You are on page 1of 93
NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS TECHNICAL NOTE No, 1832 SMALL BENDING AND STRETCHING OF SANDWICH-TYPE SHELLS By Eric Reissner Massachusetts Institute. of Technology | 20000803 215 Washington March 1949 : Reproduced From for Public Release DISTRIBUTION STATEMENT A istribution Unlimited Approved Distri Best Available Copy NACA TN No, 1832 TABLE OF CONTENTS Statics of SandwichType Shell Coordinate system on shell Statics of face layers . . Statics of composite shell Streas distribution in core layer Beau uw Strain Energy of SandwichType Shell»... eee eee ee eee Strain energy of face layers +e eee ee eee eee ee 16 Strain energy of core layer... eee ee ee eee eee UT Strese-Strain Relations for Composite Shell .........44 19 II — APPLICATIONS OF GENERAL THEORY 6...) eee ee ee eee ee 25 bret] oan ga0dd0cc000000000n00Gcu0 fy Rectangular plates sv eee eee ee ee eee BD Bending of rectangular plate with simply mupported edges’... 29 er ee eens B Circular plates; rotational aymetry .. 1-2-4522 e+ 36 Deflection of circular plate with built-in edges... +... 38 Cirouler Rings... ee eee ee eee ee ee ee ee ee eee AL Ring sector acted upon by ond tending momenta: 1. sls. 3 Closed ciroular rings acted upon by uniform radial oad |)... bb Ring sector acted upon by radial loads dy and q7, uniform 4n circumferential direction and with vanishing resultant q. 47 Bending of semicircular ring ty end shear forces ....... 46 Bending of complete circular ring under action of two concentrated radial forces at @=4n/2 .+++ee+eee SL Ciroular Cylindrical Shelle . 1. se eee ee eee ee ee ee Bh Axisymetrical deformation of circular cylindrical shell’... 58 Infinite circular cylindrical shell with periodic load dimursbuticoye, menepep es eet ee ereme tet tt Gl Finite circular cylindrical shell acted upon by edge moments and forces sw ee te ee te ee te et et ee 6 Semi—infinite shell acted upon by edge bending moment and shear force sss sees eee eee r errr ee eene Infinite circular cylindrical shell acted upon by Toa etre Nad aga eee geoccouoooG 4h) NACA TN No. 1632 Page Sphevionl fella... ees ee ee te eee eee eee Uniform stress distribution ina spherical shell ....... 77 Reduction of axieymetrical problem to two simitaneous equations for Og and Pg + +--+ sees rere reese D Analysis of edge effect for spherical shell. ......... 83 PUMEES Ge ee ee oe NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS TECENICAL NOTE NO, 1832 SMALL ENDING AND STRETCHING OF SANDWICH-TYPE SHELIS By Eric Reissner ‘SUMMARY A theory has been developed for small bending ani stretching of sendwich-type shells. This theory is an extension of the known theory of homogeneous thin elastic shells. Tt vas found that two effects are important in the present problem, which have not been considered previously in the theory of curved shells: (1) The effect of trensveree shear deformation and (2) the effect of transverse normal stress deformation. ‘The first of these two effects has been known to be of importance in the theory of plates and beams. The second effect vas found to occur in a manner which is typical for shells and has no counterpart in flat-plate theory. The gensral results of this report have been applied to the solution of problems concerning flat plates, circular rings, circuler cylindrical shells, and spherical shells. In each case numerical examples have been given, illustrating the magnitude of the effects of transverse shear and normal stress deformation, The result of this investigation indicate the necessity of taking account of transverse shear ani nornal stress in sandwich-type shells, as soon as there is an orierof-mgnitude difference between the elastic constants of the core layer and of the face layers of the composite sholl. It was found that the changes due to transverse shear and normal stross deformation in the core may be 0 large as to be no mere corrections to the results of the theory without transverse core flexibility. The actual magnitude of the changes 1s greatly dependent on the geometry and loeding condition of the structure under consideration so that no general rules may be given which indicate for which elastic modulus ratio the changes begin to be significant. Solutions of problems in the present theory may in general be obtained by mathematical methods which are similar to those employed in the theory of plates ani shells without the effect of transverse shear and normal stress deformation included. The present work does not include consideration of buckling and finite deflection effects. 2 NACA IN No. 1832 INTRODUCTION In this report an extension of the classical theory of emall ending and strotching of thin elastic shells is considered. Instead of @ homogeneous shell, investigation is made of a shell constructed in three layers: A core layer of thickness h with elastic constants Ey, Ge, and V, and two face layers of thickness t with elastic constants Ep, Gp, ani Vp. In the developments certain restrictive assumptions are made which somewhat limit the general applicability of the resulte. In so doing formas are obtained which fare as compact as possible while still describing the essential characteristics of the sandwich-type shell. ‘The thickness ratio t/h 1s assumed small compared with unity; at the same time the ratio Ept/Ejh 1s assumed large compared with unity, This latter assumption means that the face materiel is so mich etiffer than the core material that the contribution of the core layer to atress couples and tangential streas resultents of the composite shall ie negligible, It is known thet for flat plates these assumptions necessitate the taking into account of the effect of transverse shear Geformation. (See, for instance, reference 1.) The same would be expected to be true for curved sholls, and the prosent report, therefore, gives # system of equations in which this effect is incorporated. A further effect which, it appears, has not been considered previously in the analysis of small deflections of sandwich structures is the effect of transverse normal stress deformation. In the present: report it is shown that this effect arises in a manner which is typical for shelle and has no counterpart in plate theory. It may be likened, roughly, to what happens in the bending of curved tubes. ‘The process by which the general results of this report are obteined is as follows: First, each of the face layers of thickness t fe asgumed to behave like a thin shell without bending stiffness. ‘The Joads applied to these face shells, henceforth called face menbranes, are of two kinds: (1) External loads and (2) loads caused by the stresses in the core layer. Next, the core layer of thickness h is assumed to behave 11ke a thres-dimensional elastic continuum in which those stresses vhich are parallel to the faces are negligible compared with the transverse shoar and normal stresses. On the bagis of these two assumptions three steps are carried out, Firet, the equilibrium equations of the core layer and of the face layers are obtained. Thon fan appropriate expression for the strain onergy of the composite structure ie derived, Finally, Castiglieno's theorem of minimm complementary energy is used to obtain the relations which connect stress reeultents and couples of the composite shell with the quantities which describe the state of deformation of the composite shell. NACA TN No, 1832 3 . The system of equations which is obtained in the foregoing manner is specialized for the following cases: ’ (1) Flat plete (2) Cirewler ring (3) Circular cylindrical shell (4) Spherical shell with axteymmetricel deformation In each case @ mumber of problems are solved explicitly and the eprreciable effect of transverse shear and/or normal stress deformation ig illustrated mmerically. ‘This work was conducted at the Massachusetts Institute of Technology under the sponsorship and with the financial assistance of the National Advisory Committes for Aeronautics. . SYMBOLS 1 a core-layer thickness t face-layer thickness fs bp curvilinear coordinates on middle surface of composite shell. ¢ @1etance coordinate measured along normal to middle surface of shell 4) % coefficients of linear element on middle surface of shell Ry Ry principal radii of curvature of middle surface of shell. Som direct etrese resultanta in upper face membrane; n= 1,25 m= 1,2 Namt direct stress resultants in lover face membrane . Pros Pad ‘tangential components of external load intensity on upper and lover membranes a aus % normal components of external load intensity on upper and lower membranes By, Bo tm 0% = OtEp NACA TN No, 1832 components of transverse shear stress in core layer component of transverse normal stress in core layer values of transverse shear stresses for ¢ = + h/25 n=1,2 values of transverse normal stresses for = + h/2 values of transverse shear stresses at middle surface of shell. ‘transverse shear stress resultants direct stress resultants parallel to middle surface for composite shell; n = 1,2; m= 1,2 stress couples for composite shells n = 1,2} m = 1,2 tangential components of external load intensity for composite shells n = 1,2 normal component of external load intensity for composite shell external load intensity term defined by equation (22) etrain energy elastic moduli of 1eotropic face-layer material; vay, ft elastic moduli in transverse direction of core-layer material effective tangential components of displacement of elements of composite shell effective normal component of displacement of elements of composite shell effective components of change of slope of normal to middle surface of composite shell component of strain (tm = n/Eo) NACA TN No, 1832 5 De = (1/2)t(h + t)2Ry © = c#/(1 — y2) D bending stiffness factor (D = D*/(1 — y2)) Cartesian coordinates in plane of flat plate polar coordinates in plane of flat plate a radius of ciroular ring, cylinarical shell, and spherical shell x, 0 surface coordinates on cylindrical shell das Yes Mae paransters defined by equation (63) baat 1 half wave length of sinusoidal load distribution m, my quantities defined by equation (197) k complex quantity defined by equation (200) g, 6 surface coordinates on spherical shell o quantity defined by equation (7h) x parameter defined by equation (190) ‘T — GENERAL [HEORY Statics of SandwichType Shell In order to derive a complete system of equations for the shell composed of face layers and core layers it is necessary first to consider separately the statics of the face layers ani that of the core layer of the shell. Combination of the resulta obtained for these two components of the composite structure mst and will lead to those differential equations of equilibrium which hold for elemente of @ shell, whether this shell is of homogeneous or nonhomogeneous construction, In addition, however, relations ere obtained which are characteristic of the eaniwich-type shell, Coordinate system on shell.— A curvilinear coordinate system (1, Ea, §) 18 introduced as follove: Tet £] and fo be coordinates 6 WACA TN No, 1832 on the middle surface of the composite shell and let [ be the distance of a point of the shell from ite middle surface, measured along the normal to the middle surface. ‘In order that this system of coordinates be an orthogonal system, choose the £1, fp curves as lines of curvature on the middle surface (in the case of shells of revolution the lines of curvature are identical with the meridians ani parallels on the middle surface). ‘The linear element in the foregoing system of coordinates is of the form 2 2 as? = 22 + Fr) a? + ak(r+ §) ate? + at? @) where aj and ap are the coefficients of the linear element on the middle surface and Fy and Ro are the principal radii of curvature of the middle surface (see fig. 1). Formas for the calculation of the quantities ap and R, are contained in texts on differential geometry. They are collected, together with other results, in reference 2, which deals with the theory of homogeneous thin shells. Statics of face layera.