Lecture 10: Advanced Topic in Column Buckling

Structural Mechanics 2.080 Lecture 10 Semester Yr Lecture 10: Advanced Topic in Column Buckling 10.1 The Tallest Column In 1757 the Swiss mathematician Leonard Euler presented the famous solution for buckling of a pin-pin column under compressive loading at its end. He also formulated and solved the much more difficult problem of a clamped-free column loaded by its own weight. The practical question was how tall the prismatic column could be before it buckles under its own weight. In order to formulate this problem, the equation of equilibrium of a beam/column in the axial direction must be re-visited. Instead the equation N 0 = 0 or N = const, we must assume that there is a body force q per unit length q = Aρ, where A is the cross-sectional area of the column and ρ is its mass density. Then, the equilibrium in the axial direction requires that N 0 = q or N = qx + C (10.1) In the coordinate system shown in Fig. (10.1), the axial force must be zero at x = l. P l q x Figure 10.1: Column loaded at its tip (left) and loaded by its own weight (right). The distribution of axial force along the length of the column is N (x) = −q(l − x) (10.2) where the minus sign indicates that N is the compressive force. As before, the input parameters of the problem are E, I and q and the unknown is the critical length lc . The derivation of the buckling problem for a classical column presented in Lecture 9 is still valid but the axial force in Eq. (9.20) is no longer constant and thus should be kept inside the integral. For the present problem the first variation of the total potential energy is δΠ = − Z l 00 Z M δw dx + 0 0 10-1 l q(l − x)w0 δw0 dx (10.3) Structural Mechanics 2.080 Lecture 10 Semester Yr Integrating the right hand side of Eq. (10.3) by part, one gets Z l  00  M + q(l − x)w0 δw dx + Boundary terms = 0 (10.4) 0 where Boundary terms = −M δw0 l 0 + M 0 δw l 0 + q(l − x)w0 δw (10.5) at x = 0, δw = δw0 = 0; and at x = l, M = 0, V = M 0 = 0 and l − x = 0. Therefore the boundary terms vanish (see the dedication in Section 3.5). Using the elasticity law, M = −EIw00 , the local equilibrium equation for the column self buckling becomes   d4 w d dw EI 4 + q(l − x) =0 (10.6) dx dx dx Integrating once, we get d3 w + q(l − x)dw = 0 (10.7) dx3 The integration constant is zero because the shear force vanishes at the free end x = l. The governing equation is the third order linear differential equation with a variable coefficient. The solution is no longer represented by the harmonic function. The way to solve the problem is to introduce two new variables r 2 q(l − x)3 dw ξ= , u= (10.8) 3 EI dξ EI Then, Eq. (10.7) transforms to the Bessel equation d2 u 1 du 1 + + (1 − 2 )u = 0 dξ 2 ξ dξ 9ξ (10.9) Omitting the details of the calculation, the critical length of the column is found to be lc3 = 9EI 2 j1 4q 3 (10.10) where j 1 = 1.866 is the root of the Bessel function of the third kind. Finally 3 lc3 = 7.837 EI q (10.11) The total weight of the column material is Nc = lc q. In terms of the total weight, the critical length is EI lc2 = 7.84 (10.12) Nc For comparison, the length of the free-clamped column at buckling loaded by the same weight is π 2 EI EI lc2 = = 2.47 (10.13) 4 Nc Nc The bottom of both column rsees the same weight, but the critical length of the column 7.84 undergoing self-buckling is = 1.78 times taller than a similar cross-section column 2.47 loaded at its tip. 10-2 Structural Mechanics 2.080 Lecture 10 Semester Yr Example A steel tubular mast solidly built-in the foundation and is free on its top. The cylinder is t = 3 mm thick and has a radius of R = 50 mm. What is the critical length of the mast to buckle under its own weight? The total weight of the mast is Nc = Alρ (10.14) where A is the cross-sectional area, A = 2πRt. The second moment of inertia of the thinwalled tube is I = πR3 t. From Eq. (10.12) lc2 = 7.84 EπR3 t 2πRtlc ρ (10.15) from