Analytical Mechanics This page intentionally left blank
Download 10.87 Mb. Pdf ko'rish
|
T
A A x G –T B –T A p T B B (x B , y B ), x B > 0 Fig. 16.3 Suspended cable: catenary. 16.6 Lagrangian formalism in continuum mechanics 687 We now write the Euler equation for the functional x B 0 (f − λ) 1 + f 2 dx.
Setting f − λ = g, the latter becomes F (g, g ) = x B 0 g 1 + g 2 dx,
(16.69) for which the Euler equation takes the form gg = 1 + g 2 . (16.70) Recalling the identities cosh 2 α = 1 + sinh 2 α and (cosh α) = cosh α, the general solution of (16.70) is g(x) =
1 µ cosh(µx + c), from which f (x) = λ + 1 µ
(16.71) The conditions determining the constants µ and c and the multiplier λ are the boundary conditions at A, B: λ +
1 µ cosh c = 0, (16.72) λ +
1 µ cosh(µx B + c) = y
B (16.73)
and the constraint (16.68): = x B 0 1 + sinh 2 (µx + c) dx = 1 µ
B + c)
− sinh c]. (16.74)
Using the first two we eliminate λ: y B = 1 µ [cosh(µx B + c) − cosh c], (16.75)
and between the latter and (16.74), isolating the hyperbolic sine and cosine of µx B + c, taking the square, and subtracting, we find µ =
e −c − y B − e c + y
B (16.76)
688 Lagrangian formalism in continuum mechanics 16.6 cosh v h (v) 0
0
Fig. 16.4 (note that we must have > |y B |). In addition, adding term by term equations (16.77) and (16.75) we obtain µ( + y
B ) = e
c (e µx B − 1),
(16.77) and we finally arrive at the equation for µ: cosh µx B
1 2 µ 2 ( 2 − y 2 B ). (16.78)
Setting ν = µx B , we compare the functions cosh ν and h(ν) = 1 + 1 2 γν 2 , with
γ = ( 2 − y 2 B )/x 2 B > 1 (Fig. 16.4). Since h (ν) = γ > 1, we have h(ν) > cosh ν in a neighbourhood of the origin. Then the equation cosh ν = h(ν) has a first root ν
0 . Since also sinh(ν) = h (ν) for some ν ∈ (0, ν 0
for ν > ν, there cannot be other roots. Hence µ is uniquely determined by (16.78) (the sign must be chosen compatibly with the orientation of the axes). The constant c can then be found from (16.77), while λ can be obtained from (16.72). The problem is therefore solved. To determine the tensions at the two endpoints, note that for the second cardinal equation to be satisfied, the tensions act along lines (tangent to the cable) that must intersect on the vertical through the centre of mass of the cable (see Fig. 16.3). Denoting by p the weight of the cable, it is sufficient to decompose p along the two directions tangent at A and B, to obtain −T A and −T B . The equilibrium profile of a suspended cable is a curve called a catenary. Remark 16.2 It is a useful exercise to verify that for every arc of the catenary, the tan- gents through the endpoints intersect on the vertical through the centre of mass
16.6 Lagrangian formalism in continuum mechanics 689 of the arc. Imposing this condition it is in fact possible to obtain independently the equilibrium profile. P roposition 16.1 The horizontal component of the tension is constant along the cable, and the vertical component is equal to the weight of the arc of the catenary between the point considered and the vertex. Proof Suppose first of all that the vertex V belongs to the cable. Since the tension at the vertex is horizontal (Fig. 16.5), imposing the equilibrium conditions on each arc P V we see that only the vertical component of T(P ) varies to balance the weight of the arc P V . If V does not belong to the cable, it is enough to consider the ideal extension of the cable along the same catenary and apply the same reasoning to the extended cable. Remark 16.3 The property that the horizontal component of the tension is constant depends only on the fact that the external forces distributed along the cable (in this case, gravity) are vertical. Indeed, if ρF is the (linear) density of the external forces, the equilibrium equation of the cable is clearly −ρF(s) = d ds T(s), (16.79)
where s is the curvilinear coordinate along the cable. (B) Suspended bridge We now consider the problem of a bridge suspended by a cable through a series of hangers numerous enough that the weight can be considered to be distributed along the cable (Fig. 16.6).
