PRINCIPLES OF STRUCTURAL STABILITY THEORY (WAVELANDBUCKLING OF COLUMNS 4.1. INTRODUCTION Slender columns are subject to a type of behavior known as buckling. As long. as the load on such a member is relatively smail, increases in the load result, only in an axial shortening of the member. However, once 2 certain ertical Joad is reached, the member suddenly bows out sideways. This bending gives rise to large deformations, which in turn cause the member to collapse. The load at which buckling occurs is thus a design criterion for compression members, Tension members as well as short stocky columns fail when the stress in the member reaches a certain fimiting strength of the material. Once this limiting strength of the material is known, it is a selatively simple matter Co determine the load-carrying capacity of the member. Buckling, however, does not occur as a result of the applied stress reaching a certain predictable strength of the material, Instead, the stress at which buckling occurs depends on a variety of factors, ineluding the dimensions of the member, the way in which the member is supported, and the properties of the material out of which the member is made, The determination of the buckling stress is thus 2 relatively complex problem, It is the solution of this problem that will be our main concern in this book. 1 buckling does not take place because a certain strength of the material 1Buckling of Columns chr is exceeded, then why, one may’ ask, does a compression member suddenty buckle? Although itis not possible to answer this question dizeetly, one cay ‘make certain observations about the buckling phenomenon and thus explain atleast partially what is taking place. One such observation, which elucidates #p an excellent manner the phenomenon of buckling, is given ia the following Paragraph quoted from Structure in Architecture by Salvadori and Helle. (Ref, 1.1): “A slender column shortens witen compressed by a weight applied {0 its top, anc, in so doing, lowers the weight’s position. The tendency of all weights to lower their position is a basic law of nature. It is another basis low. cof nature that, whenever there isa choice between different paths, a physica Phenomenon will follow the easicst path. Confronted with the choice of bending out or shortening, the column finds it casier to shorten for relatively small loads and to bend out for relatively large loads. In other words, when the load reaches its buckling value the column finds it eas by bending than by shortening.” The buckling ioad thus appears to be the limiting load under which ax compression in an unbent configuration is possible. It will be assumed her and proved later that the transition from the st bent configuration at the buckling load occurs because the straight configuration ceases to be Stable. fn Article 12 the idea that the buckling load marks the limit of stability of the unbent configuration will be devefoped into a procedus the buc to lower the toad al for evaluating 1.2 METHOD OF NEUTRAL EQUILIBRIUM ‘The concept of stability is frequently explained by consi ing the equilibrium Of @ rigid ball in various positions, as de picted in Fig. 1-1 (Refs, 1.2 and 13), Fig. 1-1 Stability of equilibium i @ (Adapted fromm Ret. $2), Although the ball isin equilibrium in each position tion reveals the existence of important di UV the ballin part (a) i displaced slightly from ts original position of equilib rium, it will return to that position subsequent to the removal of the dieturb. ing force. A body that behaves in this manner is said 10 be in a state of arable equilibrium. By comparison, the bal in part (b), if cis displ fr its position of rest, does not return, but instead continues to move farther shown, a close examina erences among the three situations. displaced move farther away. b disturbance has move The ball in F region between anc and C itis unstable ball isin a state of ne column bi unstable st that load of the ball in Fig. | small Joads, but itis neutral equilibrium ¢ jn the column, simil ‘which the straight ¢ at which neuteal equ critical loa. ro determine th which the member¢ ent configuration. critical loads is 1.3 CRITICAL U sh ven as sim fairly complember suddenly lireetly, one can ind thus explain which elucidates in the following Sori and Heller weight applied tendency of all other basic law aths, a physical 1 the choice of on for relative et words, when > lower the load der which axial e assumed here iteonfiguration on ceases to be imit of stability 2 for evaluating he equilibeium of equilibria 12, close examina. atee situations. tion of equilib of the distur’- state of stable slightly 1 move farther anna Critical Load of the Euter Column — 3 away from the original equilibrium position. The equilibrium of the ball in part (b) is a very precarious one. It is called unstable equilibrium. Part (c) depicts yet another possible type of equilibrium. Here the ball, after being displaced slightly, neither returns to its original position nor continues to move farther away, Instead, it remains at the position to whick the small disturbance has moved it, This behavior is referred to as neutral equilidium, The ballin Fig. 1-2 is in equilibrium at any point along line ABC. In the Fig Stabhity surface a A and B the equilibrium is stable, and in the e and C it is unstable. At point 2, the transition between the two regions, the ball isin a state of neutral equilibrium. In Article 1.1 it was pointed out that a column buckles at a certain load because the straight configuration becomes, tnstable at that load. The behavior of the column is thus very similar to that of the ball in Fig. 1-2, The steaight configuration of the column is stable at small loads, but it is unstable at large loads. If itis assumed that a state of neutral equilibrium exists at the transition from stable to unstable equilibrium in the column, similar to that present for the ball in Fig, 1-2, then the load at which the straight configuration of the column ceases to be stable is the load at which neutral equilibrium is possible. This load is usually referred to as the critical load. ‘To determine the critical load of a column, one must find the load under which the member can be in equilibrium both in the straight and in a slightly bent configuration. The technique that uses this eriterion for evaluating, critical loads is called the method of neutral equilibrium. 1.3. CRITICAL LOAD OF THE EULER COLUMN Even as simple a structural element as an axially loaded member behaves in @ fairly complex manner, It is therefore desirable to begin the study of columnsBucking of Columns cht with @ very idealized ease, the Euler column.* The axially loaded member shown in Fig, 1-3a is assumed to have a constant eross-sectional area and to van oo 13 Euler col, -y bbe made of a homogeneous material. ta addition, four assumptions are made: |. The ends of the member are simply supported. The lower end is at~ tached to an immovable hinge, and the upper end is supported so that it can rotate freely and move vertically but not horizontally, ‘The member is perfectly straight, and the load is applied centroidal axis, The materia along its ‘obeys Hooke’s law, The deformations of the member are sm: all enough so that the term (0° is negligible compared to unity in the expression for the curvature J + G7 )P*, Hence the curvature ean be approximated by y".4 In accorda ee with the criterion of neutral equilibrium established in Article 1.2, th tical load is that toad for which equilibrium in the slightly “The Euler column takes is name from the man who, in the year 174, presented the first accurate stability analysis of a column, Although i is customary touky to robe te a ‘iewply supported column as an Euler column, ler in fact analyzed a member bred a one fend and free a te other in his famous retse, which can be found in Re 1 {The atention of the reader is drawn co the fact thatthe symbols)” andy” ate used to denote the second and fist, walives of y with respect to fret bent configuration st taken as shown in the distance « from the ¢ Equating this express Equation (11) solaton, et us di foregoing simpliying behavior and’ stall Becomes avaiable Solution is dul hinged supports are sis ofthe equation Ine the task of "Te salon of Bg, (1.1) becomes. The solution of hon efficients is always Eq. (1.3) leads to m Making use of the and the fact that be satisfies Eq. (1.3) » rewritten in the focnt ded member area and to mptions are dis ate orted so that ed along its hat the term ‘he curvature ed by yt stablished in 1 the slightly presented the ser xed atone 14 5 are used co are Critics Load of the Euler Column & bent configuration shown in Fig. 1-3b is possible, If the coordinate axes are taken as shown in the figure, the internal resisting moment at any section, a distance x from the origin, is Me = Ely" Equating this expression to the externally applied bending moment, Py, gives, Ely! +P 0 aay Equation (I.1) is a linear differential equation with constant coefficients As such, its solution is readily obtained, However, before considering this solution, let us digress somewhat and see what form Eg. (1.1) takes if the foregoing simplifying assumptions are not made. If the assumptions of elastic behavior and small deflections are not made, the modulus E in Eq. (1.1) becomes a variable, and curvature »” is replaced by p"j(l + (y'P?. The equation then has neither constant coefficients nor is it linear; as a result its solution is difficult to obtain, If the assumptions of concentric loading and. hinged supports are not made, there are additional terms on the right-hand side of the equation. This makes the equation nonhomogeneous, but does not make the task of obtaining a solution difficult ‘The solution of Eq, (1.1) will now be obtained. Introducing the notation ef (1.2) Eq, 1) becomes, wt By =0 as ‘The solution of homogeneous linear differential equations with constant co- efficients is always of the form y = e™, Substitution of this expression into Eq, (1.3) leads to m= -Lik, Hence the generai solution of Eq, (1.3) is, ya cet Making use of the relation en 608 ka of sin kx and the fact that both the real and imaginary parts of a complex function that satisfies Eg. (1.3) must also satisty the equation, the geeral solution can be rewritten in the Form ye d sin ke + Bos kx «es© Buckling of Columns To evaluate the arbitrary constants A conditions B, we make use of the boundary 0 ax=o cs 6 axel The first of these conditfons when substituted to Eg. (1.4) leads to B Consequently, Asin kx 06 From the second condition one obtains Asin k= 0 This relation ean satisfied in one of wo ways; either A sin ki ° 14 = 0, keand consequently P can have any value, This result is known as Uivial solution, because it confirms what is already known, that a column is in equilibrium under any axial loud P as long as the member remains per Teetly straight. If sin kl = 0, then kts mm where 1 = I, 2,3... Substitution of this expression into Egs.(J.2) and (1.6) leads to «7 ang Asin BE (8) At the foads given by Ea, (1.7) the column ean be in equilibrium in a slightly bent form. The shape of the deformation is given by Eq. 8). However, its amplitude is indeterminate, since 4 can have any value when sin ki ~ 0 ‘The vaiue of P, obtained by setting m equal to f, is a9) Ant This load is known a neutral equilibrium i ccolumn ceases to be ‘The behavior of Fig. 14 Behavior of ccan be summed up straight. At the Eul the column can rem sminate amplitude, exists at the E from stable to west Equation 1.71 loads larger than Without going into bbe assumed to be significance as far ‘The Euler load as the eritical load terms, it bas been changeably. He a column suddenly that the term criti sium is possible fo words, buckling i loaded durin idealized theoretic thus be referred 1 LINEAR Ct AN EIGEN 14 The smaildefietchy se use of the boundary ay (14) teads to a6) his result is known as cnoven, that a column member remains per- ‘o Eqs. (1.2)and (1.6) a7) (a8) librium in a slightly (1.8), Howaver, its when sin & anid Liner Column Theory—An Eigenvalue Problem — 7 This load is known as the Euler load. It is the smallest load at which a state of neutral equilibrium is possible, Hence itis also the smallest load at which the column ceases to be in stable equilibrium ‘The behavior of the Euler column, represented graphically in Fig. 1-4, Fig. 14 Behavior of Eule column, oy ccan be summed up as follows. Up to the Euler load the column must remain straight. At the Euler load there exists a bifurcation of equilibrium; that is, ‘the column can remain straight or it can assume a deformed shape of indeter. minate amplitude. This behavior signifies that a state of neutral equilibrium exists at che Buler load and that the Euler load, therefore, marks the transition from stable to unstable equilibrium, Equation (J.7) indicates that for values of a greater than |, thereexist other loads larger than the Euler load at which neutral equilibrium is possible Without going into any detail at this point in the book, these larger loads will be assumed to be valid: mathematical solutions to Eq, {1-1}, but devoid of significance as far as the physieal phenomenon of stability is concerned The Euler load obtained in the foregoing analysis is sometimes referred to as the critical load and sometimes as the buckling load, Regarding these two terms, it has been suggested by Hoff (Ref, 1.5) that they not be used inter- changeably. He advocates that the load under which an actual imperfect column suddenly bows out iatcrally be referred to as the buckling load and that the term critical load be reserved for the load at