which one gets s lc = 3 3.92ER2 = 65 m ρ (10.16) The above solution applies to a prismatic column of a constant cross-section. Approximate solution can be derived from the Trefftz condition δ 2 Π = 0. Starting from Eq. (10.3) and performing the second variation one gets Z EI l l Z δw00 δw0 dx + 0 0 q(l − x)δw0 δw dx (10.17) The critical compressive body force is then Z l φ00 φ00 dx 0 q = EI Z l 0 (10.18) 0 0 (l − x)φ φ dx As compared with the standard Trefftz formula for tip loaded column, there is the term (l − x) in the denominator. As an example consider the simplest parabolic deflection shape φ = x2 (10.19a) φ0 = 2x (10.19b) 00 φ =2 (10.19c) Introducing the above expression into Eq. (10.18), the critical buckling weight per unit length is 12EI q= 3 (10.20) l 10-3 Structural Mechanics 2.080 Lecture 10 Semester Yr 12 − 7.837 The error in this approximation is = 53% which is not good. As a second trial 7.837 consider a power shape function with a fractional exponent α φ = xα 0 (10.21a) α−1 φ = αx (10.21b) φ00 = α(α − 1)xα−2 (10.21c) The resulting solution is 2EI α(α − 1)(2α − 1) (10.22) l3 2α − 3 The critical buckling parameter attains a minimums at α = 1.75. The minimum buckling load is EI qmin = 9.8 3 (10.23) l The error is slashed by half but it is still large at 25%. In the third attempt, let’s consider the trigonometric function q= πx φ = 1 − cos  π  2lπx 0 φ = sin 2l 2l  π 2 πx φ00 = cos 2l 2l (10.24a) (10.24b) (10.24c) In addition to satisfying clamped kinematic condition at x = 0, the cosine shape gives the zero bending moment at the top. Substituting Eqs. (10.24) into the Trefftz condition, Eq. (10.18), the following closed-form solution is obtained q= π4 EI EI = 8.29 3 3 2 l 2(π − 4) l (10.25) which differs by only 6% from the exact solution. The true shape of the column which buckles by its own weight is the Bessel function but the trigonometric function provides a very good approximation. For over 200 years the Euler solution of buckling of a column under its own weight remains unchallenged. In 1960 Keller and Niordson asked the question by how much can the height of the column be increased. If the same volume of material is distributed as a constant cross-section prismatic structure of the radius r = 0.1 m, the length of the column would be V 1 l= 2 = = 32 m πr π0.12 and the weight per unit length of a still column will be q= V 7.8 × 104 = = 24 N/m l 32 10-4 Structural Mechanics 2.080 Lecture 10 Semester Yr Using Eq. (10.12) we can check if such a column will stay or buckle under its own weight lc2 = 7.84 EI N πr4 where I = , N = V ρ and E = 2.1 × 1011 N/m2 . Substituting the above values, the 4 critical length becomes lc = 26 m. This means that the 32 m prismatic column will buckle and cannot be erected. By shaping the column according to Fig. (10.2) its length can be increase by a factor of 86/26 = 3.3. If the cross-section is variable, this question has led to a very complex mathematical problem. Some aspects of this solution are still studied up to now. The problem is wellposed if the optimal solution is sought under a constant, given volume of the material. There is no simple closed-form solution to the problem so the answer is obtained through numerical optimization, see Fig. (10.2). 100 80 60 z 40 20 0 0.5 0.5 0 y 0 -0.5 -0.5 x Figure 10.2: The shape of the tallest column. Note that the height of the column was scaled down to fit on the page. To give you an idea, the steel column of the total volume of 1.0 m3 and the bare radius of 10 cm could be as high as l = 86 m. 10.2 Deflection Behavior for Beam with Compressive Axial Loads and Transverse Loads Consider a simply supported beam with a fixed load f applied at the middle as shown in Fig. (10.3). Additionally, the beam is subjected to a compressive axial load P . The total 10-5 Structural Mechanics 2.080 Lecture 10 Semester Yr EI P f L/2 L/2 x Figure 10.3: Simply supported beam with intermediate transverse load. potential