— The face layers aro treated as thin shells of thickness t and it ie assumed that the bending stiffness of these thin shelle about their ovn middle surface may be neglected. Because of thie neglect from now on they will be designated as face membranes. The middle surfaces of the face membranes evidently are given with reference to the three-dimensional system of curvilinear coordinates ty $=RMn+t) and S=- Zn ++). From equation (1) it follows that the Linear element on the midle surfaces of the face nembranes ta given by 2 2 a2 abet? at 2s a2 a het) ato? + at? as’ a2 + ae) Al sake Gigs) ate? + at (2) ‘The componente of external load intensity on the upper end lover nenbranes are designated bY yyy Pays md 9, and PY Py, Pops OMA ayy respectively (fig. 2). The core-layer stresses which act on the upper and lover monbrenes are given as Tyty, Totus amd Oty and by Tbs Tobrs mie, of course, means that no local buckling phenomena are considered in the present work. NACA TN No, 1832 7 and of;, respectively. Finally, the direct stress resultants in the upper and lower face meubranea are designated by Mian, Trou» Noms and Nop, and by My17, Nyo7, Nozz, and Nop7, respectively (fig. 2). There are then three equations of force equilibrium for the elements of each of the tvo meubranes, Writing - bet yy = On{l + a) ont = oa(! - ta) the equations for the uppor-face moubrane are the following”: Miu , eal >a aqua + STEM + thon Sp ~ Meow = + queen (Pim ~ Tafa) = 2 (4) + aquagu(pou — Tatu) = 0 (5) * Meu Sey “he + ufeu + Norn 3g) Maan % X, ‘The corresponding equations for the lower-face membrane are Bap iNy ay 3, da, day, Seton y BMEDD y ryoy FMD —Myoy HEP + ogyeen (ar + tags) = 9) 1, iN. ay Regt, See 4 noyy Hat — mar GLE + raeea(ber + Taba) + 0) 2tnese are obtained from the corresponding equations of reference 2 with Gq changed to dm, end with stress couples and transverse shear stress resultants onitted. To make up for this omission, the loads on tho two meubranes are assimed to act at their middle surfaces; this means terns of the order t/R are neglected (but not terms of order n/R). 8 WACA TN No, 1832 sealig iy *2( a) -an “| =o (9) ‘As bending moments and transverse shears ere assumed not to be acting in the individual membranes the moment equilibrium equations ‘become the eymetry relations Mau = Noru (10) Mer = Nair Before analyzing the state of stress in the core layer it is convenient to see what relations follow from equations (4) to (9) for the composite shell. Statics of composite shell.— It may be seen that, in view of the fact that all face-parellel stresses in the core layer are neglected, ‘the following expressions for the face-parallel stress resultants and couples of the composite shell are obtained: mh @ + bgt) + @ - be (12) Tho = (+ Bet!) + (0 Make*)hen es) va (Matin + (Mn 0 my tg (6 + a - O- 2m] (a5) Me = Be [0 Bena - toy] co) MACA TN No, 1832 9 Mor = M4t{(t aA )man - (0 - Bai$)o| an Yeo = REA( «Mattoon - (0 ~ Bate \toor] (28) In the same way the following expressions are obtained for components of external force and moment intensity: moe (Rae Maan IG Ran 0 = GONE MR OREM we Metl 6 Se (ARE Akt] cn Further, a load term of the following form will be encountered: aera] a ae which bears a relation to equation (20) similar to that which equation (21) bears to equation (19), This last term would represent, for a homogeneous shell, the average transverse normal stress at any station of the shell, assuming that the loads qy and q7 alone are responsible for this stress, For a homogeneous isotropic shell this term is of no importance. For a sandwich-type shell, as vill be seen, it may sometimes be of importance, In order to obtain force and moment equilibrium equations for the composite shell the face-membrane equilibrium equations (4) to (9) are combined suitably, Adding equations (4) and (7), and (5) and (8), respectively, the two equilibrium equations for the force components parallel to the midéle surface of the shell are obtained. In order to reduce them to known form (see reference 2) the following relations are used between the core—layer-surface shear stresses Tht, and Tate and tho transverse shear stress resultants @ and Q. 20 HACA TH No, 1832 (23) PIP (AEE Mae CAE Baan © Bgl Bak) «Ba rata «(2s Aasns] = Equations (23) and (24) will subsequently be shown to be in agreement with the usual definition for the transverse shear stress resultante by consideration of the stress distribution of the core layer. With equations (23) and (24), there are obtained by combination of equations (4) and (7), and (5) and (8) — carrying out addition as well as subtraction — the following four equations: DegHaa « BAPED s nyp Ht — My Ps perl + ma) «© (25)* Salta BAPE ny FP — my, FA + caselG + me) <0 (26) Sean , Bae 4 ayy FA — Moe SH + easel) = Ce ete . AMEE ney SEE tha FEE + ersalte ~ &) = © oe ‘Two further equations are obtained ty adding and subtracting, respectively, equations (6) and (9), Adding equations (6) and (9) and taking account of equations (11), (ik), and (20), there follows: ‘N; Ns ova + e)- 0+ GaGa - 6-Bet)A- sae] 0 (29) NACA TN No, 1832 nu In order that this reduces to the correct equation of transverse force equilibrium a given in reference 2, jes (+B) — Pat (30) ~ sal + aa Pant _ Bot , Bante 3 OED Bquation (30), Just as equations (23) and (24), can again be verified independently by conalderation of the state of stress in the core layer. On the basis of equation (30), equation (29) te written in the form (31)* dan +e anda + F Ry Tee we $2) + oxse0 The last equation, uss of which 1s required for the sandwichtype shell and which has not previously been given, is obteined by subtracting equation (9) from equation (6), Taking account of equations (15), (18), and (22), there results PEG HE) mine [6 Aa Pat + (0 aS) - Bagtlen] <0 (32) Provisionally, there is written Qs re + Bat logy + @- ret) ~ Bator = 2otm (33) ACA TW No, 1832 we and it will subsequently be shown that of, represents the ‘ralue of ot at the middle surface of the shell. Combining equations (33) and (32) yields tq + pL Gf + 2) a= 0 Gh Equation (34) has no relation to the sixth equation of equilibrium for an elenent of the shell which expresses the condition of moment equilibrium about the normal to the middle surface, ‘That equation which, as is known, ie an identity when resultante and couples are expressed. in terms of stresses does not occur in the present derivations, or rather it de conteined in equations (12), (13), (16), and (17), which give explicitly the slight differences between Mp and Noi, and Mp and Mp. Stress distribution in core layer.— In order to verify independently equations (23), (24), and (30), as well as for the subsequent derivation of appropriate stresé-strain relations, 1t is necessary to determine the distribution of stress in the core layer. Assuming that the components of stress 0), dj, and Typ in the core which vould contritute to stress resultente and couples of the composite shell are of negligible importance,3 these components of stress may be set equal to zero and only the components of transverse shear stress and transverse normal stress Tit, Tot, and ot may be retained, The differential equations of equilibrium for these three remaining components of stress in the eysten of curvilinear coordinates defined by equation (1) are obtained, from the general form of these aifferential equations in reference 3, in the following form: (35) (36) 3rt is for this purpose that the order-of-magnitude relation RE,/tBp << 1 is assumed. NACA TN No. 1832 13 BlneG + )G +e] - REG Dr] * ZG i fa] au: (37) ‘The values of the three stress components at the middle surface (¢= 0) are designated by the subscript m, Integration of equations (35) to (37) then gives Tt = ————Afa_ (38) (a+ tf)°@ + t/m) “£ (39) one (2+ t/Re)P(a + £7) Ca) apes aR) ER] ‘The transverse shear stress resultants Q) and Q are obtained from equations (36) and (39) in the forn! (ast) /2 aa wut( + h)s¢ = Anta 2-2 (hat) /2 eR, oR, _ hs t) Tate (a) ‘the integration mst be extenied over the thickness of the core Jayor and also over half the thickness of the face layers, in accordance vith the prior assumption that the stresses Tatu, Tatz, dtu, end ot; may be taken to act at the middle eurfaces of the respective face menbranes, NACA TN Wo, 1832 How, as intended, the proof is carried out of equations (23), (2k), and (30), which vore used to obtain the differential equations for the ‘ composite shell, To verify equation (23), from equations (36) and (39) for the left-hand side of equation (23), the following equation is obtained: Tnbm Tnbm Tempe y— Bet end this, in conjunction with equation (1), verifies equation (23). To verify equation (24) in the same manner, from equations (38) and (39) for the left side of equation (24), the folloving equation ie obtained: _ (ht Tbe a Tho nat Ra and this, in conjunction with equation (41), verifies equation (24). To verify equation (30), equation (40) 4s used to write for the left side of equation (30) 2 (cotita \ , 2 /ertatn \ , 3 foatite \, 2 /eatetm oe srlrlLrLlL te Ry Rp Ry Ry of Petite _\ , a f Atom Stl (nse) F2[, _ (nat? =) i ana this, in conjunction with equation (41), verifies equation (30). NACA TN No, 1832 b ‘The section on the stress distribution in the core layer is concluded by listing the form which equations (38) to (40) for the stresses in the core layers assume for "thin" shells, that ie, for shells for which h/R <1, From equations (38) and’(39), in conjunction with equation (41), 1t follows that a "18 = ESS (ha)* ret rH From equation (40), in conjunction with equation (411), 4¢ follows that £ oa: ayaget = atptty — Fo Ge + 3 ) (3a) It 4e necessary to note for some of the following considerations that, in view of equation (31), instead of equation (li3a) there may be written Mr , Mee _ % = oy - Lo + ) (430) Tt is seen that in this approximation the transverse shear stresses are uniform across the. thicknoss of the core layer, while the trensverse normal stress is composed of two terms, one uniform across the thickness and the other varying linearly acrose the thickness. No further calculations are needed with reference to the state of stress in the composite shell. ‘he next step is to complete the system of differential equations for stress resultants and couples by deriving an appropriate system of stress-strain relations, Strain Energy of SandwichType Shell In calculating the etrain energy of face menbranes and core layer it is assumed thet both are isotropic and elastic, with elastic costants Ep, vp =v, Gp = Ep/2(1 +) and Me, Ve, Ge = By/2(1 + Vo). 