whieh neutral equilil rium is possible for a perfect member according to a linear analysis. In othet words, buckling is something that can be observed when a real column is Joaded during a test, whereas the term eritical load refers to the solution of an idealized theoretical analysis. The Euler load obtained in this article should thus be referred to as the eritical load of the columa, 1.4 LINEAR COLUMN THEORY: AN EIGENVALUE PROBLEM deflection column theory presented in Article 1.3 is based on a ferential equation and is, therefore, known as the linear column2 Bucking of Columns a7 theory. By comparison, the large-deflection theory of columns to be studied later is based on a noniinear differential equation and is referred to as 2 nonlinear column theory. Although the smali-deflection column theory ean be considered tinear when contrasted with large-dellection theories, it is not linear in the same sense of the word as simple beam bending theory. In th fatter, equilibrium. is based on undeformed geometry and deflections are Proportional to the applied loads, which is certainly not true for the Euler column, Thus Tinear column theory is evidentiy different from both simple- flexure and large-deformation column theory. As a matter of fact, it does belong to an entirely different class of problems known as eigenvalue prob lems. ‘These problems are characterized by the fact that nonzero solutions for the dependent variable exist only for certain discrete values of some arametcr. The values of the parameter, for which nonzero solutions exist, are known as eigenvalues, and the solutions as eigenveetors. Only the shape and not the amplitude of the eigenvector can be determined in an eigenvalue Problem. In a stability problem such as the axially loaded column, the loads im BUiP at which nonzero defections ate possible are the eigenvalues, and the deflected shapes that can exist at these loads are the eigenvectors, The smallest eigenvalue is the critical toad and the corresponding eigenvector is the buckling mode shape, 4.6 BOUNDARY CONDITIONS ‘The first step toward generalizing the results obtained in Article 1.3 is to con- sider boundary conditions other than hinged-hinged supports, Case 1. Both ends fixed ‘fa column is builtin at both extremities, it can neither translate laterally Ror rotate at these points. As a result, bending moments, M,, axe induced at each end of the member, as shown in Fig. 1-52, when the column is deflected slightly. Equating the external moment to the internal moment, ata section a distance x from the origin (Fig, 1-5b), one obtains M, M or ie C.10) where k* = PEL. The solution to Eq, (1.10) consists of a complementary and « particular part, The former is the solution of the homogeneous equation, It is given by 5a, U.4). The particular solution is any solution to the entire equation, such An tS Inteanns to be studied 1 referred to as a umn theory can theories, it is not ng theory. In the d deflections are rue for the Euler ym both simple- + of fact, it does cigenvaiue prab- ‘onzere solutions > values of some © solutions exist, 5. Only the shape in an eigenvalue ‘olumn, the loads eigenvalues, and ‘igenvectors. The 1g eigenvector is ile 1.3 is t0 con: ts, 10) ind a particular 1. [tis given by equation, such Are 15 Boundary Conditions Se! ot) be ( le | [a FP eae <= Fig. 1-5 Foxed-fixed eoluran.10 Buckling of Columns 5 y= Mg/EIK*. Thus the entire solution is = Asin kx + Boos kx + Mo Fm Asin ke #008 kx + Me ay Where 4 and & are arbitrary constants to be evaluated using the boundary conditions 6 y=0 ate=0 y=0 atx=? ‘The frst two conditions are satisfied if 4=0 and Hence y Moa — cos kx Med = cos kx) ‘The last condition leads to the transcendental equation cos k= 10 ‘The smallest nonzero root to this equation is oe from which seer oe me Mos — cos 288 ad ya Mf os 28) 19 Equation (1.12) indicates that the eritical load of @ column with fixed ends is Four times as large as the critical load of a hinged hinged column, Using Eq. (1.13), it can be shown that inflection points, that is, points of vero moment, exist at x= H/4 and x= 3//4. The central portion of the member, between the quarter points is chus equivalent to a hinged-hinged ‘column of length 1/2 (Fig. 1-5e), whose eritica| ioad is EL = Fat (4) ‘The critical load of the pseudo hinged-hinged column that exists between the inflection points of the fixed-fixed column is thus equal to the critical load of the fixed fixed column, It will be demonstrated in the succeeding pages that are 15 the critical Joad of ¢ column. The length engaht of the membe Case 2. One end fix The column shov and rotate at its upp ment 6 at the uppe Equating the interna x from the base (Fi, exncc---- where k= P/EL ‘The bs ndary con ‘They are satisfieddan ing the boundary a. uy with fixed ends is olume, that is, points of | portion of the a hinged-hinged aaa) 1 exists between >the eritical load seding pages that ars Boundary Conditions 11 the critical load of any column can be obtained from an equivalent Euler column. The length of this equivalent Euler column is called the effective length of the mem Case 2, One end fixed! and one end free ‘The column shown in Fig, 1-6a is builtin at the base and free to translate ‘and rotate at its upper end. A small lateral deflection gives rise to a displace- ‘ment 5 at the upper end of the member and a moment PS at the base, Equating the internal moment to the external moment, at a section a distance x from the base (Fig. 1-66), leads to. x Ls \ i 5 | | bP ee a jae 1 in Equivaient euler Sa hs ° lr e eo te Ely" + Py = Pb or y’ key = KS LS) where kt = PIET, The solution 19 Ba, (1.15) is = Asin kx + Bos kx +5 a6) The boundary conditions st the base of the member are atx They are satisfied if12 uckting of Columns cht Aso Hence ~ 005 kx) ‘The boundary condition at the upper end of the member psd axat is satisfied if cos k= 0. ‘The smallest nontrivial root of this equation is Moh which leads to a7 and ) G.18) Equation (1.17) indieates that the critical toad of a column fixed at one end and free at the other is one fourth the Euler joad. From Eq. (1.18) it can be shown that the deflection curve of the column consists of one quarter of a sinewave or one halt the deficet of hinged hinged column (Fig. i-6c). The effective length of the equivalent Euler column is therefore equal to 2/, and the critical load of a fixed column can be expressed in the form fixed-free ed (119) Case 3. One end fixed and one end hinged The column shown in Fis 17 is hinged at the base and builtin atits upper tend to a support that is constrained to move along the axis of the menibee A small fateral deflection gives rise to a moment Mg at the fixed end. and cheat forces of magnitude Maj at each end of the member. Equating the internal ‘and external moment, at a section a distance x from the base, leads to Ely! Py = Max 4 pty a Ms i (1.20) | Arts where k* = P/EL Th ‘The boundary condi and are satistiee if Hence ‘The boundary condi leads to the transces; Ares Boundary Conditions 12 ain ~ (18) Fig. 7 Fixing column, column fixed ut on _ where k? = P/ET, The solution to Ea, (1.20) is it ofthe fied-free fsisnesiete y= Asinks + Boos kx + Mes 2b th ofthe equivatent iadcnk aan ‘The boundary conditions y=0 atxmo ny and Y=0 axas are satisfied if B=0 and A a built in atts upper eee °P Rreasi sof the member. A xed end, and shear Hence uating the internal base, leads 10 ‘The boundary condition pao a leads to the tran i cendental equation 20) tan k= ki14 Buckling of Columns The smallest noozero 108 of his bation bea which ea to roe am ME= 4 sonsn (aan2)) and ya Mls Lona (rz) 02s) Equation (1.22) gives the evitieat load of a colum his xed atthe other. The corresponding deficction curve has an inflection point tx = 0.7! Hence the effective length i 0.7), and the critical load can be given in the form inged at one end and P= Sie 0.24) Case 4. Elastica ly restrained end in most actual structures the ends of columns are neither hinged nor xed. Instead the columns are usually rigidly connected to other members, which Permits a limited amount of rotation to occur at the ends of the columns Supports of this type are referred to as elastic restraints. They are se named because the restraint that exists at the end of the column depends on the clastic properties of the members into which the column frames Let us consider coluran that is hinged at the base and clastically restrained by 2 beam at its upper ead, as shown in Fig, |-8a. The beam ig assumed to be fixed to a rigid support at its far end, For simplicity, the Fength and stiffness ofthe beam are taken as equal (othe length and stifinass respec. tWvely of the column, Jn order to linearize the problem, itis assumed that there is no bending in the horizontal member prior to buckling. However, when the ettical load is {eached, the bending deformation of the column will induce bending in the beam. Due to its stifness, the beam resists being bent by the column and exerts a restraining moment Mf, as shown in Fig. 1-86, on the column. In ccaleulating the critical load the forces produced by the bending deformations 4s well as the deformations themselves are assumed to be infinitesimal, The shear forees @ acting on the beam are therefore negligible compared to P. and it is reasonable to assume that the axial force in the column remaing equal to P during buckling, Taking the coordinate axes as for & sezment of the column gives shown in the figure, moment equilibrium AS ‘The boundary conditic is satisfied if and the condition a 11 requires that Hencecht x] (0.23) hinged at one end and .e has an inflection point critica! load can be given (1.24) either hinged nor fixed. > other members, which xe ends of the columns, ints, They are so named column depends on the umn frames, be base and elastically Fig, I-8a. The beam is simplicity, the length ‘ath and stiffness respec- at there is no bending in when the critical load is | induce bending in the Dent by the column and 8b, on the column, In © bending deformations to be infinitesimal. The 2ligible compared to P, in the column remains moment equilibrium ar t5 Boundary Conditions 18 eral fe eee wc, > tay le | : pega ps0 or (0.35) where A? = PJEL The solution of Ea. (1.25) i y= Asn e+ Beas 4 ME (126 The boundary condition at the lower end of the member =0 atx=0 is satisfied if B=0 and the condition at the upper end ys0 ate= requires that Hence16 Bucking of Columns cnt Since the column is rigidly connected to the beam, the rotation of the column at its upper end must be equal 10 the rotation of the left-hand end of the beam. For the column, the slope at x = /is M (4 — oud) PAT ~ “and | oF Ailé-aa) aan ‘The rotation 6 of the leftchand end of the beam is obtained from the slope deflection equation, Thus from which & (1.28) Mt EET ca 7 or 1.29) ‘The negative sign in front of the right-hand side of the equation is needed because the slope as given by (1.27) is negative, whereas @ is positive Equation (1.29) is the stabitity condition for the column being studied. It is convenient to rewrite the equation in the form _ 4k on k= tl ‘The smallest root satisfying this expression is kl = 3.83 from which Pago TEL 0.30) Ifthe upper end of the column were hinged, the critical Toad would be 2*E1/E, and if the upper end were fixed, it would be 20.2E//", tis not surprising that the critical load for the elastically restrained end, given by Eq. (1.30), should fall between the two limiting cases of hinged and completely fixed ends. Completely rigid, hinged or free and conditions are rarely found in actual ante engineering structures. | ing to these idealized e: and Jower bounds betw fail 1.8 EFFECTIVE-LEN! AND DESIGN ¢ Figure 1-9 depiets the ¢ eee a ‘boa eine nae in | flee | Fig. 19 © | critica! | the boundary conditic that the Euler formui provided the correct equally valid for any ‘igure, Thus gives the critical load provided 4 3s the eff equivalent Euler colt For design purpech. tation of the column left-hand end an ined from the slope~ (1.28) (1.29) equation is needed 8 is positive. an being studied. It from which 4.30) ‘id would be x £11, 5 not surprising that ¥ Bq. (1.30), should lely fixed ends, rely found in actual are t8 Etfective-Length Concept and Design Curve 17 engineering structures. However, as demonstrated, the solutions eorrespond- ing to these idealized end conditions ave useful because they provide upper and fower bounds between which the critical loads of most actual columns fall 1.6 EFFECTIVE-LENGTH CONCEPT AND DESIGN CURVE Figure 1-9 depicts the deflection curve and the effective length for several of forme eee Fig. 19 Bective length of for various boundary consitions, | | See | | | the boundary conditions considered in Article 1.5, For each ease, itis shown that the Euler formula ean be used to obtain the critical load of the member, Provided the correct effective length is used. The effective-length concept is equally valid for any other set of boundary conditions not included in the figure. Thus asp tives the critical toad for any column, regardless of the boundary conditions, Provided 4 is the effective length of the member, chat is, the length of the equivalent Euler column. For design purposes, tis convenient to have a graphical representation of18 Buckling oF Columns che Eq. (1.31). Making use of the relation T= Ae in which A is the cross-sectional ares and r the radius of g cam be rewritten in the form RE Ir) For the clastic range of a given material, that is, a fixed value of £, Eq, (1.32) can be used to obtain a curve of critical stress versus slenderness ratio, Air. This has been done for steel ( = 30 x 10° psi), and the resulting curve is shown plotted in Fig. 1-10. The curve in Fig. I-10 is known as a column 8 sooth al eee oe Slenderness rai. Ar Fig. 140 Column curve for structural sec design curve. One of the gteat advantages derived from the effective-length concept is that one can obtain the critical load of any columa, of a given ‘material, regardiess of the boundary conditions, from a single design curve. 