energy for this mechanical system is   Z L Z L 1 1 0 2 L 00 2 Πtotal = EI(v ) dx − P (v ) dx − f v 2 0 2 0 2 (10.26) If f = 0, we are looking at a classical buckling problem; viz., the beam remains straight until a critical load is reached after which the beam bends suddenly. The critical load for the configuration shown is Pcr = π 2 EI/L2 . Let’s investigate the behavior for f 6= 0. The stationary points of the potential energy still give the solutions v(x) which satisfy equilibrium. Let’s compute an approximate solution using the form  x v(x) ≈ C sin π (10.27) L The derivatives of this function are  x π v 0 (x) = C cos π L L  π 2  x v 00 (x) = −C sin π L L Inserting these into the potential energy yields Z L  π 4  x 1 Πtotal = EI C 2 sin2 π dx L L 0 2   Z L    x 1 π 2 2 L/2 −P C cos2 π dx − f C sin π L L 0 2 L    Z L   1 π 4 2 1 1 2πx = EI C − cos dx L 2 2 L 0 2    Z L   1 π 2 2 1 1 2πx −P C + cos dx − f C 2 2 L 0 2 L  π 4 1  π 2 2 1 = EI C 2L − P C L − fC 4 L 4 L The stationary condition yields  π 4 dΠtotal 1 1  π 2 0= = EI CL − P CL − f dC 2 L 2 L  (10.28)  π 4 1 1  π 2 = C EI L−P L −f =0 2 L 2 L 10-6 Structural Mechanics 2.080 Lecture 10 Semester Yr and thus C= f EIπ 4 −P 2L3 f 2L/π 2 = EIπ 2 −P L2 f 2L = 2 π Pcr − P π2 2L (10.29) The approximate solution has the form v(x) ≈  x 2L f sin π π 2 Pcr − P L (10.30) l The central deflection wo = v(x = ) is 2 wo = f l3 48.7EI 1 P 1− Pc (10.31) For zero axial load, Eq. (10.31) predicts a linear relation between the lateral point load and π4 ∼ deflection wo . The approximate coefficient = 48.7 is very close to the exact value 48 2 for the pin-pin column loaded by the point force f . The linear relation holds also for any constant value of P/Pc . A much more interesting picture is obtained by fixing the lateral load and changing the axial load. Equation (10.31) can be written as wo = η 1− P Pc , where η = f l3 48.7EI (10.32) which is plotted in Fig. (10.4). Note that the positive force is in compression while the negative in tension. Application of the lateral force deflects the beam by the amount η. Then, on application of the in-plane compressive load, the beam-column behaves as an imperfect column. By reversing the sign of the in-plane load from compression into tension, the central deflection becomes smaller and vanishes with P/Pc → ∞. This is fully consistent with our everyday experience that by tightening the rope/cable, its deflection is reduced. 10.3 Snap-through of a Two Bar System This is a very interesting problem, because it summaries and even extends our knowledge. There are three hinges so that each rod is a pin-pin column. The rods are elastic characterized by the bending rigidity EI, axial rigidity EA. The initial stress-free configuration is defined by the height w̄o , which was previously called the amplitude of initial imperfection. 10-7 Structural Mechanics 2.080 Lecture 10 Semester Yr P Pc 1 ⌘ 2 0 wo η -1 Figure 10.4: Relationship between the axial load and lateral deflections. Here, w̄o should be regarded as the initial shape of the structure. Upon application of the load, a compressive axial force develops in the rod, their length shortens allowing for a straight (flat) configuration. The system snaps into a new configuration where tensile force develops in the rods. Depending on the slenderness ratio, they may buckle sometime during the loading process. Pre-buckling solution Due to the unmovable hinges, the in-plane components of the displacement is zero, u = 0. The strain in the bars develops by the presence of finite rotations 1 1  = (w0 )2 − (w̄0 )2 2 2 (10.33) In the pre buckling configuration the rods are straight, so w0 = wo , l w̄0 = w̄o l (10.34) and Eq. (10.33) reduces to 1  wo 2 1  w̄o 2 − (10.35) 2 l 2 l The plot of the dimensionless strain versus the ratio wo /w̄o is shown in Fig. (10.6). From the elasticity law, the axial force in the rod is   ( EA  wo 2  w̄o 2 compressive for − w0 6 wo 6 w̄o N = EA = − = (10.36) 2 l l tensile for wo < −w̄o = 10-8 Structural Mechanics 2.080 Lecture 10 Semester Yr P I, A, h l l w̄o wo Figure 10.5: Initial and current shape of the two bar system. Equilibrium between the external load P and the membrane force N requires that wo P = −2N (10.37) l Eliminating the force N between Eqs. (10.36) and (10.37) yields     P l 1 wo 2 1  w̄o 2 − = EA − 2 wo 2 l 2 l (10.38) or in a dimensionless form P̄ = δ(δ̄ 2 − δ 2 ) where (10.39) wo w̄o P , δ= , δ̄ = (10.40) EA l l The equilibrium path given by Eq. (10.39) is the third order parabola with three roots at δ = 0, δ = ±δ̄, see Fig. (10.7). The loading process starts at A and the portion of the trajectory AB is stable. The point B is the instability point. In the process is force controlled, there is a jump to the next equilibrium configuration which is point E. So the system “snaps” into a tensile configuration and this transition is in reality a dynamic problem. The process can be displacement control and then the force P̄ is the reaction force which is positive on the segments ABC and EF but negative on the segment CDE of the trajectory. This means that an opposite force P̄ is required on CDE to keep the system in static equilibrium. By contrast, in the force controlled process the inertia force is equilibrating the system at any time. The maximum force occurs when dP = δ̄ 2 − 3δ 2 = 0 (10.41) dδ P̄ = 10-9 Structural Mechanics 2.080 Lecture 10 Semester Yr 2l2 ✏ w̄o2 Tension 0 -1 wo w̄o 1 Compression Figure 10.6: Transition from compressive into tensile strain. P̄ Force control B F C ¯ A ¯ 0 E δ Displacement control D Figure 10.7: Equilibrium path in the snap-through problem. √ 2 The maximum occurs at δ = δ̄/ 3 and the maximum force is P̄max = √ δ̄ 3 . 3 3 Any time during the loading phase AB there is a possibility for the rods to buckle. The instant of buckling is detected by equating the axial force from Eq. (10.37) to the critical buckling force of the pin-pin column N= Pcr l π 2 EI = 2 wo l2 (10.42) The dimensionless version of this equation is P̄cr = δ Pcr = 2π 2 2 EA β (10.43) l I is the slenderness ratio, and r2 = is the radius of gyration of the crossr A section. In the coordinate system (P̄ , δ), the buckling point is determined by the intersection where β = 10-10 Structural Mechanics 2.080 Lecture 10 Semester Yr of the straight line, Eq. (10.42), and the third order parabola 2π 2 δ = δ(δ̄ 2 − δc2 ) β2 (10.44) The displacement to buckle is s δ̄ 2 − δc = 2π 2 β2 (10.45) and the corresponding buckling force P̄c is 2π 2 P̄c = 2 β s δ̄ 2 − 2π 2 β2 (10.46) The graphical interpretation of the above analysis is shown in Fig. (10.8). Stocky column P̄ Slender column K L P̄C ¯ δC ¯ 0 δ Figure 10.8: Non-linear pre-buckling path intersects with a linear post-buckling path. There is a family of straight lines with the slenderness ratio as a parameter. The critical slenderness ratio for which buckling will never occur is 2 βcr = 2π δ̄ (10.47) This situation corresponds to the straight line tangent to the third order parabola. Of practical interest is the situation in which the bifurcation point occurs before the maximum force √ 2 is reached at δmax = δ̄/ 3 and P̄max = √ δ̄ 3 . The corresponding minimum slenderness 3 3 ratio, calculated from Eq. (10.42) is √ 3π 2 βmin = (10.48) δ̄ 10-11 Structural Mechanics 2.080 Lecture 10 Semester Yr To sum up, there are three ranges of the slenderness ratio: P̄max δmax Table 10.1: Ranges of buckling response βmin < β < ∞ β = βmin β = βmin s 2π 2 2π 2 2 √ δ̄ 3 δ̄ 2 − 2 No buckling static 2 β β 3 3 s or dynamic 2 2π δ̄ equilibrium path √ δ̄ 2 − 2 β 3 For the square cross-section h×h, the critical combination of the geometrical parameters h w̄o = 36π (10.49) h l From the above solution, we conclude that snap-through of the bar system without buckling will occur only for very shallow systems. 