6 NACA TN Wo, 1832 Poteson's ratio for the face menbrenes is written without @ subscript, pecaise, in view of the assumed stress distribution, there 1s no explicit cecurrence of Poisson's ratio v, for the core layer. The strain energy for the composite shell is the sum of the strain energies for the face menbranes and for the core layer n= np + ie (ua) For the purpose of obtaining stress-strain relations, both xp and xo are expressed in terns of stresses rather then in terms of strains. strain energy of face layers. Considering that the element of area on the middle surfaces of the menbranes is of the form aqeo(t & hy *) @ £ 2a) at, dip and that the stresses im the membranes are the stress resultants divided by the membrane thickness +, there ie, from well-known principles, the following relation: np = aff May? + Nopu® — @Myzulloon + 202 + WMyex'] «MEE Mago + al dein? 4 Mop? — 2M WNopy + C1 + wer] = @ - »eta = hg aree af, dtp (45) Equation (45) is transformed into an expression containing stress resultants and couples of the composite shell by means of equations (11) $0 (18) viich lead to the relations NACA TN No, 1832 ae 26 + Baas = Ma tee Ma (46) 2G -*at*mar =~ poe a with corresponding formas for Myp and Nop.” In thie transformation the cases are Limited to those for watch b/) into equation (49) yields A 2 sa + alee LG +6] (na t)ea, Bo] SB H+ ER” Re x abaya aby dbp (50) ‘The integration with respect to $ is carried out and equation (50) becomes role 242 \, 2 FER ape GB HE peo at ome Tt was to be expected that the terms containing the modulus of rigidity G, would occur in the foregoing form, The contribution of NACA TN No, 1832 19 the present report up to this point, besides giving the new equation (34) for tm, 18 thought to be the determination of the form in which the effect of transverse normal stross deformability manifests itself in the strain energy of the sandwich shell. Strese-—Strain Relations for Composite Shell In vhat follows a system of stress-strain relations for the com Posite sholl is obtained by the use of Castigliano's theorem of minimum complementary energy. The manner in which the theorem is used here appears to have been employed first by E, Trefftz (reference ) for the purpose of avoiding geonetrical considerations in the derivation of the stress-strain relations for thin homogeneous shelle with enall defor- nations, without consideration of the effecte of transverse shear and normal stress deformation. Assuming for the present purpose that all boundary conditions for ‘the sholl under consideration are stress conditions, the theorem consiste in the statement that among all statically correct states of stress the actually occurring state of stress makes the strain energy of the system a minimm, In the application of the theorem the fact is taken into account that statically correct states of stress only are to be compared, by means of the Iagrangian multiplier method. Before minimizing x« an integral is added to it which contains the six equilibrium equations (25) to (28), (31), and (34), each of the six equations mltiplied by a Iagrangian miltiplier. it cen then ‘be shown, by using Castigliano's theorem with prescribed boundary displacenents instead of with prescribed boundary stresses, that each of the six mltipliers has the meaning gf one of the displacement quantities which occur in the shell problem.’ With the foregoing understanding of the meaning of the mltipliers, the miltiplicr of equation (25) is designated by uz; that of equation (26), by ug; that of equation (27), by 1; that of equation (28), by Pp; that of equation (31), by ws and finally that of equation (34), by Kk, It 4s imovn that uj, up, and v represent the effective components of displacement in the §), >, and ¢ direc- tions, reepectively, Further, it is known that By and Pp represent the angles through which the normal to the middle surface of the shell turns toward the $; and £2 curves, respectively, There is no Svor the special case of the flat plate thie has been carried out explicitly in reference 1. For the case of the homogeneous shell, without effect of trensverse shear and normal stress deformation, the proof has been given in reference 4. Ths proof for the more general case which is here considered is not included ee it does not offer any clearer insight into the problem and tends to lengthen the analytical discussion, NACA TN Wo, 1832 20 tmmediate atmple goometrical interpretation for x and, while such Interpretation in terms of an average transverse normal’ strain might : be deduced herein, k 1e conoidered as an auxiliary variable presently to be eliminated. Combining now equations (44), (48), (51), and (25) to (28), (31), and (34) in the manner indicated, the following variational equation results: 1 en {a aa? + Wop? - eum ating + 2(2 + VIM D4] + Delia? + Moo? — evihyMop + 2(2 + v) Myo” Moo’ Mee he Xoo aaa 2) ayag ab) dbp + ayap(m — %)| + Bo| + asofe- «] +S afte + RGR )- 4 at, ato = 0 (52) NACA TN No, 1832 aL The variations in equation (52) are carried out and integration is done by parte to eliminate derivatives of variations in the double integral. The line integrals along the boundary which occur due to this integration by parte vanish, because it has been assumed that all strosses are prescribed at the boundary ani therefore their variations vanish at the boundary. ‘The resultant variational equation te Nya — veo dur _ ug aay ow + Noo tn aie a BE als a + ic (@t + HE) Fo +h ke th + Re a] * fi, 2a — gp 282 , Fe B02 * mts Se 4 oe, * Bes 8% nade] vel oe . Poa Sty + ag ayop aby aby = 0 (53) 22 TACA TN No, 1832 As all nine variations in equation (53) are independent of each other, it follows that the contents of all nine brackets in equation (53) mst venish separately. ‘Thus the following nine stress-strain relations are obtained for the sandwich shell, indicating with an asterisk those which appear in final form, maf, , (r+ tor) _ Neel, _ (ns toe om reg?) \ 2B Brod fen es ste &1 36, * sad Be * Ry * 1eECRL (oh) fg oa (55) 20+) wy, = DB) , 2 D (2 oe he 2e@ +2t@ (56)* ee a a (57) Mpg Yew se Chey ae] Oe eee DR = a6 Bip * Hop BE, He Hs OR (38) 20+) yy, 2 (2), 22 B (Be * ar? he = BG) * Sake) NACA TN No, 1832 23 2 eg dw 4 Gre ta, ar 8 (60)* eh hw wang ee 8e-# (a+ a k Ss -- dats Sa Tt may be verified that the meaning of the quantities u,, Up, W, B1, and Pp 4s as has been indicated by comparing equations (54) to (61) with the corresponding equations of reference 2 for the homogeneous shell with Eo = Go =a. The system of equations (54) to (62) may be brought into a slightly more concise form as follows: Define the quantities 21, Xo, and Aye dy Me Re (63) and eliminate k from equations (57) and (58) by means of equation (62) and the equilibrium equation (34), Retain equations (56) and (59) to (61) in the foregoing form and write for equations (5!) and (55) 2 -(v-4 = ox(2 Sun, v2 Se, ww @+da)m « 3 ne)itoe oS + aks st + Geer, (eu) ok NACA TN Wo, 1832 2 ~(v-2 = on/p 202 , HL Bee x’ +3 ame ( } a) = *(@ 58 aay Sey *) a Equations (57) and (58) become, if D*/(h + t)Ey Btn + t)Bp/Bo, according to equation (47), @ + %a)iha — (v - %2)Mee = (1 +2) Moe - @ — Mae)Maa With these last transformations there is obtained a systen of equations which is formally equivalent to the corresponding system of equations for the homogeneous shell, ‘The 5 equilibrium equations (25) to (28) and (31) end the 8 etrese-strain relations (56), (59), (60), (61), and (6h) to (7) are used for the determination of 13 quantities: Five atroas resultants y;, Nop, Nyp, %, and Qj three stress couples My), Mya, and Myo; and five displacements and changes of slope uy, Us, W, By, and fp. The quantity ot, which occurs in the sixth equilibrium equation (equation (34)) may be determined directly, once the shell bending and stretching problem has been solved. It is seen that the effect of transverse shear deformation enters equations (60) and (61) only and that, when G, = =, these equations give the values of the known theory of homogeneous shells without transverse shear deformation (references 2, 3, and 4). ‘The effect of transverse normal stress deformation enters equa~ tions (6h) to (67) only. It is seen that it is, in part, responsible for the occurrence of apparent stiffness factors C¥/(1 +X) and D*/(1 +). Thus, according to equation (63), the effect of finite Ez, 16 to mike the shell more flexible in bending and stretching than it would be with Ey = ©, This effect, however, is present only in curved structures ani not in plates and straight beams, as the quantities % have one or both of the radii of curvature in the denominator. A further effect of finite Ey is occurrence of the NACA TN No, 1832 25 external load terms q and 6 in the stress-strain reletions, Both these effects represent, roughly speaking, what happens to the shape of an element of the composite shell if the length of the core fibers in transverse direction 1s changed, without eny stretching or compressing of the face-membrane elements, Having derived the general system of equations for the small ‘onding and stretching of seniwich-type shells, it remains to apply those equations to specific problens which may te of interest and to determine the quantitative effect of the terms which are characteristic of the eandwich-type shell. Some of this work 1s done in part II of ‘the present report, which follows. Tt may be stated once more that for these specific applications the five equilftriun equations (25) to (26) and (31) and the eight strese-strain relations (56), (59), (60), (61), and (64) to (67) are used. IT — APPLICATIONS OF GENERAL THEORY Flat Plates ‘The problem of the flat plate is considered first in order to show that the results of reference 1 are contained in the present results and in order to solve some problems in the theory of plates which have not been solved in reference 1. Rectangular plates.— Using notation which is customary in plate theory there is set &1 ba-¥ O00 ee ee yew ene Py = Bx Ba = By Mi="% Me = My Nae = Ny = & (68) S=% Mi Ye = My Meo = My mem dy mem my 26 NACA IN No, 1832 The equilibrium equations (25) to (28) and (31) become an, a BE SE sao ahy , a oy, = ofa nO ~Y tm The stress-strain relations (56), (59), (60), (61), and (64) ‘to (67) become my y= oF BE 1 2 Hy — vu, =o 2(1 + Vly = o@ +z Ge = (8 + 1)06(6 + &) a = (n+ #)60(6 + 2) (@) (70) (7) NACA TN No. 1832 aT 28; — VM, = De Se My — Voy Be _ pe 38 My - my = DF ST (73) 2(1 + vy =D a , 2) As in the small-deflection theory of homogeneous plates, the equations for stretching (equations (69) and (71)) are independent of the remaining equations for transverse vending. Equations (69) and (72) for the stretching are not affected by the elastic properties of the core layers. Equations (70), (72), and (73) have been treated in reference 1 by moans of a strese function ¥, which, together with the deflection w, was taken as one of two besic variables. In whet follows an alternate ‘treatment is given, in which tho problem is reduced to three similtaneous equations for the quantities Px, By, and wv. On the basis of these three simltaneous equations a problem nob considered in reference 1 1s treated, nanoly, the bending of a rectangular plate which 1s simply supported on all four edges.” To reduce equations (70), (72), and (73) to three simltaneous equations for x, Py, and w, firet a quantity @ 18 defined by 3s. a= oe, (74) Introducing equation (72) into the first of equations (70), in view of equation (7k), there is obtained + Vw = —a/(h + t)G, (7) Next, Q, Me, and My are taken from equations (72) and (73) and the result is substituted in the second of equations (70). This gives, after slight transformations, ‘Tmmis samo problem has also been solved by L. H, Domell by a method which differs from the one employed here. (See reference 5 where the case of the homogeneous plate is considered.) 