1.7 HIGHER-ORDER DIFFERENTIAL EQUATION FOR COLUMNS. In Article 1.5 buckling loads were obtained for columns with various boundary conditions. In each case, a second-order differential equation, valid Ant? only for the member ‘Timoshenko and Gere any column regardiess equation will be intre sets of end restraints. For the general e ‘these exists a moment Fly, 1 ‘choosing approprie boundary conditio internal moment, 2 Differentiating twi and introducing ttayration, Eg. (1.31) (1.32) xed value of E, Eq. asslenderness ratio, | the resulting curve known as a column the effective-tength column, of a given single design curve, mns with various tial equation, valid Ar n7 Higher-Order Differential Equation for Columns 78 only for the member being analyzed, was used. However, as shown by Timoshenko and Gere (Ref. 1.2), a single fourth-order equation applicable to any column regardless of the boundary conditions can also be employed. This equation will be introduced now, and its solution obtained for two specific sets of end restraints For the general case of a columa with unspecified boundary conditions, there exists a moment and « shear at each end of the member (Fig. I-11). By Fig. &11 Column with unspecified boundary conditions choosing appropriate values for these moments and shears, any desired set of boundary conditions can be satisfied. Equating the external moment to the internal moment, st a section a distance x from the origin (Fig, 1-11b) leads to Ely” + Py = ~Qx + My 33), Differentiating twice with respect to x, one obtains Ely" + Py" =0 34) and introducing the notation20 Buckling of Columns hy the equation becomes oN 0 (35) ‘This fourth-order differeatial equation is applicable to any set of boundary conditions. Its general solution is ©, sin kor + C, cos kx yx + Cy 4.36) The arbitrary constants in the sofution are determined from the boundary conditions of the specific case under investigation Case 1. Hinged-hinged For the hinged-hinged column (Fig. {-12) the defiection and bending moment are zero at both ends of the member. Hence Fig. 112 Hinged-hinged column yay 0 ax=0 437) y=O, yO atxal 7 Substitution of the two conditions at x O into Ea, (1.36) leads to Gas Hence From the remaining two conditions one obtains Cysink = C\F= 0, —C,k sink Ar 12 Since kis nat ide C=C, =0 oi trivial solution of ec straight, From the se which for # = 1 lea Case 2. Fisedfree For the fixed-fre Fig, F13. Fised-fee the base. At the fre ‘component Py’ of! ‘The conditions and from the cond(1.33) b any set of boundary a (1.36) sd from the boundary ellection and bendi inged-hinged column 437 1.36) leads to ann? Higher-Order Differential Equation for Columns 21 Since k* is not identically equal to zero, these relations are satisfied either if C, =C, =0 or if sin kJ =C, ~ 0. The first alternative leads to the trivial solution of equilibrium at all loads, provided the member remains straight. From the second conditions, one obtains k= om, 123 which for m = 1 leads to the critical load 2.38) Case 2. Fixed-free For the fixed-free column (Fig. 1-13), the deflection and slope are zero at Fig. 113 Fixed coluran, the base. At the free end, the moment vanishes and the shear is equal to the component Py" of the applied load. Thus y= 0, atx=0 vp key =O, atx ‘The conditions at the fixed end lead to Gas KC, and from the conditions at the free end one obtains ; sin kt + C, cos ki = 0, c=022° Buckling of Columne chy Combining these results gives e C20 = —Cy Creoski~ 0 [Asin the provious example, there are two solutions, the trivial one, obtained by leting Cz =0, and the nontrivial solution, cos ki = 0, The later leads to kf = Ca De from which, for m = 5, (1.39) For both of these cases, the results obtained using the fourth-order equa- tion are the same as those obtained previously by means of the second-order equations. The disadvantage of using second-order equations is that each different set of boundary conditions considered requires the setting up of a new equation, This is obviously not necessary when using the fourth-order equation, since the latter is equally applicable to all boundary conditions. However, there are instances, such as the case of the fixed-fixed column, where the evaluation of the four arbitrary constants needed in the fourth= order equation solution is considerably more involved than the determination of the two constants required in the second-order equation solution 1.8 LARGE-DEFORMATION THEORY FOR COLUMNS. In the preceding articles the behavior of columns was studied using a linear differential equation, The linear equation was obtained by employing an approximate expression, ¥y/dx?, for the curvature of the member. Since this approximation is valid only when the deformations are small, the results ‘obiained from the linear equation are limited to small deflections. fn this article the limitation of small deffections will be removed by using an exact expression for the curvature. Consider the simply supported column shown in Fig. 1-14. Aside from the assumption of small defiections, which is no longer being made, all the other idealizations made for the Euler column are still valid. Thee member is initially assumed to be perfectly straight and loaded aiong its centroidal axis, and the material is assumed to obey Hooke's law. Ie'the x-y coordinate axes are taken as shown in the figure and the column, is in equilibrium in 2 bent configuration, then the external moment, Py, at ‘any section, is equal to the internal resisting moment —EI/R. ‘Thus column (Adapted Sr in which Y/R ist The solution the rate of chang sand the distance ture is Substitution of Equation (1.41, as at small defe is nonlinear, at of Eq. (1.()5t¢ presented lnere Ref. 1.6 Tntroducinent 1=0 rivial one, obtained ). The latter leads to (2.39) > fourth-order equa- of the second-order vations is that each the setting up of a ng the fourth-ord sundary conditions, fixed-fixed column, ded in the fourth- inthe determination on solution, died using a tinear 4 by employing an member, Since this small, the results deflections, In this d by using an exact “14, Aside from the made, all the other 2 member is initially roidal axis, and the are and the columa nal moment, Py, at WR. Thus Art. 18 Lerge-Deformation Theory for Column 23 a Held Large defections of a column (Adapted from Ref. 1.0 ia fe (1.40) in which 1/2 is the curvature, The solution of the equation is facilitated if the curvature is expressed as the rate of change of the slope. Letting the slope at any point be given by 8 and the distance along the curve from the origin to the point by s, the curva~ 1a x-a Substitution ofthis expression in Ea, (140) leads to er + py =o asl) Equation (1.41) can be used to study the behavior of columns at large as well as-at small deformations. Because dB{ds is used in place of ®y/dx#, Eq. (Al) is nonlinear, and its solution is a good deal more difficult to obtain than that of Eq, (1.1), the linear equation of the small-deformation theory. The analysis presented here follows the general outline of the solution given by Wang in Ref. 1.6. Introducing the notation k* = P/EL, Eq. (J.41) takes the form ie Ba uy 0 any24 Bustling of Columns cnt Ifthe equation is differentiated with respect to s, and dvds is replaced by sin 8, one obtains +R sin 8 (1.43) ‘Moltiplying the first term by 2(c8/ds) de and the second term by the equivalent ‘expression 2d and integrating both terms gives [GBB ace [nee sn oe (1.48) see BG) and sin 8 40 = —mogeneous equa- sth the coefficients edto P= mE ‘ore vanish, Hence art. 1.10 Initially Bene Columns 34 where F, = x#E//I%, the expression for C can be rewritten in the form Tay Consequently, (1.66) and (67) ‘The arbitrary constants in (1.67) are evaluated from the boundary conditions. ‘The condition y=0 atx leads to and from the condition one obtains 0 = Asin kl ‘Thus either 4 or sin k/ must vanish. Letting sin kf =0 again limits the solution for y to P = Pp. As before, this is undesirable, and consequently 4 Substitution of A= 0 and B=O into Eq, (1.67) leads to the bending deflection (1.68) The total deflection from the vertical is obtained by adding this expression to the initial deflection, Thus or 1.69)32, Buckling of Columns car Figure 1-17 gives a graphical representation of Eq. (1.70). The variation ° i Fig. 117 Load-deflection curves of Midheigntaelecten 8 initially bet colaros of the midhcight deflection with the load ratio P/P, is shown plotted for two different values of initial imperfection. Based on these curves, the behavior of the imperfect column can be summed up as follows. Unlike the perfect column, which remains straight up to the Euler load, the initially deformed nber begins to bend as soon as the load is applied. The deflection increases slowly at first and then more and more rapidly as the ratio of the applied load to the Euler load increases. The larger the initial imperfection, the larger the total deflection at any foad, However, as the applied foud approaches the Euler load, the deflection increases without bound regardless of the magnitude Of the initial imperfection. The carrying capacity of an imperfect column is thus always smaller than the Euler load, regardless of how small the initial imperfection is. If the initial distortion is sizeable, the column experiences fairly large deformations at loads considerably below the Euler load. carefully constructed column that is fairy straight to begin with does how- ever not deflect appreciably until the applied load is quite close to the Eul Joad, Although the failure mechanism of colurans has not yet been discussed. ‘one can surmise that large deflections produce inelastic strains, which in turn lead to collapse. Members with large initial imperfections can thus be expected to fail at loads considerably below the Euler load, while relatively straight columns will support axial loads only slightly less than Py. 41.47 ECCENTRICALLY LOADED COLUMNS Ins ticle 1.10 an initially bent member was used to investigate the bebavior of an imperfect column. The effects of imperfections on column behavior ean also be studied by considering the straight but eccentrically loaded member shown in Fig. 1-18, Itis assumed that the member is initially straight, that the material obeys Hooke’s jaw, and that deformations remain small. Equating ana Fee £48 Eccenticaly 1 internal resisting applied moment Pt Letting kt = P/EL, ‘The general solutios ‘The arbitrary const the condition leads to and from the condi cone obtainsont (2.70), The variation Load-deflection curves of ‘colar. shown plotted for (wo urves, the behavior of 5. Unlike the perfect the initially deformed te deflection increases tio of the applied load perfection, the larger oad approaches the less of the magnitude 1 imperfect column is how small the initial © column experiences w the Euler load, A begin with does how- site close to the Euler ‘ot yet been discussed, strains, which in tum fections can thus be load, while relatively sss than Ps. vestigate the behavior column behavior can ically loaded member ally steaight, that the ‘main smail. Equating Art tt Eccentricsily Loaded Columns Fig. 118 Recenticlly loaded cole the intemal resisting moment —Efy”, at an applied moment Ple + ») gives Ely" + Plety Letting k= EI, one obtains The general solution of this equation is = Asin kx + Boos kx ~e The arbitrary constants are evaluates the condition yaO weno leads to Bae and from the condition ya0 ates one obtains fm elm cos kt 2 section, to the corresponding amy am (173) mm the boundary conditions. Thus“ Buckling of Columns cht Substitution of these results in Eq, (1.73) gives any ‘An expression for the midheight deflection » = é is obtained by letting x= 2, Thus +550 sn 1) 075 Inada th deities 2sin2 Kt conkt= 1 = Dine Ad Mog and sins = 2sin cost a, (75 can be rest to on (175 whee Pe = mE Figure 1-19 gives « graphical representation of Eq, (1.76). The variation ° ee aa Fig. 119 Load-deflection cures of cecentricelly loaded columns, Misheight detietion & of 5 with F/Py is shown plotted for two values of the eccentricity e, Compati son of these curves with those of initially curved columns, given in Fig, 1-17, indicates that the behavior of an eccentrically loaded column is essentially the ‘same as that of an initially bent column, In both cases, bending begins as soon as the load is applied. The deflection increases slowly at first and then more and more rapidly as P approaches P,, At P = P,, the deflection increases without bound, Coiumns with large eccentricities deflect considerably at loads well below the Euler load, whereas columns with very small eceentric- Are 133 ities of loading do not t Euler load, 1t will later collapse. A column with hoe expected to suppor! folber hand, 2 member siderably below Pr, 4.42 SUMMARY OF BEHAVIOR It has been shown that the load is applied, the Joad approaches the E rapidly. Several impor tions, First and forem: ceancept of @ perfect m imperfect columns, pre load is thus # good apr ceolume can support wi forms the basis of « use toad has been deter system can be in equil ‘at which neutral equil the eritical load can The critical load imperfect system iner thestrctaral emer and then determines ¢ A third conclisiot fof an imperfect syste initial erookedness © forence between & pe