Fig. 16.5 690 Lagrangian formalism in continuum mechanics 16.7
Fig. 16.6 Knowing the length of the cable and the weight of the load (such that we can neglect the weight of the cable), we want to determine the profile of the cable and the tension along it. Since we neglect the weight of the cable, the force acting on each arc ds is γ (γ = weight per unit length of the load). Therefore the specific force applied to the cable is γdx/ds, and, if y = f (x) describes the profile of the cable, this force has the expression γ/ 1 + f
2 . The dependence on f prevents the definition of a specific potential. The problem must therefore be solved by applying equation (16.79) directly. We again find that the horizontal component of the tension is constant: [T (s)x (s)] = 0. (16.80) For the vertical component we find [T (s)y (s)] = γx (s). (16.81)
Eliminating T using (16.80) and (16.81), we find that the profile of the cable is a parabola. The rest of the problem is left as an exercise to the reader. Note that it is not possible to use the variational method because in this case the forces are non-conservative. 16.7 Problems
1. Following the example of (16.36), write down the Lagrangian density for the vibrating string. 16.8 Lagrangian formalism in continuum mechanics 691 2. Extend the previous Lagrangian density to the case in which the function u depends on two space variables x 1 , x 2 and deduce the equation for the small vibrations of perfectly flexible elastic membranes. 3. Write down the Hamiltonian density corresponding to the Lagrangian dens- ity (16.48) for sound waves and show that Hamilton’s equations reduce to wave equations. 16.8 Additional solved problems Problem 1 Use equation (16.79) to solve the problem of a suspended cable, computing also the tension T(s). Solution
Writing the two components of T in the form T dx/ds, T dy/ds, the equations to be integrated are [T (s)x (s)] = 0, [T (s)y (s)] = ρg. Since d
= 1 + dy dx 2 −1 / 2 d dx , eliminating T we arrive at a differential equation for the function y(x): d 2 y dx 2 = c 1 + dy dx 2 1 / 2 , where c is a constant to be determined. This can be integrated by separation of variables, etc. Note that the vertical component of the tension is ρgs if the origin of the arcs is chosen at the vertex. This is therefore equal to the weight of the cable between the point P (s) and the vertex (whether or not this belongs to the cable). Problem 2 A heavy cable of linear density ρ and length is fixed at the endpoint A, and runs without friction on a pulley B at the same height as A. The cable is kept in tension by a weight p applied to the other endpoint. Find the equilibrium configuration. Solution We know that the profile of the cable between the two suspending points A, B is given by (16.71). Since A and B are at the same height, the catenary is
692 Lagrangian formalism in continuum mechanics 16.8 symmetric with respect to the y-axis, and hence we write (16.71) in the form f (x) = 1 µ (cosh µx − 1) − β. (16.82) Denoting by 2a the distance between the points A, B (Fig. 16.7), we must impose the condition f ( ±a) = 0, and hence 1 µ
− 1) = β. (16.83)
We also know that the tension at the point B is given by p + ρg( − λ), where λ is the length of the arc AB: λ = 2
a 0 1 + f 2 dx =
2 µ sinh aµ. (16.84) If ϕ is the angle that the tension in B forms with the x-axis, we have tan ϕ = f (a) = sinh aµ and sin ϕ = sinh aµ/cosh aµ. We now use Proposition 16.1 to write ρg µ
− 2 µ sinh aµ sinh aµ
cosh aµ , (16.85) which, setting ν = aµ, γ = dx(p/ρg), δ = dx(γ + l)/a, reduces to cosh ν + 2sinh ν = δν. (16.86) For ν > 0 the left-hand side of (16.86) and all its derivatives are positive. Denote it by χ(ν) and define ν 0 such that χ (ν 0 ) = δ. This equation has only one positive solution ν 0 (δ), as long as δ > χ (0) = 2. It is easy to verify that e ν 0 = 1 3 δ + δ 2 − 3 . (16.87)
We can now conclude that equation (16.86) is solvable if and only if δ ≥ δ
0 , with δ 0 defined by δ 0
0 (δ 0 ) = χ(ν 0 (δ 0 )).
(16.88) We leave as an exercise the proof of the existence of a unique solution δ 0 > 2
to (16.88). For fixed parameters l, a, ρ, the inequality δ ≥ δ
0 becomes a condition on the weight p: if this is too small, there cannot exist a solution. Equation (16.88) has the unique solution ν = ν 0 (δ
), if δ = δ 0 , and has two solutions, ν 1 (δ), ν 2 (δ), such that ν 1 (δ) < ν
0 (δ) < ν
2 (δ), when δ > δ 0 .