28 NACA TN No, 1832 Peg Pr — 20h + HOB + tes = a(n + 2a] ate | (76) In an analogous manner the following further equation is obtained: (™ hy hy — 20h + toe By +R fPBB, ~ a(n + dege] + my In order to solve equations (75) to (77) two equations are next obtained involving w and @ only, Differentiating equation (76) with respect to x and equation (77) with respect to y end adding the two resultant equations, in view of equation (7h), gives Bit on oth + 105 fo + 9H) 4 SE SE ° end, miking use of equation (75), a= 3f 3 3] (78) The following procedure may now be carried out: (a) Solve equation (78) for ®, (») with thie value of © solve equation (75) for wv, (c) substitute @ and w in oquations (76) and (77) and solve for PB, and By, and (4) eliminate extraneous terms in Pp, and By by considering equation (74). Before deriving the solution of a problem along these lines, the explicit differential equation for w which follows by combining equations (75) and (78) may be given Pew = 2 , &)) - v2 _ afr x2 = th + t)Ge () Note that the effect of transverse shear occurs on the right side of the equation only. In order to compare the magnitude of the q terms on the right of equation (79), aegune that relevant changea of @ cecur over distances of order’ 1 (where 1 may or may not be a fepresentative dianeter of the plate). ‘Then, as order-of-magnitule relations, there results NACA TN Wo, 1832 29 Gaz) a _ ofa tes te: “(Pnc, From equation (80), tt follows that transverse shear ceases to be @ secondary effect as soon as 1 is of order fit \ip/Go or of emaller order. Bending of rectengular plete with simply supported edges. The edges of the plate are assumed to be at x= 0,a and y = 0,b and along these edges momente and deflections are assumed to vanish, Further, 2 2 (81) pee mi mi where Ay = Bx/a (82) Bn = ax/b ‘From equation (78), it follows that on BY) pig ats tar tn ony + (83) Mae + ty where a, 48 a harmonic function, Putting equation (83) into equation (75), 30 NACA TN No, 1832 ee) > ten fae wag] er (h + t)G¢ vhich is integrated to “aYigs where wp is the general solution of V°wy =-an. It is to be expected and may be shown explicitly thet for the plate which is simply supported all eround ¥;, =, = 0 and, as in the Navier solution for the plate without transverse shear deformation, the particular integral is the complete solution of the problem. ae oft «2 Sa ane ot tay + Mp (8) + we) Equation (84) may be rewritten in the more explicit form oe yy tent + F tye She + ele n2)] oF [a2 + 22 (22/02)]* x ein ME x ain BE y (85) When Gy = ©, equation (85) reduces to Navier'e solution wy. Equation (85) 4e more readily interpreted by meana of the ratio w/iy of deflection with and without transverse shear deformation. On the basis of equation (85), there may be obtained the following equation (86), 8Ssetting 2/2(1 - v2) = 5.4 and (Bp/c,)t(n + t)/e® = B, equation (86) takes on a form which contains as a special case the result of equation (18) of reference 5. NACA TN No, 1832 32 (n+ t)tEp ny “3a_-)a = a(1- Pay For ‘ho cose of 8 uniform ond intensity q/ Constant and for the conter of the plate (x = a/2, y = b/2) equation (86) becomes > > sin mx, oe 2D ah? eee Ee t “ dt im? 4 n2(02 /2) Hoe 2 Gwe, “3 y sary See vein Bee in ein ma (87) mal? + n2(a? /e2)] ‘The ratio of the series is 1.98, when a/b = 1, and the ratio of the series is 1.11 when a/b = 1/2, For the case of a concentrated load at the center of the plate the deflection ratio at the point of load epplication assumes the form (sin mx/2 gin nx/2)? awl See Ton -v e, y (ein ma/2 ein nx/2)® uP + 22 (e2/22)]® Now it is easily shown that the mmerator series in equation (88) does not converge and consequently w/iy = = in thie case, A nore detailed consideration shows that in any plate theory which takes transverse shear deformation into account the deflection under the point of application of e concentreted load mst decoue infinite in contrast to what happens when transverse shear deformation is not taken into 2 NACA TN No, 1832 account, This difference, of course, vanishes as soon as the load Intensity becomes finite, end then the theory with transverse shear defornation taken into account 18 more accurate than the theory which does not take into account this effect. For the sske of numerical illustration take again the square plate (a/ = 1) with uniform load distribution. According to equation (87), The deflection at the center is increased because of transverse shear by the factor ae Ee (ha tit Herb eTgee (87a) Take h = 1.0 inch, t = 0.1 inch, a = 10 inches, Ef/G, = 200, ant ¥ = 1/3. Thea, according to equation (87a), w/wy = 1 + 2.3, 80 that in this case the deflection with transverse shear is more than three times the deflection when shear deformation in the core is neglected. Returning now to equation (81) for w and equation (83) for and substituting these two equations In equations (76) and (77) in order to determine the changes of slope fy and fy, after slight transfor— nations there results be BD tea coe tax ata ty (rn? + un?) (89) 3) ae Equations (89) are remarkable for the reason that they are not affected py transverse shear deformability. According to equations (73), the same is then true of the bending and twisting couples Mr, My, and Mry- Tt is not easy to see why, in thie statically indeterminate problem, the magnitude of the internal forces, as vell as that of the deflections, does not depend on the elastic properties of the core. The analysis, however, shows that the distributions of Mr, My, and Mry, and therewith of Q, and Qj, remain the came as those obtained under the assumption that Oo = ©, In this connection the following remark uay be made. NACA TW No. 1832 . 3 Evidently the following three boundary conditions, w = My = By = 0 along the odges x = 0,a, have been satisfied. In order that the last of those three conditions be satisfied there are necessarily nonvanishing edge values of the twisting couples M,y. Tho same is true in the theory without transverse shear deformation, where, however, no alternative possibility exists, as in that theory only the boundary condi— tions w= My = 0 are relevant. For the present system of equations three boundary conditions mst be formated for every plate edge, Thus, it is possible although mathematically complicated to solve the problem of the rectangular simply supported plate with the edge condition B, = 0 replaced by the condition M,,= 0, In that case, which will not be pursued here, there evidently will be a distribution of internal stresses which is modified by the effect of transverse shear deformation. Cylindrical bending of plates.— As a further relatively simple example of application of equations (70), (72), and (73) problems are considered for which a Py = 0 OC )Px = a( )fax = ()' May = = my = By = 0 (90) My = Vy and where consequently the problem reduces to the following system of equations: Qt+q=0 ie Orie ee) (92) & = (h + t)G (Bx + w') (0. - v?)Me = Dea," Note that in order to obtain the problem of the sandwich beam from equations (90) and (91) the only changes which are necessary amount to setting v= 0 in equations (91), WACA TN No. 1832 3 To get into evidence the effect of finite values of G_ in equation (91), the following system of equations is deduced from equation (91): (Iv) © + Patt Dw caret ao 2a (92) wo Da Mr 7 rs te (93) Sees +m (94) + & By =— Ww rs tee (95) Solutions to the following rroblems are listed: (1) Simply supported plate of span 7? carrying a load _g = 99 cos sx/%. Boundary conditio: w(at/2) = My(42/2) yah, 2 Er s tt] c08 x/2 (96)20 > 2a-Aa 2 | (ryt Aa the problem is statically determinate as far as moment and force are concerned there is no modification of My and Q due to the finite value of Go. 10me factor in brackets may again be written in the form 1 +°5.4p, with p= (e/e)f[a + ove], using the notation suggested in reference 5. NACA TN No, 1832 35 (2) Simply supported plate of svan 2 carrying a uniform osd_a = ao. 4 4 2 00 eal go - ¥ ot ales )-tboa Salle i en From this for the center deflection, ye gov 2k, hs t)t W(0) = Bp hb + wae te] (98) Tt is soon that the correction factor for the center deflection is almost the same as that for the cosine load curve (equation (96)), the only difference being a change of the factor x°/2 = 4.93 into 2k/5 = 4.80, that 1s, a reduction of the shear correction factor by at most 3 percent ie prosent.12 (3) Built-in plate of span 2 carrying a uniform load q A Hy £7 ES =0 te ountary conte ons are: w(47/2) = p, (42/2) (and . Sa bale’ | (99) From this there follows for the center deflection, 4 Ld Eg (a+ t)t wo) BE vas cg at ow) Comparison of equations (100) and (98) shows that for the builtin plate the effect of transverse shoar deformation is very mich more Uyote that according to equation (87a) the shear correction factor for the square plate of width a= 1 is more then twice as large as the shear correction factor for the plate strip of width 1. 36 NACA TN No, 1832 pronounced than 1t is for the simply supported plate, a factor 24/5 in the latter case being replaced by a factor 2h in the former cage.J2 Aa a further result in this problem of the built~in plate, by putting equation (99) into equation (93), it is found thet the moent function My does not contain eay terms depending on the effect of transverse shear deformation, This again is somewhat surprising as in thie case 1t te not possible to determine the moment function by statics alone. Circular plates; rotational symetry.— As no examples of solutions of circular sandwich-plate problems have as yet been published and as 4t 1a of some interest to determine in which way the shear correction factors change in going from a problem for the plate strip to the corresponding problem for the circular plate, the equations for axiaymuetrical transverse bending of circular plates are briefly discussed. Poler coordinates r,@ are introduced and notation which is customary in plate theory is used. As a consequence of equa tions (70), (72), and (73), the following system of equations is obtained: are ar ea 0 (101) Sie _ vy — rq, + rm, = 0 Q, = (b+ t)Ge (8, + aw /ar) (102) 12, gomewhat similar percentage increase mst take place in going from equation (86) for the rectangular plate with ell four edges eimply supported to a formla (which has not yet been derived) for the rectan— gular plate with all four edges bullt in, 13ag a problem where the moment @istribution ie’in fact dependent on the effect of transverse shear there my be mentioned the problen of the cylindrically bent plete with both ende built in, which carries a Joed q, = a,x instead of the load ap = 49. ‘This problem also may be solved by means of equations (92) to (95). NACA TW No, 1832 37 Mp — WMg = D* a8y/ar (203) Mg — Wp = D*By/e According to equation (79), the equation for the deflection w ¥ALL be eaqeh2 eed. wy =a +h) Se mE (104) where V° = (1/r)a[r a( )/ar]/ar. Having found w by means of equation (10k), A, may be determined from fy = = Hy 3B fs ae (105) and therewith M end Mg are obtained from equations (103). Tn the present problem it seema to be somewhat nore convenient to Proceed as follows: Combine equations (101) and (103) to obtein as equation for the change of slope fr, (106) Having fr, M, end Mg are found from equation (103) and OQ, from the second of equations (101), (107) 38 NACA TN No, 1832 oe = DG SE) + mp (208) Finally, vith this value of G,, w is found by integrating equation (102), = [ror meee) fe (209) + ter © (h + t)Go Deflection of circular plate with built-in edge.— The bending is now considered of @ plate with tranaverae load q = an(r/a)” and with m= 0, First, from equation (106), 3 n+3 og E logy E + ——Se®>? _, +038 8b * Cin se) (110) Dip = op E+ og ‘Attention is restricted to complete plates with no concentrated oad at the center, and consequently it is necessary to set co, = c3 = 0 in equation (110), This gives (1108) a2 (n + 4)(n + 2)? Putting equation (110a) into equation (109), there results for the transverse deflection : ane? ey + a} (n+ 4)(n + 2)? eles 2a aes *t ee) | oy NACA TN No, 1832 9 ‘Taking the case of a plate of radius a with built-in edge, that 1s, with the boundary conditions By(a) = ¥(a) = (az) there results 3 +3, - —se le -r Te = Tate aL) 4] (a3) ena e/ad™h a3 (ofa)? =a (aes tte {ia (n + 4)? 