former must be diste present in the latter a not surprising that © of which cause bet bochavior of an impe 413 INELASTIC In each of the inves ‘made that the mate tion to be valid, thecnt } a7) sbiained by letting } (1.75) (1.76) 1.16). The variation ad-defection curves of leaded columns, tricity ¢, Compari- given in Fig, 1-17, im is essentially the vding begins as soon first and then more deflection increases ct considerably at very small eccentrie- are 199 Inotastic Buekling of Columns 36 ities of loading do not bend appreciably unti} the load is fairly close to the Euler oud. It will later be shown that excessive bending causes a column to collapse. A cohimn with relatively small eccentricities of loading can therefore be expected to support foads only slightly less than the Euler load. On the other hand, a member with large eccentricities will collapse at loads con: siderably below P, 1.12 SUMMARY OF IMPERFECT COLUMN BEHAVIOR thas been shown that a slightly imperfect column begins to bend 2s soon as the load is applied, that the bending remains relatively insignificant until the load approaches the Euler Load, and that the bending then inereases very rapidly, Several important conclusions can be reached from these obsetva~ tions. First and foremost, the Buler theory, which is based on the Fetitious concept of a perfect member, provides a satisfactory design criterion for real imperfect columns, provided the imperfections are relatively minor, The Euler load is thus a good approximation of the maximum load that a real imperfect column can support without bending excessively. Second, the foregoing results Form the basis of @ useful method of stability analysis. Up to now, the eritical oad has been determined exclusively by finding the load at which a perfect system can be in equilibrium ina slightly bens configuratfon, that is, the load. at which neutral equilibrium is possible. Now, a seeond criterion for finding the critical load can be stated as follows: The critical load is the load at which the deformations of a slightly perfect system increase without bound. To apply this eriterion, one gives the structural member or system to be investigated a small initial deformation and then determines the load at which this deformation becomes unbounded. A third conclusion drawn from the Faregoing analyses is that the behavior bof an imperfect system can be simulated either by giving the system some initial crookedness or by applying the load eccentrically. The essential di ference between a perfect and an impesfect compression member is that the former must be disturbed to produce bending, whereas bending stresses are present in the latter asa direct consequence of the applied load. Its therefore ot surprising that either eccentricity of loading or initial crookedness, both of which cause bending, can be used with equat success to simulate the behavior of an imperfect system, 1.13 INELASTIC BUCKLING OF COLUMNS In each of the investigations presented heretofore, the assumption has been made that the material behaves according to Hooke's law. For this assump- ‘ion to be valid, the stresses in the member must remain below the propor-98 Buckling of Columns cht _Ze Fig. 120. Coluran curve, tional limit of the material, Figure |-20 shows the Euler curve and a horizontal line representing the proportional limit of the material. It is evident from the figure that, for slender columns, the applied load reaches the Baler load before the axial stress exceeds the proportional limit. The finear elastic analysis is therefore valid for slender columns, and the Euler load represents the correct, buckling load of such members. One the other hand, the axial stress in a short column will exceed the proportional limit of the material before the applied (oad reaches the Euier load. Consequently, the results of the elastic analysis are not valid for short columns, and the buckling load of short columns must be determined by taking inelastic behavior into account Before considering the theory of inelastic column behavior, let us briefly review its historical development. The Euler formula was first derived by Leonhard Euler in 1744 (Ref. 1.4). It was, at the time, mistakenly assumed that the formula applied to short as well as slender columns. When test results during the nineteenth eentury indicated that the formula was uncon- servative for short columns, Euler's work was believed to be completely erroneous and was discarded for a iengthy period of time. Finally, in 1889 two ‘men, Considére (Ref. 1.7) and Engesser (Ref. 1.8), reached the conclusion that the Euler Joad as presented by Euler was valid, but only for slender columns. They also realized that Euler's formula could be applied to short columns if the constant modulus £ is replaced by an affective modulus that depends on the magnitude of stress at buckling. Engesser believed the tangent modulus (0 be the correct effective modulus for inelastic column buckling Considére did not reach any specific conclusions regarding the value of the effective modulus. He did, however, suggest that as @ column begins to bend at the critical load there is a possibility that stresses on the concave side Increase in accordance with the tangent modulus and that stresses on the convex side deorease in accordance with Young's modulus. This line of reasoning is the basis of the double modulus theory, according to which the ant 108 effective modulus is modulus, As soon acknowledged its va effective modulus bi it was not until von double modulus the For roughly the modulus theory as for inelastic bucktir inelastie column be the double modul, Gouble modulus the constant as the colt at the critica! load. cause a decrease in the concave side at axial load to incre: bend, and that no cross section. If th tbehavior of al five The tangent 1 double modulus t hhas thersfore {inelastic bueklin, theories continue. ‘The succeedin: ling, Eirst, the et using infinitesima deformations are maximum carry load, as is the ca used to develop ¢ 4.14 DOUBLE In this article th proportional fim modulus theory 1. The eolun 2. Both ends 3. The defor mated by LEBZ cht curveand a horizontal Itis evident from the s the Euler toad before lear elastic analysis is represents the axial stress ina ce material before the fe results of the elastic ickling Toad of short for into account sehavior, let us briefly wvas first derived by 2 mistakenly assumed columns, When test e formula was unean- ved to be completely 3. Finally, in 1889 two vached the conclusion ‘but only for slender Id be applied to short effective modulus that cr believed the tangent sstic column buckling. utding the value of the column begins to bend son the con 6 that stresses on the modulus. This line of ceording to which the e correct Art 1.16 Double Modulus Theory 37 effective modulus is a function of both the tangent modulus and the elastic modulus, As soon as Engesser became aware of Considére's theory, he acknowledged its validity and went on to derive the first correct value of the effective modulus based on the double modulus theory (Ref. 1.9). However, it was not until von Karman in 1910 (Ref. 1.10) independently rederived the double modulus theory that it gained widespread acceptance. For roughly the next 30 years, the double modulus theory, or reduced ‘modulus theory as itis sometimes called, was accepted as the correct theary for inelastic buckling, Then, in 1947, Shanley reexamined the mechanism of inelastic column behavior and concluded that the tangent modulus and not the double modulus is after all the correct effective modulus (Ref. |.11). The double modutus theory is based on the assumption that the axial load remains constant as the column passes from straight to a slightly bent configuration at the critical toad, Only if this assumption is made does bending necessarily cause a decrease in strain on the convex side of the member while strains on the coneave side are increasing. Shanley points out that itis possible for the axial load to increase instead of remaining constant as the column begins to bend, and that no strain reversal need thorefore take place at any point in the cross section. If there is no strain reversal, the tangent modulus governs the behavior of all fibers in the member st buckling, ‘The tangent modulus theory leads to a lower buckling load than the double modulus theory and agrees better with test results than the latter. It has therefore been accepted by most engineers as the correct theory of inelastic buckling, Nevertheless, discussions regarding the merit of each of the theories continue, The succeeding articles are devoted to a detailed study of inetastie buck: ling. First, the critical toad of an initialy perfect member is determined, using infinitesimal deformation theory. Then, initial imperfections and finite deformations are introduced in order to determine whether or not the ‘maximum carrying eapacity of an inelastic column coincides with the critical toad, as is the ease for elastic columns, Finally, the theoretical findings are used to develop design criteria for real engincering materials. 1.14 DOUBLE MODULUS THEORY In this article the critical load of a column whose axial stress exceeds the Proportional limit prior to buckling is obtained by means of the double ‘modulus theory. The analysis involves the following assumptions: 1. ‘The column is initially perfectly straight and concentrically loaded. 2. Both ends of the member are hinged ‘The deformations are small enough for the curvature to be approxi mated by 9”38 Buckling of Columns chr 4. The same relation exists between bending stresses and bending strains as exists between stress and strain in simple tension and compression, 5. Plane sections before bending remain plane afler bending; hence Jongitudinal strains vary linearly as their distance from the neutral surface. ‘The critical load is obtained by means of the concept of neutral equilib- ium. Accordingly, the critical load is defined as the axial load at which equilibrium is possible both in the original undeformed position and in an adjacent slightly bent configuration, This definition implies that the axial load remains constant as the member moves from the straight to the deformed position Consider the column shown in Fig. 1-21a. The final deformed configura: tion of the member is reached by applying an axial force to the initially ee 8 ie if I ee Deformations fee ‘Reduced modus theory (Adapted from Ref. 1.12) aa 98 straight columa, 1c and then letting th stant, In other wor 1-21b, The axial s assumed to be ab During bending, « side of the column member. The fina column is shown stress consists of ‘The deformation ¢ Fig. 1-2ic. ‘The elastic mo a fiber unloads. F column, witich oc crease in strain, (On the concave s instantaneous rel: eis therefore 20" the critical load a very small compe o., can be assum the stress is inere where £, is the s curvature is giver at distances 2, a1 Stresses and str: measured to =p", Bags. (1.79cnt an 148 Double Modulus Theory 39 and bending strains straight column, letting the force increase until it reaches the critical load, on and compression. and then letting the member bend slightly while the axial foree remains con- fter bending; stant. In other words, the behavior of the calumn follows curve 0-1-2 in Fig, fom the neutral 1-21. The axial stress, o,,= P./4, that is present when bending begins is assumed to be abové the proportional limit of the material (Fig, 1-2le), at of neutral equilib During bending, there occurs a small increase in the sttess on the concave axial load at which side of the column and a slight decrease in the stress on the convex side of the i position and in an member, The final stress distribution on a typical eross section of the bent plies that the axial column is shown in Fig. I-21d. At any point in the ross section, the total ‘ght to the deformed stress consists of a uniform axial stress, a.,, and a variable bending stress. ‘The deformation corresponding to this bending stress distribution is shown ia deformed configura- Fig. I-2ie. ‘ree to the initially ‘The elastic modulus E always governs the relation of stress to strain when 4 fiber unloads. Hence the decrease in stress, oy, on the convex side of the colurnn, which occurs as the member bends, is related to the corresponding decrease in strain, ¢,, by the relation a= be, am oui On the concave side, bending causes an increase inthe total stes, and the instantancous relation of the tending stess 2, 10 the corresponding strain 1s therefore governed by the tangent modulus. Sits deformations beyond the eitical load are assumed tobe infinitesimally smal, bending stesoe are very small comptred to ¢,. and the tangent modulus E, corresponding to Gu. can be assumed to apply over the entie part of the cross section weere the sress i increasing. Hence @ a= Be (78) where £, is the slope of the stress-strain curve at o = 9... Noting that the ae curvature is given by ddjds (Fig, [-21e), the bending strains ¢, and e, located at distances 2, and z, from the neutral axis are (1.79) (1.80) a Stresses and strains are positive when compressive and. measured to the right of the neutral axis. Approxi ios —y" Eqs, (1.79) and (1,80) become is positive when ng the curvature by aaa" ast et 1, (1.82)40 Buckling of Columas ons Equilibrium between the external load and the stresses on any section will Row be considered, It has been assuttied that the axial load remains constant during bending. The resultant of the tensile and compressive stresses due to bending must therefore vanish. Thus [rocaas rods ° (1.83) {in view of Eqs. (1.77), (1.78), (1.81), and (1.82), Bq, (1.83) becomes by" [jad ay" [naa 0 a.ga) Introducing the notation ae] 4, O.