10.4 Dynamic Snap-Through The present lecture notes are restricted to static and quasi-static problems. However, the nature of the snap-through problem calls for the consideration of the full dynamic analysis. Assume that the loading of the two-bar system is load controlled. There is a stable 2 equilibrium path on the portion AB. When P̄max = √ δ̄ 3 is reached, the system jumps 3 3 instantaneously to the next equilibrium point F in the static solution. The magnitude of the force is the same, but the corresponding displacement is determined from the solution of the cubic equation 2 √ δ̄ 3 = δ δ̄ 2 − δ 3 (10.50) 3 3 This equation has three real roots δ̄ 2δ̄ δ1 = √ , δ2 = δ3 = − √ 3 3 (10.51) By adding inertia forces into the equation of equilibrium, the snap-through process occurs in time. The bar system is first accelerated on the portion BCD of the descending force and then decelerated on the rising portion DEF. The dynamic solution is straight forward if the distributed mass of the rod is lumped into two discrete point masses m = lAρ, as shown in Fig. (10.9). By adding d’Alambert inertia forces into static equilibrium, Eq. (10.37), one gets −P − 2m ẅo wo = 2N 2 l where now P is positive in tension. 10-12 (10.52) Structural Mechanics 2.080 Lecture 10 Semester Yr P m m wo Figure 10.9: The equivalent two massless bars and two lumped masses. Eliminating the axial force N in the bars between Eqs. (10.36) and (10.52), one gets the following differential equation −P̄ −   l2 ρ δ̈ = δ δ 2 − δ̄ 2 E (10.53) where the dot denotes differentiation with respect to time. It is convenient to introduce the t l E dimensionless time t̄ = , where t1 = is the reference time, and c2 = is the speed of t1 c ρ the longitudinal stress wave in a bar. In the force controlled system, the exciting term is 2 constant P̄ = √ δ̄ 3 . In the new dimensionless coordinate, Eq. (10.53) takes the form 3 3 −P̄ − δ̈ = δ 3 − δ̄ 2 δ (10.54) where the dot denotes differentiation with respect to the dimensionless time t̄. Using the chain rule of differentiation, dδ̇ dδ dδ̇ dδ̇ = = δ̇ (10.55) δ̈ = dt̄ dδ dt̄ dδ one can get a solution on the phase plane (δ, δ̇) rather tan in the time domain. Substituting Eq. (10.55) into Eq. (10.54), the following equation is obtained
−P̄ dδ − δ̇dδ̇ = δ 3 − δ̄ 2 δ dδ (10.56) which can be readily integrated to give 1 δ 4 δ 2¯δ 2 −P̄ δ − δ̇ 2 = − +C 2 4 2 (10.57) The integration constant C is determined from the initial condition that the velocity δ̇ is 1 zero when the deflection reaches δ = √ (point B). The solution for the velocity δ̇ is 3 s       δ 1 1 δ 2 1 δ 4 2 δ̇ = 2δ̄ −P + − + (10.58) 2 δ̄ 4 δ̄ 12 δ̄ 10-13 Structural Mechanics 2.080 Lecture 10 In terms of the normalized velocity (10.58) reads Semester Yr δ̇ δ = v̄ and the normalized deflection η = , Eq. 2 2δ̄ δ̄ s v̄ = 2 1 1 1 − √ η + η2 − η4 + 2 4 12 3 3 (10.59) The plot of v̄ versus η is shown in Fig. (10.10). 0.8 0.6 0.4 0.2 -2 -1 1 2 Figure 10.10: The plot of v̄ versus η in dynamic snap through. 1 The polynomial in η under the square root in Eq. (10.59) has two real roots, at η = √ 3 √ and η = − 3. The dynamic motion starts at B, increases slowly, reaches a maximum √ in F and falls rapidly to zero at the point G with the coordinate ηf = 3. Note that the dynamic deflection overshoots considerably the deflection reached in the static problem 2 ηstat = √ = 1.15. 3 At the final stage when the motion of the system stops, there is enough tensile energy stored in the bar to initial free vibration with the forcing term P̄ removed. The solution to 3 this phase is given by Eq. (10.57) with P̄ = 0, and the new integration constant C1 = 4 so that continuity of velocity is achieved. The plot of the free vibration of the system, governed by r 1 2 1 4 3 v̄ = η − η + (10.60) 2 4 4 is shown in Fig. (10.11), in comparison with the dynamic snap-through plot. 10-14 Structural Mechanics 2.080 Lecture 10 Semester Yr 1.0 0.8 0.6 0.4 0.2 -2 -1 1 2 Figure 10.11: The plot of v̄ versus η in free vibration, in comparison with the dynamicthrough plot. 10-15