7 eee (ry a@(h + wal ) | (ak) From equation (11M) there follove for the deflection at the center of the plate 4 De(0) = one p+ ose a(n + 2)(n + 4) Consider the following special cases: (1) Uniform load distribution qq = qo. From equation (115), it follows for the ratio of deflection with and without transverse shear deformation that (0) eae Ep (n+ t)t Tol 5. Ge, a 40 NACA TN No, 1832 Bquation (116) may be compared with equation (100) for the deflection of the infinite plate strip of width 7 with built-in edges. . Setting 1 = 2a, it is seen that, while the transverse shear correction factor for the strip has a value 6, the corresponding factor for the circular plate is 8. This is consistent with the earlier comparison ‘between the simply supported strip and the simply supported square plate, except that there the change is from 4.8 to 9.7. (2) Linearly increasing load distribution a = ay equation (115), 1t follows that w(0) 25 rm et (x7) Molen 3 Ge, * shoving that the correction effect is only slightly greater than in the case of the uniform load distribution. (3) load increasing linearly from edge to center, g = a9 + ay(r/a). (a= -a0)- From equation (115), it foLlova ty qipersoeition tat that _ 23age" [, , 3000 __Er (hn + t)¢ me(0) = Sel + Be Gave, (8) Comparing the factor 3000/387 = 7.76 which occurs in equation (118) with the corresponding factors 8 and 8,33 in equations (116) and (117) Tt ie seen that, in the foregoing thres problems at loast, there is little difference between the transverse shear stress correction factors . in the case of three different loading conditions for the circular, clamped—edge plate. The fact that this agreement should not be expected fo hold gonerally follows again by considering the case of a point load at the conter of the plate, for which the shear correction factor would again be infinite, NACA TN No, 1832 aa The examples of this section should be augnented by the solution for the circular plate of radius a, which carrios a load distributed uniformly over a smaller circle which is concentric with the boundary of the plate. Circular Rings As the simplest example of a curved sandwich structure there are considered in this section stresses and deformations of -circuler rings in their own plane. As was found ‘in the general developments of part I of this report, in a curved sandwich structure there will be the effect of both transverse shear and normal stress deformation. There are set for the relevant coordinates end variables fy = 08 a=l Reza wey Pl = 8 My =6 a=a Mi = M (aig) PL=P mora 3( )/a8 = a( )/a do = ( )*/a dg = = Alle + t)t/e2] (G/B) The equilibrium equations (25) to (28), (31), and (34) reduce to the following equations: 0 ° N+Q4ap (120) 0 ° Q-N+aq le effect of transverse shear stress deformation on homogeneous cirouler rings has been considered by L. Beskin in reference 6. ye. NACA TW No, 1832 Mt aQ+am=0 (121) otm = 8 — [i/(n + t)a] ‘The stress-strain relations (56), (59), (60), (61), and (64) to (67) reduce to the following equations: (oda dofe sv gt] (az) a= (n+ t)ce[p + owt - v)] (223) (2 + aM = 2 w(pt + 8/8) (a24) The load terms p, a, m, and 6 are given, according to equations (19) to (22), by (125) or a ACA TN No. 1832 43 Ring sector acted upon by end bending moments. As e first problem on circular rings, which illustrates the effect of transverse normal stress deformation, there is taken this basic case for which, as is imown, there mat be tho same stress distribution at all sections @ = Constant of the ring. According to equations (120) and (121), = % (226) Oty = — Mo/(h + ta Equations (122) to (124) become view=0 84 (ut —v/a=0 al (1 + %)Mp = D*B'/a (128) ‘The significant remilt of thie consideration is contained in equation (128), which may be written in the alternate form pe pt De Mem =e = (129) olsen rsp heme & c ‘Thus, in this case of pure bending the traneverse flexibility of the core is responsible for a reduction of the bending etiffness factor D¥=2 t(n + t)®8 which is obtained exactly when Eq = 0 and practically when E, is of the same order of magnitude as Ep. Equation (129) shows that the reduction of D¥ is elgnificant vhonever E, is so small that the ratio E,/%p is of the sane order of magnitude aa the ratio (h + t)t/e?. ne NACA TN No, 1832 As @ mmerical example take the following values: h = 0.9 inch, +t = 0,05 inch, a = 20 inches, and Ep/B> = 1000, for which a (hs tit Fe a, 0.95 x 0.05 . ga Bt Fn Bx OTE x 1000 60595 indicating a reduction in bending stiffness of about 6 percent. Changing a from 20 inches to 10 inches changes the effect from 6 percent to 24 percent. Chenging Ep/l_ from 1000 to 2000 incressse the effect from 6 percent to 12 percent. Altogether it may be said that this effect 18 of noticeable magnitude for sone goonotrically reasonable structures vhen the modulus ratio E/E 1s of the order 1000 or more, Assuming aluminum face leyers with Er = 107 pst, this means that Ey 104 psi, which is well within the range of some prodent-day core-layer materials. Comparing equation (129) with the earlier formlas for the effect of transverse shear stress deformation, for instance with equation (116) Invhich a represents the plate rediua and obeorving Unt Gp % 2 Fe, tt te esen that the correction terms are of the sane form, the difference peing an appreciably lerger numerical factor in the expression representing ‘the shear effect. Closed circular ring acted upon by uniform radial load. Having rotational symetry, 4/40 = 0 ani v=—-=0, Also set p=m=0, ‘The remaining equations permit the determination of the stresses in the face end core layers in a way which depends on the extent to which the load is applied to the outer (upper) and inner (lower) face membranes. Equation (12) becomes ag (130) From equations (121), it follows that ~ [rym + t)a] (231) om NACA TN No, 1832 45 ‘The stress-strain relations (122) to (12k) give eee sda Geet-shg- Sede] a Dt ig Mey 7X aie (133) A closed circular ring eubjected to @ uniform radial load distribution q is stressed not only by a uniform axial force W = aq, as would be expected, tut in addition is stressed by a uniform bending moment M, the magnitude of which is given by equation (133). The explanation of this result is that for a ring with relatively soft 5 core the circumferential stress distribution depends on the extent to which the external redial load is applied to the inner and outer forces, respectively. Roughly speaking, for a sufficiently flexible core layer the load q, goes predominently into the outer face layer, while the load q; goes prodominantly into the inner face layer. According to equations (46), in the present case for the stresses in the two face layers, (134) According to equations (130) and (133) and in view of the definitions of D* and 2, this may be written (135) 46 NACA TN Wo, 1832 Combining next equations (131) and (133), for the transverse normal stress in the coro layer, the following expression is obtained: Stay ey For a specific example assume that the radial load is applied entirely to the inner face of the ring so that qy = 0 and, according to equation (125), a= Q-B24)q, (37) o--30-Ag#n With q and s given by equations (137), equations (135) and (136) become t (138) mae Roe ste =~ 30—Aartat a) It is seen that the flexibility of the core layer increases the circumfer— entiel stress in the loaded face layer in the ratio (1+ 2%)/(1 +2) and decreases it in the unloaded face layer in the ratio 1/(1 +A), where is defined by equation (119), compared with the equal values of these stresses vhen Ey = « Considering once more the numerical data under the section entitled Ring sector acted upon by end bending moments,” it is found, for instance, that the stress in the inner face layer may ve about 6 or 12, NACA TN Wo, 1832 aT or 2h percent higher than the corresponding stress calculated without taking into account the transverse flexibility of the core layer. Ring sector acted upon by radial loads gy and qz, uniform in circumferential direction and with vanishing resultant q.— Again it is assumed that d()/aé=0,m=p=0 and now in addition that gq = 0. 80 that, according to equation (125), the only nonvanishing load term is s. Further, it is assumed thet the ends @=-+ a of the ring sector are free of stress, that is, N(éa) = Q(ta) = Min) = 0. ‘The ordinary theory of circular rings would then indicate the absence of deformations in the entire ring. In the present case there is found a type of deformation peculiar to the sandwich ring, which may perhaps ‘be compared to the action of a Bourdon gage. Solving first equations (120) and (121) and satisfying the end conditions of the ring sector, N=Q=M=0 (ako) Sm = 8 The stress-strain relations (122) to (12h) are then vaw aba wt (ak) BY = -8/E, Assuming 8 independent of 8, from equation (141) there 1s obtained by integration, with constants of integration Ay, A, and A3, aoe Ba- form Ve eg @ + Aya + Ap con @ + Ay in 6 (abe) w=a Bs fo sin 6 —A3 cos 6 Ee 48, NACA TW No, 1832 As a specific example consider a completo ring, slitted radially at the section @ = x, so that a= x. Prescribe furthermore the symmetry conditions §(0) = v(0) =w(0) = 0, Under these conditions there is obtained from equation (142) EB = -20 Eov = -as(@ — sin 6) (143) Eyw = as(1 — cos 6) From equations (143), it follows thet the redial slit, which is of zero width before the loads a, and 7 are applied, opens under the action of the loads to a width given by via) ~ v(x) = 2x0 B= 2ra(t + hee (a) For a merical example take a = 10 inches, h = 1 inch, t = 0,05 inch, Eq = 10,000 pai, and gy = 20 psi, and obtain v(—n) — v(x) = 0.132 inch (245) The foregoing three examples of ring analysis have been discussed in gome detail, because they illustrate relatively simply the effect of transverse normal stress deformation in the theory of curved sandwich structures, without involving at the same time the effect of transverse shear stress deformation, Bending of semicircular ring by end shear forces.— Now a problem is considered in which both the values of E, and G, affect the result of the analysis, In the equilibrium equations (120) and (121) all external load terms are sot equel to zero and then, by integration and from the boundary conditions, that is, from HEA) MEAD-e a 8) = #8 (146) NACA TN No, 1832 9 the following expressions for N, M, and Q are obtained: : Q@= % sing N = Q% cos 9 (147) M = a0 cos @ The stress-strain relations (122) to (124) become (1 +2/3)Q com @ = (C¥/a)(v + w) Q@ sin 6 = (n+ t)Go[B + (wt — v)/a] (148) ~(1 + d)Qa cos @ = (Dt/a)pt Integration of the lest of equations (148) gives Dep = -a2(1 + A)Q sin 6 (149) where a constant of integration has been eliminated by means of the symetry condition (0) = 0, Substituting equation (149) in the second of equations (148), Yur — ¥) = @ sin ole: doa #0» er re 2) Simltansous solution of equation (150) and the first of equations (148) for v and w gives as general expressions for v and ¥, 4 1 AO cos 8 + Ay sin @ + Ay cos 8 (152) < A9 gin @ — (A, + B)cos 6 + Ap sin 6 50 NACA IN No, 1832 where A, and fig ere arbitrary constants of integration and A and B are found to be ooe3[, gus t)t/fp Be) pe | Bef a SPE +f) + Bel +3) aoa3[, a (n+ t)t (Bp | Be) pe 2») wee 2 0 Fat +3) Ag further conditions, it ie prescribed that v(0) makes Ay = Ay = 0 in equation (151), There remains (152) v= AO cos & (253) w =A gin @— Bcos @ of particular interest are the values of ¥(x/2) and w(0), the first of these giving the radial deflection of the point of load application, the second giving the change of radius at right angles to the applied Joad, It ie found that (154) (155) Equations (154) and (155) contain the interesting result that, for this problen, transverse shear and transverse normal stress affect the outcome formily in nearly the same way. If the generally unimportant terms with D*/s®C* are omitted, which amounts to the usual assumption of circumferential inextensibility of the ring, then the effects of finite B, and G, occur in exactly tho same wey. WACA TH No, 1832 52 For a numerical example take h = 0.9 inch, t = 0.05 inch, a = 20 inches, Ep/E, = 1000, and Ep/G, = 2000, This gives The factors in brackets in equations (154) and (155) become Ht 2 2) SO pone An ey amd Halt pan) 6.6 * 16.8 ae +2 - —)- 1* 38 * 168 iol +x za) ae ‘Thus, in the present example the flexibility of the core is responsible for an 18-percent increase of deflection-load ratio, and of thie 12 percent is due to transverse shearing and 6 percent to tranéverae normal stress. Compared with these two effecta the effect of circumferential extensibility of the composite ring is seen to be negligible, As a further numerical illustration, it is noted that reducing the ring radius a from 20 inches to 10 inches, with all other data unchanged, cherges the 18-percent correction to a 72-percent correction. Bonding of complete circular ring under action of two concentrated radial forces at_0 = 2n/2,— The solution of thie problem may be obteined $y auperposition of the solutions for the semicircular ring under the action of end shear forces Qp (equations (146) to (155)) and under the action of end loading moments Mo (equations (126) to (129). 