~ | zsdA 2.85) for the moments of the area on either side of the neutral axis about this axis, Eq, (1.84) can be written in the form FQ, + EQ. = (1.86) Equation (1,86) is used to determine the location of the neutral axis, Since £,%E, O, + Qs #0, and the neutral axis does not coincide with the centroidal axis for inelastic bending, A second condition of equilibrium is that the bending stresses balance the externally applied moment Py, Thus Joon dat fosssaa + by 9° as? ‘Making use of Eqs. (1.77 (1.78), (1.81), and (1.82), one obtains v(E fads 6, f° stad) + By = 1.88) The two integrals inside the parentheses are equal to the moments of inertia of the arcas on either side of these quantities by cutral axis about this axis. Representing dd and (1.89) Eg. (1.88) takes the form (El, + Bh) + Py =0 2.90) ant 148 Finally, by introdi cone obtains Eauaton (192) is Beste it the ie wh Eq, (11) fort That has been rep (190 that he valve material and on the than ‘longa given Independent vib constant coeffi rods 2 The ¢ sss needs the 9 The load given by E loud, Since E, < E, Euler load, For design pury where (¢,)ecis the ¢ To obiain the criti E,. This is accom reduces to where e, and e, ate tension and comp cross section, i8 gi cone obtains#5 on any seetion will nad remains constant sssive stresses due to 83) 83) becomes , (1.84) 1 1.85) axis about this axis, (1.86) xe neutral axis. Since ot coincide with the ing stresses balance 0 (1.87) obtains =o (1.88) Je moments of inertia tis axis. Representing, 4 (1.89) (1.99) a na8 Double Modulus Theory 47 Finally, by introducing the notation El + By asp cone obtains Ely" + Py= 92) Equation (1.92) is the differential equation for the bending of a column stressed into the inelastic range of the material, Comparison of this equation with Eq. (1.2) for the elastic column indicates that the two are identical except that E has been replaced by £,, the reduced modulus. It is evident from Eq. (1.91) that the value of £, depends on the stress-strain characteristics of the material and on the shape of the cross section, and that B, is always smaller than E, Along a given column, £, is constant; that is, itis not a funetion of the independent variable x. Equation (1.92) is therefore a linear equation with constant coelficients, and its solution is identical to that of Eq. (1.1) for the elastic column, except that the elastic modulus E is replaced by the reduced modulus £,. The critical load of an initially straight column whose axial stress exceeds the proportional limit prior to buckling is therefere equal to 1.93) ‘The load given by Eq. (1.93) is commonly referred to as the reduced modulus load. Since £, < E, the reduced modulus load, P,, is always smatler than the Baler load For design purposes, Faq, (1.93) is usually rewritten in the form (de 41.99) Oy where (¢,), is the critical stress corresponding to the reduced modulus Toad. To obtain the critical stress for a given member, it is necessary to evaluate E,, This is accomplished as follows. For @ rectangular section, Eq. (1.86) reduces 10 where c, and c, are the distances from the neutral axis to the extreme fibers in tension and compression, respectively. Noting that hy, the total depth of the cross section, is given by one obtains sea42, Buclling of Columns cht art tas strain reversal anyw VE, by E, at all points in i: Im the problems ¢ Using these relations, the expression for the reduced modulus given by Fa. load at which equil (1.91) can be simplitied to However, in the pre 4e rember bends subse TET TER (1.95) now stable at loads eee ical load is no ot {In a similar manner, the reduced modulus of an idealized 1 seotion, that is to the problem ath ‘ovo flanges of equal area connected by a web of negligible thickness, is found brium can cake plac tobe patigen of the mem : definition will nove t = ee (1.96 Consider a colur “t axial load P equals ition to a slight Since £, depends on the tangent modulus atthe eitical stress, Ea, (1.94) position to a slight connot be used directly to solve for (o,).. Instead, the variation of E. with Ps an ig 1 is obsained from the stress-strain curve of the material and a column bbending moment a design curve, similar to the curve in Fig. J-10, is consteucted. The eritical stress corresponding to any value of Ir can then be obtained directly from the design curve k The tangent modulus theory of inelastic column buckling will now be con- Sider, In detetmining th cities load by this theory, the five essumptions taade inthe double modula thoory sted on page 3. are retained However, one assumption made previously, that the ania Toad remins constant asthe 1 Golumn pases from the straight toa slightly bent postion of equilibrium, ne ; longeeepplis. Instead, the tangent modulus theory assumes thatthe el lr load increases during the transition from the siaight tothe slighty bent Position It also assumes that the inercase in the average axial sess greater than the desrease in stress due to bending a the extteme fiber on the conven Side ofthe member. Hence no srain revert iakts pce on the conven side The comresive stress increases a all points, and the tangent modus gov es the tlation of ses o strain forthe entice eos sation The difference dete the tangent modulus theory and double modules 'cory can be summarized a5 fliows The double modulus theory assumes thatthe axial load remains constant as the column moves from the straight to a slightly bent positon, a the eel load. Hence the compressive sess inereasesacconding to Bon the concive side of the member and decreases according to £ on the convex side. Inthe tangent modulus theory, the axel Joad i assumed to increase during the transition to the bent form, Theres no Straight Yom) 4ulus given by Fg (1.95) ined J section, tha bile thickness, is found 4.96) itieal stress, Eg. (1.94) variation of E, with raterial and a column astructed. The critical sined directly from the dling will now be con- tne five assumptions are retained, However, mains constant as the on of equilibrium, no assumes that the axial W to the slightly bent nial stress is greater ne fiber on the convex .ce on the convex side, tangent modulus gov- tection, y and double modulus odulus theory assumes voves from the straight the compressive stress member and decreases slulus theory, the axial bent form. Th are 138 Tangent Modulus Thoory 43 strain reversal anywhere in the member, and the increase in stress is governed. bay E, at all points in the cross section. {In the problems considered heretofore, the critical load was defined as the load at which equilibrium in the straight configuration ceases 10 be stable. However, in the present case the axial load is assumed to increase as the member bends subsequent to reaching the critical load, Hence the member is. now stable at losds above the critical load, and the old definition of the critical load is no longer valid. A definition of the eritical Joad, more suitable to the problem at hand, is the fowest load at which a bifurcation of equili- Drium can take place, that is, the smallest load at which the deformation pattern of the member suddenly changes. The eritieal load satisfying this definition will now be determined. Consider a column that is initially straight and remains straight until the axial load P equals the critical load. The column then moves from the straight Position to a slightly bent configuration, and the axial load increases from P to P+ AP (Fig. 1-22), It is assumed that AP is large enough, relative to the bending moment at any section, so that the stress at all points in the member Sivecs istbution 2-7 For eat form 2 Stress distribution “ar siolght form | Strain ¢ Stress distbation Fig. 1.22 Tangent modulus theory44 Buckling of Columns cht fnereases as bending takes place. Since deformations beyond the critical load are assumed to be infinitesimally small, the increase in stress Ao that occurs Guring bending is very small compared to the critical stress o., and E, corresponding to ¢,, can be assumed to govern the increase in stress at all points in the meraber. Since the same modulus governs bending deformations at all points in the member, the neutral axis coincides with the eentroidal axis, and bending stresses vary linearly across the section as in elastie behavior. The only dif ference between this case and clastic bending is that inereases in stress are related to increases in strain by £, instead of £, For the bent form shown in Fig. 1-22, the internal bending moment at any section is Mie = Ey ‘Taking into account thet APis negligible compared to P, the external moment {s Py. Hence moment equilibrium at any section leads to Ely" + Py = 0.97) Equation (1.97) is identical to Eg. (1.1), the differential equation of elastic buckling, except that Eis replaced by E,, The critical load obtained from the solution of Eq. (1.97) is therefore (1.98) This expression is generally referred (o as the tangent modulus load, Com- parison between Eq. (1.98) and Eqs. (1.91) and (1.93) indicates that the tangent modulus load is always smaller than the reduced modulus load and that the former, unlike the fatter, is independent of the cross-sectional shape. 1.16 SHANLEY'S THEORY OF INELASTIC. COLUMN BEHAVIOR ‘The analysis presented in this article concerns itself with the behavior of inelastic columns beyond the critical load. To earry out such an investigation, it is necessary to consider finite deformations, As long as an analysis of inelastic columns is limited to infinitesimally small bending deformations the tangent modulus can be assumed to be constant over the entite part of the ‘member where the compressive stress is increasing, This idealization was made m both the reduced modulus and the tangent modulus theory presented previously. However, for finite bending deformations, the actual complex variation of the tengent modulus in the plane of any eross section as well as An 156 along any fngiudin bucking or ite det by using imple modulus cat be ai BeAr show io F aR Fe, connected to each call is made up of t area 4j2 and a ler curve in Fig. 1-23. ‘ments, at the cente plex variation of the cross section &‘ond the critical load stcess Ae that oceues r ay and E, ease in stress at all vas at all points in the al axis, and bending havior. The only aif. nereases in stress are bent form shown in the external moment 0 97) ial equation of elastic oad obtained from the (1.98) modulus load. Com- 93) indicates that the ced modulus foad and + cross-sectional shape. with the behavior of touch an investiga Jong as an analysis of bending deformations. er the entire part of the is idealization was made ‘ulus theory presented 1s, the actual complex oss section as well as Ae 156 Shanley’s Theory of Inetastic Column Behavior 45 inal fiber must be considered. Since it is impossible to express this variation analytically, the differential equation of inelastic column buckling for finite deformations can only be solved by numerical procedures, By using a simple column model, Shanley (Ref. 1.11) has shown that the numerical complexities that arise from considering the variation of the tangent modulus can be avoided and an approximate closed-form solution is obt able. As shown in Fig model consists of two infinitely rigid legs Fig, 123 Shanley model (Adapted from Ref. £11) connected to each other at the center of the column ay a deformable cell. The cell is made up of two axial elements a distance h apart. Each element has an area 4/2 and a length # and behaves 2ecording to the bilinear stress-strain curve in Fig, 1-23, Concentration of all the deformable material, in two ‘ments, at the center of the model obviates the need for considering the com- Plex variation of material properties both along the length and throughout the cross section exhibited by a real column,48 Buckling of Columns cht The model is assumed (o remain straight until the critical load is reached. It is then deflected laterally a finite distance d The resuiting axial strains of the cell elements are e, and c:. These strains are due to bending and any change that occurs in the axial load during bending, They do not include the aaial strain present before bending begins. Ifthe lateral deflection d remains faicly small, itis related to the slope of the legs, a, by 4.99) Since Eq, (1.99) can be rewritten in the form Leet a= Glete) (1.100) Since the modet is in equilibrium in the deflected configuration, the external moment at midheight is equal to the internal moment at that point, ‘The external moment is M,= Pm Bee +0) «109 Corresponding to the strains ¢, and e, there are axial loads P, and P, in the cell elements, Like the strains, these loads are not the total loads, They represent only the change in load that takes place in each element as the column passes from the straight to the bent form. If the effective moduli in the two elements are designated by £, and E,, then P, = 08,4 (1,102) and th ners momenta the cll gual co Ma Poko Mey + (1.203) Eguatingexeral ad nea momen’ ges i: Fe sele tee, hed irom whieh «1.08 eis assumed that ¢ indicated in Fig. 1-23, ‘At this point, itis Joad of the model, T strain begins. Hence £, To study the beh ‘between the applied != assumed that the epi reaching the critcai | and £; is either equal takes place on the cor the notation and substituting Eas From Eq. (1.100) and Substitution of these To express ¢, in form, a second exp!itical load is reached siting axial strains of to bending and any ey do not include the deflection d remains (1.99) (1.100) 1d configuration, the noment at that point, «.t01) ads P, and P, in the he total loads. They each element as the ve effective modulil in (1.102) e) (1.103) (1.104) an 146 Shenley’s Theory of Inelastic Column Bohavier 47 tis assumed that the loads, P, and P,, are taken as positive ifin the direction indicated in Fig. 1-2 At this point, itis useful to obtain an expression for the tangent modulus load of the model. The modulus theory assumes that there is no strain reversal at any point in the crass section at the instant that bending begins. Hence E, = E, = £, and Eq, (1.