ed . Math. Z. 15.5) to (3 . Pay ne and H. Treff tz. pp. (McGraw-Hili Book Co.