52 MACA TN No, 1832 ‘he first step consists in determining Mp in terms of Q% such ‘that the sum of the A's from equations (129) and (149) assumes the value zero for 0 = x/2, that is, the value of the superimposed bending nonent at 0 = x/2 mist make tho tangent to the deflected ring at this point horizontal. Combining equations (129) and (149) in this namer, there is obtained Ltd wa = 0 . B49? aor Mo = (2/m)aQ% (156) Tt may be noted that equation (156) 1e a further case of a statically indeterminate problem where transverse shear and normal stress flexibility do not affect the internal force and monont distribution but effect only the state of deformation of the structure. Further, the radial deflections ¥(x/2) and w(0) due to the action of % are calculated, in order to combine them with equations (152) and (153). Integrating equations (129) and (127) with the boundary conditions +(0) = v(x/2) = 0, there is obtained for the displacements due to M, DA = (1 + Moe? (L - F cos 6) (157) Diy = (1 + )Mga?(0 - ¥ ain @) end, in particular, Dvw(0) = (1 + 2)Moa?(Z - 2) (158) p(B) = (G+ 2) Moa? NACA TN No, 1832 53 Combining equations (158) with equations (154) and (155) and taking Mo from equation (156), there follows for the resultant displacements @)- LE D0 0 the Bee Hh w(o) --92 (2-Day -ap - Bb -ajh (159) (n+ t)t Ee were rq = “Es tt a Equations (159) may be written in the alternate form has been put as a further abbreviation. w(x/2) = 0.249 ef oe s.asfte : w(0) = 0.137 ef on 3.650 . Fe + a} (261) When X= dg = 0 and when the composite ring is assumed axially imextensible, which amounts to putting D*/e20* = 0 in equations (160) and (161), then equations (160) and (161) reduce to well-known results of circularring analyste. Comparing equations (160) and (161) for the closed circular ring with equations (15!) and (155) for the open semicircular ring, it 16 noteworthy that for the semicircular ring 2 and Xg occur with equal weight, while for the closed circular ring the influence of 2g is considerably greater than the influence of 2, Thus, for the closed circular ring the effect of transverse shear deformation is mch more important than the effect of transverse normal stress deformation, while for the open semiciroular ring toth effects occur in a mich more nearly equally important way. For @ numerical example of the use of equations (160) and (161) take agein the values for the numerical example given in the section entitled "Bending of semicircular ring by end shear forces." Tania gives for the expressions in traces 5h ACA TN No, 1832 1 2%5.29 , 5.09, l+iggt ee * i707? 1 .2%3.65 , 3.65 4 59 ** 768 * 36.8 * i770 Tms, while the effect of transverse stress deformation for the open circular ring amounted to 18 percent, the corresponding corrections for the closed ring are 69 ani 50 percent, respectively. The next step in the analysis of saniwick~type circular rings would be the general solution of the system of equations (120) to (12h) for arbitrary load distributions. ‘This, evidently, te possible and further specific examples of interest might te analyzed on the basis of the general solution. Such extension of the vork of this section 18, however, left for future considerations. Circular Cylindrical Shells In this section the general system of equations of part I of thie report is restricted to the equations of the theory of circular cylindrical sholle. The treatment of sandwich-type shells of this kind 1s shown to be not appreciably more difficult than the analysis without the effect of traneveree shear and normal stress. As specific examples some problems of rotationally symmetric defornations are treated. In particular the influence coefficients are obtained for a semi-infinite shell acted upon by bending moments and transverse forces at one end of the semi-infinite shell. With these influence coefficients an explicit solution 1s obtained for the problem of the infinite circular cylindrical shell acted upon by a pressure ‘band of zero width. NACA TW No. 1832 el In the general equations of the problem there are set for the relevant coordinates and variables, & = 20 bax Ree Mi = % % = & Mo= Me WEY By = Be Ao = By (162) Mg = Dy Pi = Pe Po = Px mec On Mio = Non = Yeo Mo = Mor = eo ‘The equilibrium differential equations (25) to (28), (31), and (34) desone ON ON Beta tne ao 1 Me , se tbs tet me <0 (263) 2a , 1 29 _ Na & *a3e a 77° OMe , 2 Me 7 Be te 3g tO (164) 56 NACA TN No, 1832 Mo t= 8 - TS (165) ‘The stress-strain relations (56), (59), (60), (62), ana (64) to (67) become, with Mg = Ap = 0, dy =Zl(n + t)t/e?](Be/ee) = 2, Geom oer Gan — Wiig = on(32) (266) 2(1 + VIRgg = ot +2 @= + e¢6(¢0 « 2 kz) 308, Y (167) Gane 06(% +x) (2+ 20M ~ me = oo(2 3B 8) = Mm = »(&) (168) J a 2(1 + W)Mgo = oe +43) When = Eo = @ (and therewith 2 = 0) equations (163), (164), (166), Gen? and (168) reduce to the knovn aystem of equations in which deformations due to transverse stresses are neglected. ‘The solution of the present eystem of equations is not essentially more difficult than the solution of the system vith Gp = Eo =, In particular also here NACA TN No, 1832 5T there may be obtained a trigonometric double-series solution, as a generalization of Navier's solution for the flat plate (references 7 2 and 8), For this trigonometric double-series solution there 1s set, oD) ewes 20-) ) Pomn c08 mo ein sx/2 (169) o =) >. em stn etn mx/t we) > Vom sin m ein nx/? >> Yam 008 m6 sin mx/? a2) gq wim me con nx/2 (a70) a y I Pam Sin m9 cos mx/2 Bo = > > Boun cos mé sin nx/1 58 NACA IN No, 1832 Qs JY Sean 81m n# 08 mx/2 Ge YY an 908 28 win nx/t (Res Me) = Ss >» (Gams Nom) sin mo sin nx/2 Tyg = »y S Nom 008 m9 cos nx/7 aa (Me, ¥) > Yan» Moqn) #12 m9 etn mx/t ne yy Megum 008 139 com mzx/2 When equations (169) to (171) are substituted in equations (163) to (168) there renains for every value of m and n a systen of 13 simltaneous equations for the 13 Fourier coofficients which occur 4m equations (170) and (171). A syaten of only five simltaneous equations for tho five Fourier coofficients in equation (170) is obtained if first equations (163) and (164) are reduced to five equations for the five unknowns wv, v, u, By, and fg, by moans of equations (166) to (168). For the present, the task is not carried out of obtaining the deformation and internal stress Fourier coefficients of equations (170) and (171) in terms of the Fourier coefficients of the load terms in equation (169). Instead, the axisymmetrical case, to which equations (169) to (171) reduce when sin m@ and cos m@ are interchanged throughout, and then only the terms for m= 0 are taken, is treated separately. Axieymetrical deformation of circular cylindrical shell.— In equations (163) to (168) set 4 aC )/ee = 0 a )fex =)? Tix = % = Mox Bo = 0 (ar) Bg = Bo = 0 NACA TN No, 1832 59 and then the following system of equations has to be dealt with: Ny! + py = 0 (173) &" = (N/a) +a = 0 My" — & +m = 0 (17k) om = 8 — Mo/(h + t)a : (+ 3 vy, = orfe , Geta) (175) - Ty — Vig = C*ut @ = (n+ t)00(R, + w") (176) (1 + 2)M - VM, = D¥*s/aky (177) My — My = DB," The system of equations (173) to (177) may be reduced to two simltansous equations for py and Q,, ae follows: First, express M, in terms of , by moans of equation (175) and substitute ‘the result in equation (17k), From the first of equations (177), it follows that =? +s 178) aera (1 + A)aES : 60 NACA IN No, 1832 and this, introduced into the second of equations (177), gives ~ G+ ,, vps Mee te) (179) Bquation (179) 18 introduced into the first of equations (174) and, restricting attention to shells of uniform section properties, there is obtained (La MD gir go my — Yes (180) rea-w ie a0 +.-¥) To obtain the second of these equations, first, introduce into equation (176) the value of vw! which follow from equation (175), giving 2)" + _ CF(h + t)g! woe &[e + 3)Net mt = oe] (281) In equation (181), Wj! and My! are taken from equation (173) and, after slight transformations, there ie obtained e+ = + (180) (a+ HG, Comparing equations (180) and (182) with the corresponding equations without the effect of transverse shear end normal streas deformation, it is seon that the effect of transverse normal stress, which 4s represented by 2, merely somewhat modifies some of the coefficients of the left sides of the corresponding system of equations vith E, =, In contrast to this, the effect of finite G, is to introduce a new term into the left sides of these equations, This new term nay be of appreciable importance, as will be shown. Having solved equations (180) and (182), My and My are obtained from equations (179) and (178), respectively; Ng follows from equation (173) in the form Ng = a(@" +4) (183) NACA TN No. 1832 61 and w follows from equation (175) in the form we tefonfia +/3)0Q," + ag + vp a] (284) ‘Tho folloving examples tllustrate the use of equations (178) to (184). Infinite cfreular cylindrical shell with periodic load distribution— In specialization of equations (169) to (171), set a=aysinye | 5 = 6, sin ux r (285) Bx = Bry Coe 4x By = My, Cos UE v= ¥, sin ux w= a, cos ue (286) By = Bxy cos x % = %, cos ux | Ny = Nyy sin ux Ng = No, sin ux (187) My = Myy ein x Mg = Moy sin ux @ KACA IN No, 1832 By introducing equations (185) to (187) into equations (180) and (182), to eimilteneous equations are obtained for the expli— tudes Q,, and By, a8 follows: D* Hey Gru = en Ty ae Be (288) ote 2 ‘ft a :) a ly = - SF (hay + vixy/e) ane we cep ee % simplify the further discussion, by setting in equation (188) My = 8, = Pry 0, there is obtained for Py, and oier-v W Sue aren Deas (189) Ou = HEX J ‘The quantity K is given by -1 2 g, 2 7 [sy Pe aaa? j «s90) 3s (ns tha? So (2 + dat (ho t)?04 where uso hag beon made of the reletion y= x/2. In equation (190) the term 2/3 will usually be of little importance. The other two variable terms represent the effect of transverse shear deformation and of shell curvature, respectively, When the radius a is so lnree that v#/(n + t)®a2 << 1, the shell behaves under the action of the given load essentially aaa plate strip, The effect of transverse shoar is important as soon as the term (2/r°)(12/a2)(+/(m + t)) (Er/Ge) 4a not emall compared with 1. Before evaluating a numerical example the following further formas which are readily obteined from equations (179), (183), and (184) are Listed: NACA TN No, 1832 63 Su =UxK Meu = 3 Noy = 29,,(2 — X) (91) 22a, HEC Zquations (191) show that in this problem not only ie the deflection increased because of the effect of transverse shear, and with that the hoop stress resultant Mg, but nox also an effect 1s found on the dending-nonent distribution Mj,, in the opposite sense. The effect of transverse shear is to reduce the magnitude of the bending moments in the shell. This result 4 in contrast to what vas found for the exanples which were vorked out in the sections on plate analysia and circular ring analysis and is therefore of particular significance. Equation (191) for w,, may be compared with the corresponding expression for a simply supported plate strip of width 1, with simisoidal load. ‘the result for this case mst follow from equation (191) in the limit are and agree with equation (96), which was previously obtained, To compare the last of equations (191)’with equation (96), the last of equations (191) is written in the form 4 2 rsa vp a+ () Lea the tit _kaafe pei tet es 1M " Ls r—veyyyh ce Ay (2)? 