{04) leads to dh uh (1-105) To study the behavior of the model at finite deformations, a relation between the applied toad and the lateral deflection will now be obtained, [cs assumed that the applied load P increases as the column bends subsequent to reaching the critical load. Hence E=5, 2.106) and Ey is either equal to # or £, depending on whether or not strain reversal takes place on the convex side of the model after bending has started. Using the notation «a.107) and substituting Eqs. (1.106) and (1.107) into Eq. (1.104), one obtains (1.108) From Eq. (1.100) and Substitution ofthese relationships into Eq, (1.108) leads to Ane on BEL + Le ve] or Parlt a gye— de] (1.109) To express e in terms of d, and thus put Eq. (1,109) into a more useful form, a second expression for P will now be obtained, It has already been48 Buckling of Colurine chr assumed that the applied load P increases as the column bends, {Pit is now also assumed thet bending starts at the tangent modulus Joad P,, then Pa Pb AP (1.110) The increase in load AP that occurs during bending is given by (any 113) aia) ally, substitution of this expression into Eq, (1.109) gives the desired load-deflection relation: 5) Using this relation between the applied load P and the lateral deflection d the postbuckling behavior of the model can be studied, As assumed, the model bewins to bend at the tangent modulus load; that is, P= P, when d = 0, As d increases, the variation of P with d depends on the value of r = E,/E,. This ratio is either equal to unity or £/é, depending oon whether or not strain reversal takes place on the conver side of the model Ithere is no strain reversal, + = I, and P remains constant at P, as the mode! bends. However, bending at constant load produces strain reversal. The assumption of na strain reversal thus leads to inconsistent results and must be discarded. Assuming that strain reversal does take place, t = FIE,, and Eq (1.119) indicates that P increases with increasing deflection. It will now be shown that P increases and approaches P,, the veduecd Art 196 modutus Joad, as the « modulus theory assur and Eye, In view of Bq, (1.100 Substitution of this re the reduced modulus thus approaches the becomes very large. ‘The behavior af tt solid curve in Fig. | Fig. 124 Load-defect Shanley model progresses with ine: Jarge, the axial load ‘The most import is that E, decreases ¥ whereas it was assum tion of the postbuck! obtained if the deere somewhere betweenama bends. IF it is now tus load P, then (1.110) ad amy ve shown that (a2 1 ua) ) and (1,113) leads to (ass 1.109) gives the desired aus) ve lateral deflection d, the gent modulus load; that vot P with d depends on unity or E/E,, depending zonvex side of the model onstant at P, as the model sees strain reversal, The istent results and must be place, z= F/E,, and Eq. ection roaches P, e reduced An t16 Shanley's Theary of Inelastic Column Bohavior 49 ‘modulus load, as the deflection d becomes large compared to h. The reduced ‘modulus theory assumes that bending takes place at constant load. Hence P= Py, and Eve = Eyes, from which eB eB, Ee In view of Eq, (1.100) 14fpt) 16) the reduced modulus load, The load supported by the model (Eq. (1.113)] lus approaches the reduced modulus foad (Eq. ({.116)] as the deflection becomes very lar The behavior of the model obtained from Ea, (1.115) solid curve in Fig, 1-24. Bending begins at the ta summarized by the sat modulus load and Fig. 124 Load-defection curve of Shanley model, Detection & progresses with increasing axial load, As the lateral deflection becomes large, the axial load approaches the seduced modulus load. ‘The most important difference between an actual column and the mode! is that F, decreases with increasing compressive strain in an actual column, ‘whereas it was assumed to remain constant in the model, A better approxima tion of the postbuckding curve than that given by the solid line in Fig. 1-24 is obtained if the decrease in Z, is taken into account, As shown by the dashed ‘curve in the figure, the model then has a maximum load whose value lies somewhere between P, and P,Giiaap 50 Buokling of Columne cht Although the model represents sn extreme simplification of the actual column, it does lead to the following valid conclusions regarding inelastic column buckling An initially straight column will begin to bend as soon as the tangent modulus load is exceeded Subsequent to the onset of bending, the axisi load increases and reaches a maximum value that lies somewhere between the tangent ‘modulus load and the reduced modulus load. 3. Although there is no strain reversal at the instant chat bending begins, strain reversal does oceur as soon as the bending deformations are finite. hese thee findings, probably the most significant is that the maximum Joad of the cofumn lies somewhere between the tangent modulus load and the reduced modulus hoad. A more precise study than that conducted by Shanley involving real columns made of real materials indicates that Pp,, is usually closer to P, than to P, and that the fatio P/Pyx = 1.02 to 1.10 (Ref. 1.13), It can thus be concluded that the tangent mociulus load is very close to the maximum load that an inelastic column ean support 147 ECCENTRICALLY LOADED INELASTIC COLUMNS. In this article the behavior of an eccentrically loaded column, stressed beyond the proportional limit, will be considered, As pointed out previously, the relation of stress to strain in a columa"being bent inelastically varies in @ complex manner from point to point in the member. It is therefore not possible to obtain an exact closed-form solution to the problem of inelastic bending of an eccentrically loaded column. Instead, the problem must be solved by means of a numerical procedure. Such a method has been devised by von Kiirmn (Ref. 1.12). However, it involves laborious caleulations, and its use is therefore justified only if a very accurate solution is required, Ja ‘most instances, an approximate solution suffices, and this can be obtained without unduly lengthy calculations if one or more simplifying assumptions sare made, An approximate analysis ofthis type is presented here,” Iti based on two assumptions: 1, The columa axis deflects in a half-sinewave. "Tis method of analyzing an eecenricaly loaded inelastic column ws suggested £0 the auther by Warner Lansing of the Grammar Aiveraft Engineering Corporation. 12. The stress varie curve of the mat idealized 1 secti negligible area, stress distribu In addition to thet sections remaining pl: that the curvatuse cat Consider 4 colum ‘The stresses at any s¢ a bend Bn designated by @, anc indicated by the sol stress variation, sho element are shown strated that the cury in which €, and ¢:2 g, and a, and hist yy”, one obtains It is convenies extreme fiber stress (see Fig. 1-25d). Substitution of thi Making use of thyfication of the actual vns regarding inelastic ss soon as the tangent | load increases and b the tangent « that bending begins, Jing deformations are is that the maximum modules load and the Shanley involving real sually closer to P, than 11), Tecan thus be to the maximum load alumna, stressed beyond ed out previously, the elastically varies in a cr. It is therefore not 2e problem of inelastic the problem must be 104 has been devised vrious calculations, and, olution is required, In 4 this can be obtained implifying assumptions ted here.” It is based umn was suggested tothe ing Corporation Anta? Eccentrically Loaded inelastic Columns 51 2. The stress varies linearly across the section. The actual stress-strain curve of the material is used to obtain the extreme fiber stresses, and a linear variation is then assumed to exist between these stresses. For an idealized I section consisting of two flanges connected by a web of negligible area, the linear approximation coincides with the exact stress distribution In addition to these two approximations, the usual assumptions, of plane sections remaining plane during bending and of deformations being small so that the curvature can be approximated by the second derivative, are made Consider a column that is loaded eccentrically, as shown in Fig, 1-25a. The stresses at any section consist of an average axial stress, oy =» PIA, and a bending stress, ¢,, As indicated in Fig. 1-25b, the extreme fiber stresses are esignated by a, and o,. Between these stresses a linear stress distribution, indicated by the solid line in the figure, is assumed in place of the actual stress variation, showa by a dashed line, The deformations of a typical clement are shown in Fig. Using similar triangles, itis easily demon strated that the curvature [/r is given by Llane F in which €, and e, are the extreme fiber strains corresponding to the stresses 2, and a and his the depth of the cross section. If the curvature is replaced by 9”, one obtains yi =e ay 1t is convenient to designate the slope of the straight ine joining the extreme fiber stresses on the stress-strain curve as the chordal modulus, ox (sce Fig. 1-25d). Hence (1.118) ns) 41.120) | Mh : by52 Buckling of Columns Assumes tnsar S085 estbuion sesso 1) Sees @ {el Detormorion Fig. 125 Eccentrcalyfoaded inelastic column, art 447 Bg. (1.119) ear be ee in which the bending Equation (1.121) assuming a Tinear st member. iis similar constant modulus, 2 modulus, Few ‘Assuming that th in which 6 is the de Substitution of x = and because I= 4 Equation (1-12 average compress! only if Eew corte) "The maximum ‘Suostitution for ¢ae 1a7 Eccentrically Laeded Inelastic Columns 53 Eq. (1.119) can be rewritten in the form a2 in which the bending moment M, at any section, is equal to P(e + 3). Equation (1.121) is the inelastic bending equation that results from assuming a linear stress variation between the extreme fiber stresses of the ‘member, Its similar to the elastic moment-curvature relation except that the constant modulus, E, in the latter has been replaced by a variable chordal modulus, Fey. Assuming that the deformed shape of the column is given by y= dain a.) in whieh 6 is the defiection at midheight, Bq, (1.121) becomes M in = Substitution of x = 2 and Af = P(e + 6) into Eg. (1.123) gives 3) ie and because / ae, an uaa) Ey Wine, | Equation (1.124) gives the midheight defection 5 as a function of the average compression stress a, However, a solution for 5 can be abtained only if Bey corresponding to 8 given vaiue of ¢ is known The maxirmum stress at midheight is given by P, Pe+ oh a W Substitution for54 Buckling of Columns Equation (1.126) allows one to obtain Ecs corresponding to @ given value of Ge, and thus makes possible the evaluation of 6 in Eq, (1.124). Due to the presence of a2, it is not possible to solve Eq. (1,126) directly, instead, the ‘equation must be solved by iteration. One assumes a value for Ecu, solves for 2 in Eq, (1.126), and then using this result one obtains a new estimate of Eos from Eg, (i.118). It is assumed that the cross section of the member is doubly symmetric and that the minimum stress ¢,, in Eq, (1.118), is given by 0, = 0) ~ (6; — oy) 2p The use of Eqs. (1.124) and (1.126) in combination with Bq. (1.118) makes it possible to obtain the load-deflection curve of an eccentrically loaded column. The complex nature of inelastic bending becomes apparent when ‘one considers the amount of numerical work required to obtain 2 solution, in spite of the fact that two very sweeping assumptions have been made, Using the above procedure, the behavior of eceentrically loaded columns will now be examined, Consider a column whose cross section consists of two fianges connected by a web of negiigible area (Fig. 1-26a), Each flange has an area A = 20/n.