2t Be 1+dsr—vayt o 3 iim) S 1,2) rg pera? ox (14+ .)pe \W 2 Equation (192) reduces to the equivalent of equation (96) if in it ase, From a comparison of equations (192) and (96), it 4e further concluded that the correction due to transverse shar is greatest in thie case when a= =, go that, in this caso, the curvature of the shell tends to reduce the additional shoar deformation, below the value obtained for the simply oupparted plate strip. on NACA TW No, 1832 For e numerical example first take h = 1 inch, t = 0.05 inch, a = 10 inches, 7 = 20 inches, Er/o = 200, Br/E> = 100, v = 1/3, and X=d Lee 100 = 0,025, The factor K of equation (190) 2 vecomes 2 400 x 0.05 4(1 = 0.09) K fis 0.00 + ong x ae « HOSE) = (1 + 0,008 + 3.86 + 54.5) 0.01685 while without transverse shear and normal stress deformation ()ggetgee = (1 + 54.5)72 = 0.0180 ‘The correction in this case amounts to about 6 percent. Changing the moduli ratio to Er/Go = 2000, Er/B_ = 1000, K = (1 + 0.08 + 38.6 + 54.5) = 0,0106 instead of K = 0.01685, The correction in this case amounts to 92,0180 = 9.0106 5100 w 70 poreent. Thus eguin there is & case where omission of the effect of transverse shear deformation would give results which could not be used. However, it ie noted that the effect of transverse normal stress deformation is quite small end may here safely be neglected. If the foregoing values of K are introduced into equations (191), 4 1s eoen that the percentage corrections apply to the bending-moment value directly but that for hoop tension and radial deflection the corrections are very emall indeed. In fact, in order that there be appreciable corrections due to tranaverae shear on hoop tension and radial deflection, it te necessary that the half wave length of the eimisoidal load q be so amall that K is at least of magnitude 0.25 or more. NACA TN No, 1832 6 A case of approximately this kind 1s obtained if the half wave length 1 1s changed from 20 inches to 10 inchse and the moduli ratios ere again taken as Ep/Eo = 100, Ep/Gy = 200. Then, K = (1 + 0,008 + 0.965 + 3.41)-2 = 0.1865 whereas (®)x,, = (1+ 3.41)? = 0.227 ‘The percentage change of K ani therewith of M, 1s slightly more than 19, The percentage change of Ng and w is about ue The foregoing numerical examples show that the effect of transverse shear may be significant in cylindrical sandwich—shell analysis and that noreover its magnitude will not in general be predictable by the enalysis of an equivalent flat-plate or straight-beam problem. For the infinite circular cylindrical shell with load q =a, cos nx the essential results are given by equations (190) and (191), These results may be extended directly to the loading ct a Yee ae (193) Mn = nx/2 By superposition, from equation (191) the following formless are obtained: Me =) Ga/tg2) a c08 tax Ng =a > Qn (1 ~ Ky) 008 tp (agh) w = (02/0) y gat - (2 + 2/3)K] coo vax 66 NACA TN No, 1832 ‘The values of Kj are obtained from the formila 2% Be Wa p= v2 mo gy Pe Ort ea Ge ee] Having the olution for the infinite shell with periodic load distribution, it will be only necessary to add to thie the general solution of the differential equations without external load terms, im order to obtain the complete solution for any edge condition of the axisynmetrically stressed circular cylindrical shell of finite length. Thie additional solution will now be obtained. Finite circular cylindrical shell acted upon by edge moments and torces.— To solve equations (180) ani (182) with right-hand eldes equal to zero, equation (182) ie differentiated twice and Px'' ia substituted from equation (180), This gives a7 — om?a,11 + imho, = 0 (296) where w2\f—_4" _.i [ __s mas a/ayihs ee FW + das 8) Oo (agte) ee a+aov 1 af aaa (sr) oe GQ ONA+3) Verte VA ++ 7) SEE NACA TN No, 1832 6T The auxiliary equation corresponding to equation (196) is rh — om2r? + im = 0 (198a) my? + \/my — my (198) The solution 9f equation (196) occura in tvo different forma, depending on whether r® of equation (198b) is real or not. According to equations (197) and (198b), r2 is complex as long as my! < tiny or (1998) ah - v/a +0) < (n + t)®a®(a + 2/3) To clarify this condition, neglect 2 (which is of very Little imortence here) and oquation (199a) then becomes Z< (1990) When equation (199) hold, a quantity k may be defined by Pn ee (200) and the four roota of the characteristic equation are xk, K, -x, and -X, where a bar indicates the teking of conjugates, The solution of equation (196) may be written Oe = Cy 4 TO 4 Cpe + Tel (201) 8 NACA TN No, 1832 Where equation (199) does not hold, which 4a the cage for very small values of Ge/Ep only, all four roots of equation (198a) are real and of the form wy = Wm? + fm! = img? ke = ky (202) 3 = \fm? — \fm* — 4m? xy = 3 end the solution of equation (196) can be taken in the form x; = kegx ~x Og = Aye N + Aye 1 4 age + Ayo > (203) Before applying either solution to e specific protien, there are noted the following relations which follow from equation (200) = 2 2 1 = [nl ~ 2m) + R= VE fm? 4 Oe Semt-infinite shell acted upon by edge Yending moment and shear force.— There are the following boundary conditions, (20h) L+rA-v (0) = @% 7 while for x = © these same quantities vanish. nig(0) = Le ADE g,0(0) vo] (205) 15¥or the seme problem without the effect of transverse shear and normal stress, see reference 9. NACA TN No. 1832 @ Of particular interest in this solution are the values of deflection w(0) and change of slope ,,(0) at the section vhere the lode My) and Q are applied.16 Taking first the case Ep/G.< 2a/t for which equation (201) applies, it is seen that the conditions et infinity require that ® =T%=0 (206) go that % = Cye a Te (207) The values of fx may be obtained by integration from equation (180) tn the form et pe _ 1l+i-¥v where two constants of integration have been discarded to satisfy again ‘the conditions at infinity, With equations (207) and (208) there is obtained from the boundary conditions (equations (205)) that (1 + = % (cule) + (21/) = -t (205) 16;ithout transverse shear and normal stress deformation these relations are 2 — v2)o* 8,(0) = SX ofs- et [S=2 | in agreenent with equations (236) of reference 9, where the homogeneous shell is considered. 70 NACA TN No, 1832 This determines C, ani 0, in the form 2, o, = Ho-+ bel? x-k (210) % = Hs Liel?Mo =k Equation (210) is introduced into equation (208) and there ie obtained as the first of two "influence coefficient" formlas fue 8,(0) = — —be[ag + (ik + MQ] (211) ae ele The second of these formles follows from equations (184), (207), and (210) in the form oe wl) 2 [ha ¥ so MP [he + + Bag] (212) Equations (211) and (212) may be written in more explicit form, using equations (204) and (197). The results are (1 +.4/3)(a + 4 — v2) Bo ho fuck ae 4 = av 4, —(or/e%ee) wp Ges) asa | (283) B,(0) NACA TN No, 1832 n (1 + 4/3)(a += v4) r+% _ fa 4) how Ltd =v (c*/e2Go) @ (214) *3 ep (1+ (24 4/3) | (hs tL + 0/3) Neglecting the generally amall effect of finite E_ in equations (213) and (214), that is, putting 4 = 0 in these equations, there may be written instead _aNi= era +4 paw oe ion B,(0) — a a E+] (215)* w(0) == es w- GAG vee fey exe Veep C — ai v8 & Equations (215) and (216) contain the noteworthy fact that the correction factors for the effect of transverse shear are indepenient of the ratio t/n of face-layer thickness to core thickness, ‘The complete formilas of course mist and do contain the influence of the core thickness h. Tt ig further noted that, while equations (211) to (216) have been derived for the case that my! < km, for which the complex solution holds, thoy are also valid, as is readily shown, when dm > m't, Comparing equations (213) and (214), and (215) and (216) with the equations listed in footnote 16 it is seen that: (1) The effect of transverse shear modifies the deflection due to Q and the rotation due to My but not the other two coefficients, (2) the effect of transverse normal stress enters all four coefficients tut only ina minor vay, and (3) the reciprocity relation that the deflection due to My {a the game as the rotation due to Q is carried over fron the theory without the extra effects. 2 WACA TW No, 1832 + = 0.1 inch, For e numerical example the following data are chose: h = 1 inch, a = 10 inches, E/By = 100, Ep/Gp = 200, v = 1/3, 0.055, and, from equation (197), = 2h2 x02 and 2 = gb? 100 m= [OE X20 - 0,106 a "2" \Vioo x ex ios” °° ‘Then, according to equation (204), fel? = 0.273 +k = V2 0.18 + 0.173 = 0.84 while without transverse shear deformation (™ = 0) the value of & +E = 0.59, According to equations (211) and (212), the effect of transverse shoar in this case is to increase the rotation due to the edge moment in the ratio 0,.84/0.59 = 1.42, an effect of 42 percent. The same increase ia found for the deflection due to the edge shear force. Rotation due to the shear force and deflection due to the moments are practically unchanged. likewise, the effect of trensverse normal stress in this case is of negligible importance. Ae a further numerical exemple there ia chosen t = 0,05 inch, h= 1 4nch, a = 20 inches, B¢/E_ = 1000, z/Go = 2000, 1 1.05 x 0.05 . E 2 2:0 0.0 1000 = 0,065, and, from equation (197), =, /—2- m = 35\| Dose x 05 20 0,483 L WT = 0,05 —i = 0.218 inc xa V 1-0 and 2 NACA TN No, 1832 B From equation (204) then Jel? = 0.095 4B = Y2 \[0.23% + 0.095 = 0.82 while without traneverse shear deformation (m = 0) the value of k+K=0,hl, Thus the effect in this case is to increase edge rotation due to edge moment and edge deflection due to edge shear force in the ratio 0,82/0,4h = 1.87, an effect of 87 percent. Infinite circular cylindrical shell acted upon by transverse line load — Calculation ia restricted to the determination of deflection and ending moment at the section x= 0 where the line load of intensity 200 ie assumed to act. ‘he result of the foregoing paragraph may be used as follows. Consider the infinite shell cut in two parte ati the section x = 0 and assume & bending moment My of such magnitude that the slope (0) is zero. According to equation (211), this gives ---0 0 (217) k+E Vanes me and therewith ctw(o) (+B |x? | ms? + ome aD) G+a/3e® kv “O° Tomes hag Equations (217) and (218) become, with equations (204) and (197), % ayfoe —2sa—ve ote fa2D* (1 + A)(1 + 2/3)” (h + t)Gp 1 + 2/3 (e219) Mo. ™ NACA TN No, 1832 deka ve oe /ee OS ia 2a Dee) Egy (220) ae eae or /at pe (1 + A)(1 + 2/3) erie TB To give these formas a somewhat less unvieldy appearance, the effect of finite Eo, that is, 4 = 0, my again be neglected, as is permissible in most cases; and there may be written (221) * (222)* Some mumerical examples are given as follows. Taking t = 0,1 inch, a = 10 inches, Er/G_ = 200, and v = oF traneverse shear deformtion reduces My to 1/V2.05 times the value which holds when Gy = ®, that 1, there 1a about a 30 percent reduction in Mo, At the sano tine the deflection under the line load te 3.05/J2-05 = 2.1h times what 4t 1a when Gy = 5 that 1a, there 16 an increase of about 115 percent in w(0). Taking t = 0,05 inch, @ = 20 inches, and Er/G, = 200, Mo te decreased ty a factor //5 = 0.89, while w(0) 1s increased by & factor 1,5/\1.25 = 1.34. Taking + = 0,05 inch, a = 20 inches, and E,/G, = 2000, M is decreased ty a factor 1/\/3.62 = 0.526, while w(0) ie increased by a factor 6,25/\/3.02 = 3.29. NACA TN No, 1832 B Equation (220) for w(0) may be compared with equation (116) for the cirouler plate of radius a, This comparison shows that, while for the plate both the ratios t/a and (h++t)/a enter into the correction factor, the correction factor for the cylindrical shell conteins the ratio t/a only; that is, the corrections (tut not the results) are independent of the ratio of face-layer thickness to core thickness in this case of a cylindrical shell. Spherical Shelie In conformity with customary usage, the following notation is introduced: ty) =a bo = a8 a = gin g Rp =Rp=a Nop = Ng Me=Mar=Mp = ay %=% Ma = My Moo = My (203) Me = Mio P= 2% Py = Be my = my m= ape wav By = Bg Bo = By Attention 1s here restricted to problems with rotational eymetry and the following relations are used: a( ) Re =0 (22h) Ngo = Ge = Mig = Pg = tg = v= Bg = Tho aifferential equations of equilitrium (05) to (28), (31), and (3h) decom, setting dC AAG = aC )/ag = ()* 6 NACA TN No, 1832 = aig)" = cos fiig + ein fag + 2 sing rg = 0 (225) (en gag) - sin NY + No) +asingqg=0 (226) (em Jig)! — cos fly — 2 sin fay + 2 sing mg = 0 (227) + (Hp + le + a6 =0 (228) the strese-strain rolations (56), (59) to (61), and (64) to (67) pecone, if there 18 set in accordance with equation (63 (m+ t)t Ee 4 (229) @+ ang - @ = 2K = oftee +het 3) (230) @ a BN, 7 (@ S 2g : on(z cot fee ag #) (231) Y = (n+ HG (@g +H (232) (1+ Aug — (V—2)Me = Eby" + #) (233) 5 (1+ My — (Vg = Ee cot P+ £) (238) NACA TN No, 1832 7 There is first given e simple special solution of this system of equations and then a generalization is obtained of the two simltaneous equations for Qg ani fy which ere fundamental in the theory of homogeneous isotropic shells, Uniform stress distribution in a spherical shell. Set in equations (225) to (234) pg = mg= 0 and assum that Ng, Np, Gy, Wy, and My are independent of g. From equation (225) it follow that: Ng = Np = No (235) y= From equation (226) it follows then that No = ag (236) and from equation (227) it follows that Mg = Mg = Mo (237) Bquation (228) gives tm = 8 — 2Mo/(h + tle (238) ‘In equations (230) and (231) set u=0 for reasons of symmetry and obtain G§r- yo Fe Meas cen) B WACA TN No, 1832 or, with Ng from equation (236) end % from equation (229), (o*/a)w = 201 - v)ag (2h0) Equation (232) 1s identically satisfied when fg = 0, Equations (233) and (234), in conjunction with equation (237), give (1 +2) ¥)Mp = (D¥/a) (8/8) =a eas w Be 5 c or My = 2+ ter. (24a) 1+2-v Then, from equation (238), a (2k) Sm = te Teeny Equation (22) may be compared with equation (136) for the ciroular ring. According to equation (46), there are obtained from equations (236) and (241) the following expressions for the stress resultants in the outer ("upper") and inner ("lower") face layers: (Btn - eG +) 28 Ty anv, (243) ->35) Teeny Comparison of these results with the corresponding results for the circular ring (equations (135)) shows that for given values of q and s there is a greater difference between N, and N, in the NACA IN No, 1832 7 9 spherical sholl than there is in the circular ring, the reason being the relatively larger influence of the e-term in equation (213), For a specific example, it is again assumed that the radial load ia applied entirely to the inner face go that qy = 0 and, according to equations (20)'and (22), (2h) 2 e=- 3 - at) a Substitution of equation (2h4) in equation (243) gives G+ bets, - ( - ht) aa, yoy 2a ye 2a 1+2—¥ : (245) As 9 numerical example, taking % = 0.0595, as in the example given in ‘the section entitled "Closed circular ring acted upon by uniform radial load," and V = 1/3, it 4s found that the factor in Nz which contains the effect of the core flexibility ie (1 + 0,36)/(1 + 0.18) = 1.15. ‘Thus, where for the circular ring there was a 6-percent stress increase, there now ig a 15-percent stress increase. Reduction of axieymetrical problem to tvo simltancous equations for _Qg and fig.— The fundamental results of reference 10 for Homogeneous shells nay be readily extended to sandwich shells, as follows: Equations (225) and (226) are used to express Ng and Ng in terms of Qg. Ng = cot gag + Fy(9) (286) Ng = Oy" + Fo(9) (247) 80 NACA TW No, 1832 Tn equations (246) and (2h7) the functions Fy, and Fo are given by . wa {e cos § — pg sin p)ain gag (248) fing Fy)! + apg sin 4] (249) Next the @ieplacement components u end w ere expreséed in terms of Q%, by moans of equations (230), (231), (246), ana (247). Subtraction of equation (231) from equation (230) gives Bear — cots v) (2 + v)(tig ~ Ne) a+ v[-@" = cot ag) +7 - FJ] (250) Equation (112) ‘te integrated to (c¥/a)u = (1 + v) (ag + ¥3) (251) where F3 is given by Fy(P) - Fol) F3 = -sin pf Oa seh “ns ag (252) Equations (251) ani (252) are introduced into equation (231) end the following expression is obtained for vt (C#/a)w = (1 + 2/3)(cot ag + Gy") + Fy (253) NACA TN No, 1832 81 where Fy is given by 4 te 8 $+ (1+ woot g Fy + @ + Dy . (v - ry (254) Equations (251) and (253) ere introduced into equation (232) f and the first of the two simltaneous equations for %y and Bg is obtained in the form tore fe DOH oe) om +O sys 5] which may be rearranged to read1? eae apt ~[sotev - B= iat & wee ale + mn Bg = F5(A) (255) The function Fs is given by (2+ Fs + Ts M5 i Introducing the operator L=()' + cot $( )' — cot29¢ ) Twhen 4 = 0 and Go =e and when no external loada are present, this equation checks with the first of equations (a) on page 469 of reference 9, ge NACA TN No, 1632 equation (255) may finally also be written (L = Hn) + oe 8g = F5(9) (e57)* where a 4\_ 1 _ on Gat vA 28) ‘The second of the two eimtancous equations is obtained somewhat more directly as follows: Write equations (233) and (23+) in the form My ele + r)pg" + (v — deoot Bg — ve + 2X(1 + v) + (1+ v)e/e] (259) D¥/a eat % Pane Gl + root # Bg +(V ~ Bg + (1+ V)e/e] (260) Introduce equations (059) end (260) into the moment equilibrium equation (228) and obtain Dt/a 1-¥ 4 2n(1 + v) [ns + cot f Bg’ — (ot? +t Pg + (1+ v)a! Be + (2 + W)at/Be] ~ dy + ay = ° (261) NACA TN No, 1832 83 ‘Again, using the operator L, this my be written in the form = santa + vilag = Fe(9) (262)* ‘The function Fg is given by Fg = 02 + vit = Bb- sara s vag (263) Bquation (262) may be compared with the second of equations (g) on page 469 of reference 9. Analysis of edge effect for spherical shell.— The special case of no distributed surface load and no concentrated load at the apex of the 7 shell 1s obtained by setting =F =0 Following again a known procedure from the theory without transverse stress deformation, there may be set a % sin ¢ (264) fy =» 1 sin g + Gta, a 8 erg 2 °° 9 ome . (265) 8h NACA TN No, 1832 and g'', Introduction of with corresponding formlas for and (262) gives equation (264) into equations (257 atts (aed = 2 cot*#)q, + roy? (266) ' v= 3 2 ay" ~ (CoB b + } cota es - FB - sare vig =o (267) Attention 1a restricted to the cases for vhich F5 = Fg =0 in equations (266) and (267) and the problem is considered of the shell subject to edge loads 0%) )o and )o ata section $ = Jo, Asouning that cot fo te not large compared with unity28 and that: the effect of the edge loads is restricted to a narrow eige zone 80 tnat [aa] < |e", | Pal <« |e:"*], equations (266) and (267) may be simplified to om qt 41 ha + TB 8 = 0 (268) an - eh - +22 + vq =o (269)* Equations (268) and (269) show that the influence of finite Bo(% # 0) in the edge-effect problem consists, except in extreme circumstances, in minor modifications of the results for Ho =e. The quantity which represents the influence of finite G, and which is wy . 2t ‘het a (270) may, however, in practical cases ve large compared with unity end not of negligible influence on the results, Ben cot fy >>1 the shell is termed a "shallow" shell which is not considered in hat follove. NACA TN No, 1832 85 Equations (268) and (269) may be compared with equations (180) and (182) for the cylindrical shell. This comparison shows that the influence of finite G, in the edge-effect Problem is of the same nature for the spherical and cylindrical shells, Thus, results of the sano quantitative nature will be obtainable as in the section on cylindrical shells under the headings entitled "Finite circular cylindrical shell acted upon by edge moments and forces" and “Semi~infinite shell acted upon by edge bending moment and shear force." Tais work is not herein carried further to specific applications. Tt 16 apparent that such applications may be worked out with hardly any more difficulty than when the effect of the core deformability is not ‘taken into account. CONCLUDING REMARKS A system of basic equations has been derived for the analysis of suall-deflection problems of sandwich-type thin shells. This aysten of equations reduces to love's theory of thin shells when the transverse shear and normal stress deformability of the core of the sandwich 1s of negligible importance. The system of basic equations hae been applied to @ mmber of specific problems from the theory of plates, circular rings, circular cylindrical shells, and spherical shells, and it has been found that the effects of both transverse shear and transverse normal stress deformation may be of such magnitude that an analysio which disregards them gives values for deflections and stresses which are appreciably in error. Numerical calculations have been in the nature of sample calcu~ lations, 1llustrating both the use of the equations and the possible effecte of using then. Examples have teen chosen from the point of view of relative simplicity as well as with the thought to {lluetrate most clearly the consequences of the extra deformations which have ‘deen taken into account. It is unavotdeble that, in so doing, some of the examples may be of little interest for aircraft structural analysis and thet some problems may not have been analyzed which would have well fitted within the contents of this report and which at the same tine would have ‘been of considerable practical importance. ‘The general analysia has been restricted by the following two order-of-magnitude relations: (1) t/n<<1 and (2) tE,/nE, > 1, vhere t is the face-layer thickness, h is the core-layer thickness, Ep is the elastic modulus of the isotropic face-layer material, and Ey is the elastic modulus in the transverse direction of the core-layer material, Therewith it is felt that very likely nearly all situations have been covered in which the effect of transverse core flexibility 1a of significant practical importance. It 1g evident, however, that if desired the theory could te extended so as 86 to include cases where oF relations ere not satisf’ analysis is the omission o: effects. NACA IN No. 1832 me or both of these two order-of-magnitude ied, ‘The mein limitation of the present f ell finite-deflection and instability Massaclusetts Institute of Technology ‘Cambridge, Mass., May 26, 1947 NACA TN No, 1832 87 REFERENCES 1, Refesnor, Eric: On Bending of Elastic Plates. Quarterly Appl. Math., vol. ¥, no. 1, April 1947, pp. 55-68. 2, Reissner, Eric: A New Derivation of the Equations for the Deformation of Elastic Shells. Am, Jour, Math., vol.-63, no, 1, Jan, 1941, pp. 17718h, 3. love, A, E. H.: A Treatise on the Mathematical Theory of Elasticity, Fourth ed., The Untv. Press (Caubridge), 1934, PB. 90. 4, Trofttz, E.: Ablettung der Schalenbiegungs-gleichungen mit dem Castigliano' schom Prinzip. Z.f.a.MM., Bi. 15, 1935, pp. 101-108, 5. Donnell, L, H.: Discussion of “The Effect of Transverse Shear Deformation on the Bending of Elastic Plates" by Eric Reissner. Jour. Appl. Mech., vol. 13, no. 3, Sept. 1946, pp. AND — A050. 6, Boskin, Leon: General Circular Ring Analysis. Aircraft Eng., vol. XVII, no. 194, April 1945, pp. 126-132, 7. Retesner, H.: Forminderung einer dtinnwandigen, an den Rindern frei eufliegenden, beliebig belasteten Zylinderschale, Z.f.a.M.M., BL. 13, 1933, pp. 133-138. 8. Biezeno, C. B., and Koch, J. J.: Some Explicit Formlas, of Use in the Calculation of Arbitrarily Loaded Thin-Walled Cylinders. Nederl, Akad. Wetensch. Proc., vol. 44, 1941, pp, 505-512, + Timoshenko, S.: Theory of Plates and Shells, MoGraw-Hil1 Pook Co., Inc., 1940. 10, Reissner, H.: Spannungen in Kugelschalen (Kuppeln). Miller—Bresleu- Festechrift, 1912, pp. 181-193. NACA TN No. 1832 SE Figure 1.- Element of composite shell, showing coordinates and dimensions, eee NACA TN No, 1832 & Figure 2.- Element of composite shell, showing location and orientation of stress resultants in face layers and core layer and orientation of external loads. seg ‘on Ny vow {gtteys Twoyaoyds pue “etreys TwoTapaTTéo awrnoayo ‘sPupt autnoxpo ‘sequtd yetz Buyuxeouos ameyqoad oO woTanTos ey; 04 por Tdde ox ByTnEex TexoUOS ons, ‘uo;ywmI0zep sse.38 Tautou eszeaster 30 y00330 oy (z) puw woT;EmZoJep Twoys epreasuUT} Jo 400330 ous (T) :8TTEs Pears 50 £70043 eyy UF ATBNOTAeLd PexepFEUCO woEd, you eauy yoTYA ‘meTqoud quoBesd 643 UF yUeqZ0dzy oxe a300530 Ong 40YR PUMOS BT 4 “STTSys O74euTS UTY SnoeusHomoy Jo £l0ey} UACUR ey Jo HOTeuOAEe Ue SF Aaoou, stay ‘sTroys edéy—qorapuEE Jo BuTyOWex8 Pus Surpueq Trwus 40s podotesep ey £z0043 ¥ pay

You might also like