* and the flanges are 10 in, apart. The slenderness ratio of the column is 30.4 and the member is constructed of an aluminum alloy having the stress-strain characteristics depicted in Fig. 1-26b. Load-deflection curves wilt be obtained for eccentricities of 1.0, 0.5, and 0.1 in, To illustrate the Procedure used fo construct the curves, a set of sample calculations is given, Let it be required to obtain the deflection, 5, corresponding to o, = 38 ksi and e = 1.0 im. The first step in the calculations is the determination of a Set of values for a and Ey that satisfies Eg, (1.126). Substitution of ¢, — 38, ecir? = 0.2, and Ijr = 30.4 into Eq. (1.126) reduces it t0 the form 1 2-86 ees (1.128) ‘As & first approximation, assume o, = 48 ksi and o, = 28 ksi From Fig. 1-26b, ¢; = 0.00473 and €, — 0.00267, Using Eq, (1.118) gives Ecy — 9.7 10? ksi From Eq. (1.128) one obtains ¢ — $0.0 ks Asa second approximation, assume oz From Fig. 1-26b, ¢; = 0.00500 and €, — 0.00247 Using Eq, (1.118) gives ey = 9.5» 10° ksi From Eq. (1.128) one obtains ¢; = $0.2 ksi 50.0ksi and; As a third epproximation, assume o; = $0.2 ksi and ¢, ~ 25.8 kei From Fig. 1-266, ¢, = 0.00502 and €, = 0.00245, Using Eq, (1.118) gives Fey = 9.45 % 107 ies From Eq. (1.128) one obtains ¢, = 50.2 ksi art ta? tress o (kei Fig correct value of Ees Having obtaine: Eq, (1.124). Thus In carrying ov exist nvo solutions there is another st that satisfies Eq. ( used in the precetyg toa given value of (1.124), Due to the directly. Instead, the Iue for Bey, solves for ns a new estimate of jon of the member is a. (1.118), is given by quam ith Eg (1.118) makes { eceentrically loaded comes apparent when o obtain a solution, in ve been made jeally loaded columns scction consiss of to 2), Each flange bas an lenderness ratio ofthe luminum alloy having coad-deflection curves 1 in, Te illustrate the alculations is given. sponding t0 a, = 38 she determination of a ibstitution ofa, to the form (1.128) Ja; = 28 ksi and, = 26.0 ksi, B17 Eccentreally Loaded Inelastic Columns 55 ® ) Fig. 126 Eecentsealy loaded column—exampke Thus o2 = 50,2 ksi and Key ~ 9.45 % 10? ksi satisfy Eg, (1,128), and the correct value of Foy corresponding to a, = 28 ksi is 9.45 x 10° ksi Having obtained Ee, itis naw possible to solve for the deflection 3 using Bq, (1.124). Thus $5 x In carrying out these calculations, it soon becomes apparent that there exist two solutions corresponding to every value of ¢,, Thus for ay = 38 ksi there is another set of values for a and Ey, aside from those just determined, that satisfies Eq. (1.128). By trial-and-error procedure, similar to the one Used in the preceding paragraph, itis found that o, = 59.7 ksi and Bey =58 Buckling of columns ont 5.5 * 10” ksi is the second solution to Eq. (1.128). The deflection corr sponding to Bey = 5.5 x 10° ksi is 1 Om gg = E8Sin 338 Proceediog in 2 similar manner, sufficient data have been obtained to plot the load-deflection curves presented in Fig. 1-27. Curves giving the relation= & Detiection & in) Fig. 1:27. Failure of eccentrictly loaded columns, ship between the average stress oy and the midheight deflection é are given for three different eccentricities of loading. The data used to plot these curves are listed in Table 1-2, For each curve shown in the figure, the load inereases with increasing ax/Ela, and by ) welding it Zir < s./Zia,, However, «large number of tests have shown thes GOlbmns of intermediate slenderness ratio tend to buckle at loads significantly below those siven by the curve in Fig 1-30b, At on time, this discrepancy pas attributed entirely to unavoidable eccentricities of ioading and initial inpertections in the member. However, during the 1950s, a series of extensive investigations at Lehigh University demonstrated conclusively that residual Stresses account for a large portion of the discrepancy between the theoretical Ree ard he test results in Fig. 1-30b, Anvexcelentsurmmary ofthese findings is given in a paper by Beedle and Tall (Ref. 1.14), Residual Stresses in Hot-Rolled Steel Members Hot-olled structurai-stee! members develop residual stresses asa result of never Cooling during the manufacturing process. In wide-fange shapes, the {ips ofthe flanges cool more rapidly than the atea at the intersection of the Web and the flange, where a relatively large mass of material is situates While the fangs tips cool and contract, the flange center is still warm and foft Hence it can follow the deformation of the tips without causing sreeses to be induced in the member. Later, however, wien the flange center oncle 2d ties to shrink, the deformation is resisted by the already cold and rigid fangs tips. Asa result, equilibsium is finally achieved with the fange center ‘n tension and the flange tips in compression (Fig, I-31). The central part of Coots firs canoressive residual stress Fig. 1-91 Uneven coating of hot a" Ym, rolled wide-Sange shape the web clso cools mote rapidly than the web-flange junction, and compres: ‘ive residual stresses consequently also develop in the web. Recause of theis Proximity to the neutral axis, these residual stresses are, however, ‘important than those in the Ranges, The presence of residual stresses in struetural-szel columns has been confirmed experimentally both by the method of sectioning and by obtaining average stress-strain curves from stub column fests of very short columns, In the method of s much less tests, that is, compression ectioning, the member is cut ‘rt. 420—Buchlin into @ large number ¢ present prior to eutt after cutting, the ch present in the origin (Ref. 1.14), it has be compressive residual strength steels at the yield strength. A sim js shown in Fig. 1-3: fang. compressive stress 0 0.3e, at the flange-¥ The presence of otected by means ‘compressed in a test stress, P/A, an aver shown in Fig. 1-33, strain relationship o is free of residual sit reaches the yield sto at constant stress. F column, causing tho the applied stress 1 spreads progressive! compression to thos tensile stress. The st yielding curve betw the presence of resiae Art 130 Buckling Strength of Hot-Rolled WideFlange Steel Columns 62 eee into a large number ofiongitudinal strips, which relieves the residuat stresses ee present prior to cutting, Ifthe length of each strip is measured before and 0 after cutting, the changes in length are indicative of the residual stresses ee ee present in the original member. Based on large numberof investigations (Ref, 114), it has been concluded that the av at loads significantly rage value of the maximum : compressive residual ses in hobroced wide Range shapes of mover Wading and al strength steels at the ip ofthe Range is approximately equal tO of he acces orexiansite yield strength, A simple idealization of the residual stress pattern in a flange is shown in Fig, 1-32. It assumes a linear stress variation from a maximum usively that residual tween the theoretical ary of these findings stresses as a result of deflange shapes, the intersection of the | material is situated i er is still warm and. out eausing stresses e flange center cools ready cold and rigid {th the flange center +03 oy 4 sani The central part of ig. HAE Residual stress pattern in 1533. Stub column stess-strain ange curve (Adapted from Ref. 1.14), compressive stress of 0.30, at the flange tip 10 a maximum tensile stress of 0.30, at the flange-web juncture. The presence of residual stresses in a hot-rolled member can also be detected by means of a stub-column test, Ifa short length of the member is compressed in a testing machine and the strain is plotted against the average stress, P/A, an average stress-strain curve for the entire cross section, as shown in Fig. 1-33, is obtained. The dashed curve, which gives the stress strain relationship of a coupon, is included for comparison, Since the coupon is free of residual stresses, al its fibers remain elastic until the applied stress reaches the yield strength of the material. Subsequently, each deforms freely even cooling of hot we shape. Gaeta eae at constant stress, By comparison, residual stresses are present in the stub a eee ariel column, causing those fibers with an initial compressive stress to yield before eoreier auch eae the applied stress reaches the yield strength of the material. Yielding then spreads progressively, as the load is increased, from the fibers initially in jieaieertien tae compression to those with no initial stress, and finally to those with an initial nd by obtaining tensile stress. The stub-column stress-strain relation thus exhibits a gradual that is, compression yielding curve between the proportional limit and the yield strength, due to ieee wae the presence of residual stresses.64 buckling of Columns cht By depicting the internal stress distribution at various load levels, Beedle and Tall (Ref. 1.14) demonstrate precisely how residual stresses affect the behavior of a compression member. In their illustration they make use of an J section with a linear residual stress distribution, as shown in Fig. (-34a, Stub column a ah ar 7 ooroge stress P/A al ‘Shin e ig. 134 Inftence of residual stress on stub column strese-steain curve (Adapted from Ref. 14), Prior to the application of an external load, only the residual stresses are present (Fig. 1-340), With an external load acting, the stresses consist of the sum of the original residual stresses and the additional stress required to balance the applied load (Fig. 1-34c). Yielding commences, atthe flange tips, when the average applied stress P/4 plus the residual stress ¢, at these points is equal to the yield strength @, of the material (Fig, 1-34d). Thus ¢, proportional limit of the stub-coluran stress-strain curve, is Op =0,~ or For the ease being considered, ¢, O.7ay. As the applied stress, P/, Aan. 120 Buckling increases beyond 0.70, the flange center (Fig. to yield (Fig. 1-340). 7 above the proportions load-resisting area. B pradually yielding str gradually yielding stre Goiumn Strength Cur Structural-Stee! Men ‘The presence of impossible to determ! theory, However, the column buckling dos bending increase sin no strain reversal oct the absence of strain member. Th impor Johnston (Ref. 1-15), the inelastic range is the entire cross sectic part of the cross sect Equation (1.131) ind the Buler stress mult The ratio 1,// der the member, and or idealized J section si resistance of the we flanges about their « partially yielded (sh the xx axis, 1/78 in whieh 6 and 4vad levels, Beedle sresses affect the vy make use of an wn in Fig. residual stresses are esses consist of the 1 stress required to 5, at the flange tips, 28s, at these points, 134d). Thus 6, the ve, 8 applied stress, PA. Art,1.20 Buckling Strength of Hot-Rolled Wide-Flange Steel Columns 65 -yond 0.7, yielding spreads inward from the flange tips toward the flange center (Fig. I-i4e), At P/4 ~ ¢,, the entire cross seetion has started to yield (Fig. 1-34f), The decrease in stiffness exhibited by the stub column above the proportionat limit is thus due (o @ gradual reduction of the effective load-resisting area, By comparison, an aluminum stub column exhibits a gradually yielding stress-strain relation because each of its bers has a gradually yielding stress-strain curve Column Strenath Curves for Hot-Rolled Structural-Steel Members The presence of residual stresses in structural-steel columns makes it impossible to determine the critical load directly from the tangent modulus theory. However, the basic concept underlying Shanley’s theory of inelastic column buckling does apply. According to this concept, axial load and bending increase simultaneously after the critical foad has been reached, and no strain reversal occurs as the member begins to buckle, For a steel column, the absence of strain reversal at buckling means that only the elastic part of the cross section contributes to the internal resisting moment of the deformed member. This important observation was first made by Yang, Beedle, and Johnston (Ref. 1.15). Based on it, one can conclude that the eritieal toad in the inelastic range is given by the Euler load with the moment of inertia of the entire eross section, /, replaced by J,, the moment of inertia of the elastic part of the cross section. Thus or a3 Equation (1.131) indicates that the critical stress of a steel column is equal to the Euler stress multiplied by the reduction factor 1, ‘The ratio 1/7 depends on the residual stress distribution, on the shape of ye member, and on the axis about which the columa bends, Consider the Idealized 1 section shown in Fig. 1-35, In addition to neglecting the bending resistance of the web, itis justified Co neglect the moment of inertia of the flanges about their own centroidal axis for such sections. If the flanges are partial yielded (shaded area), and the section bends about its strong axis, the sex axis, 1/7 is given by in which 6 and 4 are the width and area of the flange, and b, and A, the6 Buckling of Cotumans cnt Elastic port vy Fig. £35. Partially yielded section clastic parts of these quantities. 1f, on the other hand, bending takes place about the weak axis, the yy axis, JVs given by 2ou/ia Soria 4 (13) Letting ¢ = A,/A. Eq. (1.131) can be written in the form (134) for sirongaaxis bending, and (1.135) for weak-axis bending. Comparison of Egs, (1.134) and (1.135) indicates that the reduction factor is much larger for weak-axis bending than for strong- axis bending and that each case will therefore have a different column- strength curve, To plot the column curves corresponding to Eqs, (1.134) and (1.135), iti necessary to know how varies with the average axial stress, P/4. This rela- tion can be determined either experimentally, by obtaining a stub-columm stress-strain curve, or analytically if the residual stress distribution is known ‘The stub-column stress-strain curve is plot of strain versus average stress. Is slope is given by do, We (0.136) ar 120 Buckling S from whict, ‘The factor ¢ is thus eg column stress-strain cu ‘Alternatively, t ca seetion and the linear 1 Fg the average applied stub column, the or shown in Fig, 1-365 ‘The total load P act in which a, is the st figure, it ean be she Substitution of thitParilly yielded section, ending takes place (1.133) (1.13 a. (1.135) indicates that vding than for stron; a different column- 134) and (1.138), itis stress, P/4. This rela ‘ining 2 stub-column distribution is known. | versus average stress. 136), Art,1.20 Buckling Strength of Hot-Rolled Wide-Flange Stoo! Golumas 67 From which 5 (1.137) The factor 7 is thus equal to the ratio’of the tangent modulus of the stub- column stress-strain curve to the elastic modulus. Alternatively, ¢ can be determined analytically using the idealized 1 section and the linear residual sisess distribution shown in Fig. 1-362, When ——— Fig. 1-36 Residual stress in idealized T section, the average applied stress #, = PIA is above the proportional limit of the stub coluinn, the cross section is partially plastic and partially elastic, as shown in Fig. 1-360, The dashed lines represent the initial residual stresses, The total load P acting on the member is given by P= afd A) + 2 4, (438) in which oy is the stress at the center of the flange. From the geometry of the figure, it ean be shown that Substitution of this relation into Bq. (1.138) feads to Ab a68 Buckling of Columns one Dividing through by the erosssectional area, 24, gives esa a, 0, ~ 04 and replacing 4,/A with one obtains t= Mayle, (1.139) Equation (1.138) gives the desired variation of z with the average axial stress ¢,. Using this relation in Eqs. (1.134) and (1.135), two columa-strength curves, one for strongeaxis bending and one for weak-axis bending, have been obtained. The curves are shown in Fig, 1-37 as solid lines. The upper curve is Sieong oxle benaing Oe)/2 Fig. 1-37 Column curves for wide-tange shapes wit (Adapted from Ref. 48. for strong-axis bending and the lower curve for weak-axis bending, These curves ate for an idealized { section with the linear residual stress distribution depicted in Fig. J-36a. A similar set of curves ean be developed for any other cross-sectional shaps and residual stress distribution, An extensive investig UHon of residual stress patterns in wide-flange shapes (Ref. 1.14) has indicated that the actual stress distribution falls somewhere between the linear one assumed here and a parabolic one with the same maximum residual com= pression stress; that is, ¢yiman) = 0.30, The curves in Fig. 1-37 therefore give @ reasonable estimate of the strength of hot-rolled wideslange structure! steel shapes. art 120 The dashed fine in strong- and weak-axis E and was first proposec evaluated from the con’ intersect the Euler eury lead to Although the maximur steel members is appr and strongeaxis bendin Equation (1.140) corre 1.21 DESIGN OF ¢ ‘The Euler equation buckle elastically, itis columns on the Euler tions, the allowable 5 where 2 is the effect allowable stress is tak 23/12. The safety fac initial eocentricities « in the Euler theory in most stee\ membe1 theory does not give During the first hal0.139) average axial column-strength nding, have been upper curve is daal stresses is bending. The: stress distribution vped for any other tensive investiga 14) has indicated cen the linear one tom residual com= -37 therefore give inge structural- Art 1.28 Design of Steet Columns 69 ‘The dashed line in the figure, which provides a compromise between strong- and weak-axis bending, was recommended by the Column Research CCouneil in the second edition of it 16), It is a parabola of the Form and was first proposed by Bleich (Ref. 1.12). The constants 4 and B are evaluated from the conditions that ¢.- = 0, at AJr = 0 and that the parabola intersect the Euler curve at the proportional limit ¢, ~ ¢,. These conditions lead to — le 02, Although the maximum compressive residual stress in the lange of structura steel members is approximately 0.3a,, a curve better suited to both weak- and strong-axis bending is obtained by letting ¢, ~ 0.50, Thus (ato), Equation (1.140) corresponds to the dashed curve in Fig. | 1.21 DESIGN OF STEEL COLUMNS ‘The Euler equation accurately predicts the strength of steel columns that buckle elastically. It is therefore customary to base the design of slender steel columns on the Euler Joad. For example, in the 1969 AISC design specifica tions, the allowable stress in compression for elastie buckling is given by 4b where 2 is the effective length of the member (Ref. 1.17). In this formula the allowable stress is taken equal to the Euler stress divided by a safety factor of 23/12. The safety factor accounts for reductions in steength resulting from initial eccentricities and other deviations from the ideal conditions assumed in the Euler theory ‘Theoretically, the inelastic buckling toad ofa steel column should be given by the tangent modulus load, However, due to the presence of residual stresses in most steel members, a straightforward application of the tangent modulus theory does not give a satisfactory estimate of the strength of short columns. During the first half of the twentieth century the significance of residual70° Buckling of Columne stresses, although suspected, was not sufficiently appreciated, As a conse- quence, no rational theory for inelastic buckling was developed. Instead, design formulas were simply obtained by fitting a curve to experimentally obtained buckling loads. Two of the most commonly used relationships were the straight-line formula op 4 BID au = Agel and the Johnson parabo on = AUN In both of these expressions 4 and B are empirically determined constants and S.F. is a safety factor. An example of the latter formula ean be found in the 1949 AISC design specifications (Ref. 1.18), where the allowable stress for inelastic buckling is given by 17,000 — o.4a5(L)" (1.142) The first theoretically based formula for short steel columns was not developed until after an extensive investigation at Lehigh University in the 1950s proved conclusively that residual stresses are responsible for the dif- ference in strength between the results of the tangent modulus theory and actual test observations. This study, which is summarized in Article 1.20, Jed to the CRC column-strength curve given by Eq, (1.140). Shortly thereafter the AISC adopted the CRC curve as its design criterion, and the allowable stress in the 1969 specifications (Ref. 1.17) is accordingly given by (1.143) To account for the fact that the sensitivity of « column to eccentricities and variations in the support conditions increases with A/r, the safety factor in Eq, (1.143) is a function of Hr. The early research on the strength of short steel columns was largely confined to hot-rolled wideslange shapes with a yield stress of 33 or 36 ksi However, since that time numerous additional investigations have been conducted involving @ variety of different shapes, steel grades, and fabrica- tion procedures (Refs. 1,19 and 1.20), It is now evident from the results of these studies that the strengths of different types of steel columns vary cone siderably and that more than a single design curve may therefore be desirable, To this end, the Column Research Couneil in the third edition ofits guide is recommending the use of three column-strength curves in place of the for mer one, each of these © category of columns. I these curves are hot-roi box shapes, and round In addition to reve: recent studies have also ‘of initial imperfections i ‘Thus the new column columns, whereas Eq. column, Am initial out < strength curves. References LLM, Satvapont anc Nil Prentice-Hal 1.2. SP. Twosmenxo York: MeGraw-1 13 NJ. Horr, The 1956) LL, Buuee, “Suria LSN. J. Hore, “Bu Society, Vol. 58,3 1.7 A. Cossobae, “R Procédés de Cons 18 F, ENoEsser, 19. F, Enorssen, “Kr 140 T. vox KAnMans Forsehungsarbeite LAB. R, Stantey Selenees, Vol. 4 Lag F. Busca, Bucki Book Company. Lia J.B, Donsro ar 114 LS, Beeoue and Division, ASCE,cht ated, As a conse: zveloped. Instead, lo experimentally relationships were ermined constants a can be found in se allowable stress (0.142 columns was not a University in the ‘onsible for the dif vodulus theory and in Article 1.20, led Shortly theresfter and the allowable given by (1.143) o eccentricities and the safety factor in slumns was largely ‘ess of 33 oF 36 ksi ns have been rades, und fabrica- from the results of columns vary con- erefore be desirable sedition ofits guide sin place of the for- cht Poferences Ti mer one, exch of these curves being representative of the strength of a related category of columns. Included in the three groups of columns covered by these curves are hot-rolled and cold-straightened members, wide-flange and box shapes, and round bers and members composed of welded plates In addition to revealing the need for multiple column-strength curves, recent studies have also demonstrated the desirability of including the effect of initial imperfections in the theory instead of in the safety factor (Ref. 1.21), ‘Thus the new column curves give the maximum strengths of initially bent columns, whereas Eq, (1.140) gives the critical load of an initially straight column. An initial out of straightness of 1000 is used int strength curves, References 1.1. M, SaLvapori and R, HeLten, Structure in Architecture (Englewood Clits, Nu: Prentice-Hall, Ine, 1963), 12S, P. Tiwostenko and J. M. Gre, Theory of Elastic Stability, 2 York: McGraw-Hill Book Company, 1961). ad ed. (New 1.3 N.J. Hore, The Analisis of Structures (New York: John Wiley & Sons, Ine. 1956) 1.4L. Bune, “Sura force de colonnes," Mémoires de’ Académie ce Berlin, 1759, ESN, J, Horr, “Buckling and Stability,” Journal of che Royal Aeronautical Society, Vol. $8, Jan, 1954 16 CT. Wao, Applied Elasticity (New York: McGraw-Hill Book Company, 1953) 1.7. Consipine, “Résistance des pices comprimées,” Cangris Snternarional des Procédés de Construction, Paris, Vol. 3, 1891 18 F, Enaussex, “Usher die Knickfastigkeit gorader Stibe.” Zeitschrift fie Archielur und lngeniewrwesen, Vol. 35, 1889, 19 F. Enoessen, “Knicktragen,” Sehweitzerische Bauzeitung, Vol. 26, 1895 1.10 T. von Kaan, “Untersuchungen ier knicktestigkei Forschumgsarbeiton auf dem * Mitteilungen aber blete des Ingenicurwesens, Berlin, No. §1, 1910 LiL BR. SuaNtey, “Inelastic Column Theory” Journal of the Aeronautical Sciences, Vol. 14, No, 5, 1947 112 F, Bustest, Buckling Strength of Metal Structures (New York: McGraw-Hill Book Company, 1952), 13 J.B, Dusero and T. W. Wiener, “Column Behavior in the Plastic Steess Range,” Journal of the Aeronautical Sclences, Vol. 17, No, 6, 1950, 14 1.5. Beepoe and L, Tane, “Basie Column Strength,” Journal of the Structural Division, ASCE, Vol, 86, No, ST7, 1960, HKLis 16 Lae 119 ut 12 Buclling of Columns cnr CH. Yano, L, 8, Besvts, and B. G. JOHNSTON, “Residual Stress and the Yield Strength of Steel Beams,” Welding Journal, Vol. 31, No. 4 1952, B. G. Jouwsrow, Guide 10 Design Criteria for Metal Compression Members, 2nd ed: (New York: John Wiley & Sons, Ine. 1966). AISC, Steel Construction Manual, Th ed. (New York: American Institute of Steel Construction, 1970), AISC, Steel Construction Manual, Sth ed. (New York: American Tastituie of Steel Consiruction, 1947), R. Brogwovoe and £, Taut, "Maximum Column Strength and the Multiple Column Curve Concept,” Friis Laboratory Report No. 337.29, Lehigh University, Oct. 1971 B. G. Jounstow, Guide 0 Desion Criteria for Meral Compression Members, Srd ed. (New York: John Wiley & Sons, Inc. 1973). R. HE. BATTERMAN and B. G. Jonxstox, “Behavior and Maximum Strength of Metal Columns,” Journal af the Structural Division, ASCE, Vol. 93, No. S72, 1967, Problems Determine the magnitude of the eritical load for the axially louded column shown in Fig. Pl-1, The column is hinged at both ends and prevented from Fig. Pt moving in the direction at its midpoint. The column is, however, free to move inthe y direction at midspan. Assume thatthe column buckles elastically Write the second-order differential equation for the bending of the column shown in Fig. PI-2 +f 4 7 Fig. P Avits lower end the prevented from rot 4.3. Find the crtial toe in Fig. PI-3. Them land to the support 1M is the moment ¢ connects the two 8 La Using the mode! it Iaeral debection w (@) the lateral defc () the foad is 200 (©) the model has Which fundamente these models? 18 Determine the ith (eines Werte sepat cof continuity at th 1.6 Using the tabulatecht a1 Stress and the ye 1952. ression Members, rican Insitute of ercan Institute of + and the Multiple 1. 397.29, Lehigh pression Members sfaximur Strength SCE, Vel. 93, No. ally Yoaded column and prevented from however, fee to move dackles elastically sending of the column cnt Probleme 73 shown in Fig. P1-2 and use it to determine the critical load of the column, Fig, PL2 Fig, PIs Atits lower end the column is completely fixed. At the upper end the columa is, prevented from rotating, but free to translate laterally. (Py, = m*ETIL*) 1.3. Find the critical toad of the one-degree-of-freedom model of a column shown in Fig, PI-3. The model consists of tWo rigid bars pin connected to each other and to the supports. A linear rotational spring of stiffaess C = 1/8, wisere IM isthe moment at the spring and @ is the angle between the wo bars, also connects the two bars to each other. (Py, = 2C/L) 1.4 Using the mocel in Fig. Pi-3, obtain and plot relationships for load versus laterat deftetion when (@) the lateral deflections are large, () the load is applied eccentrcaliy, (© the modei has an initial lateral defection Which fundamental characteristis of an actual columa are demonstrated by these models? 15. Determine the critical ioad of the column on three supports shown in Fig, Pl-4, Fig, Plot (Hine: Write separate equations for each span and make use of the conditions ‘continuity at the center support.) 1.6 Using the tabulated stainless-steel stressstrain data, construct a column design74 Buckling of Columns ont curve, that is, 2 curve of tangent-modulus stress versus slenderness ratio, otis) enfin) est) tinny 100 0.00004 22s o.00018 200 0.00008 as0 3.00025 20 0.00010 srs 0.00050, 2s 2.000115 30 0.00080 500 o.00034 1.7 Using the procedure outlined in Article 1.17 and the stress-strain data of Problem 1.6, determine the maximom load that the eccentrically loaded colunin with the idealized 1 section shown in Fig, Pl-S can support, (Asseme the saine stress-strain properties for tension and compression.) 2.4 INTRODUCTIC Fig. PLS In Chapter 1, the bel lating the governing However, in many in to obtain, and appr approximate method the method of finite two involve a consi especially suitable fo able. Tn one way or at replaces the actual cc ‘The behavior of a ¢ freedom is described the behavior of a fini algebraic equations. algebraic equations, equations, whose 80