16.8 Lagrangian formalism in continuum mechanics 693
Fig. 16.7 We note that the solutions we obtained must be checked, to make sure they are compatible with the constraint λ < l, and hence, by (16.84), we have l a
(16.89) fixing a maximum admissible value ν ∗ for ν. The solutions ν i (δ) are acceptable if they are in the interval (0, ν ∗ ). Note that for δ → ∞ we have ν 1 → 0 and ν 2 → ∞. Therefore for sufficiently large p the problem always admits a solution, in correspondence to the root ν 1 (δ). The definition of the profile (16.82) of the catenary is completed by equation (16.83), which yields the value of β. It is possible in particular to study the case p = 0, characterising the conditions on l that guarantee the existence of solutions. APPENDIX 1: SOME BASIC RESULTS ON ORDINARY DIFFERENTIAL EQUATIONS A1.1 General results In this appendix we list some results of the theory of ordinary differential equations which are especially relevant for the aims of this book. For a more detailed exposition, or for the proofs that we omit, we refer the reader to Hirsch and Smale (1974) and Arnol’d (1978b). Let A ⊂ R
l be an open set and X : A → R l
C 1 . Consider the differential equation ˙x = X(x), (A1.1) with the initial condition x(0) = x 0 ∈ A.
T heorem A1.1 (Existence and uniqueness) There exist δ > 0 and a unique map x : ( −δ, δ) → A, x = x(t) of class C 1 , which is a solution of (A1.1) satisfying the initial condition x(0) = x 0 . Remark A1.1 It is well known that our hypotheses are stronger than necessary. For the exist- ence of a solution, it is sufficient for X to be continuous, while to guarantee uniqueness one must assume that X is locally Lipschitz: for every x 0 ∈ A there exist a neighbourhood U 0 ⊂ A of x 0 and a constant K 0 > 0 such that |X(x) − X(y)| ≤ K 0 |x − y| (A1.2) for every x, y ∈ U 0
Example A1.1 The equation ˙ x = x 2/3
, x ∈ R, has two distinct solutions such that x(0) = 0: x(t) = 0 and x(t) = t 3 /27. T heorem A1.2 (Continuous dependence on the initial conditions) Assume that x 1
2 (t) are both solutions of (A1.1) in the interval [0, t ] corresponding to the initial conditions x 1 (0) = x 10 , x
2 (0) = x
20 . There exists K > 0 such that for every t ∈ [0, t] we have |x 1 (t) − x 2 (t) | ≤ |x 10 − x 20 |e Kt . (A1.3)
Remark A1.2 K can be chosen to be equal to the Lipschitz constant of X. 696 Some basic results on ordinary differential equations A1.1 Remark A1.3 The estimate (A1.3) is sharp. This can be verified by considering the equation ˙ x = kx, whose solutions are x(t) = x(0)e kt . In general the solutions of (A1.1) are not defined for every t. Theorem A1.1 guarantees existence of a solution only in an interval ( −δ, δ) ⊆ R. However it is not difficult to prove that for every x 0 ∈ A there exists a maximal open interval (t 1 , t 2 ), 0
∈ (t 1 , t 2 ), in which there exists a solution x(t) of (A1.1), satisfying the initial condition x(0) = x 0 . Note that it is possible that t 1 = −∞, or t 2 = +
∞, or both. If t 1 =
2 = +
∞ the solution x(t) is global. The following theorem illustrates the behaviour of the solutions which are not global. T
tion is (t 1 , t 2 ) for bounded t 2 , for every compact set C ⊆ A there exists t ∈ (t 1 , t 2 ) such that x(t) ∈ C. Remark A1.4 By Theorem A1.3, if x(t) is not a global solution, when t → t
2 then either x(t) → ∂A or |x(t)| → +∞. Example A1.2 The equation ˙ x = −x + x
2 , x
∈ R, has the solution x(t) = −x(0)[e t (x(0) − 1) − x(0)]
−1 . This solution is not global if x(0) > 1 or x(0) < 0. A result on the continuous dependence on the data, frequently used in Chapter 12, is the following. L emma A1.1 Let X be a vector field of class C 1 and A be an open subset of R n
sup x ∈A |X(x)| ≤ ε. Let x(t) be a solution of ˙x = X(x) with the initial condition x(0) = x 0 ∈ A, and
let (t 1 , t 2 ) be the maximal interval of definition of x(t). Then |x(t) − x 0 | ≤ εt for every t ∈ (t
1 , t
2 ). Proof The function f : (t 1 , t 2 ) → R, defined by f(t) = |x(t) − x 0 | satisfies f(0) = 0 and df dt
x(t) − x
0 |x(t) − x 0 |
A1.2 Some basic results on ordinary differential equations 697 for every t ∈ (t 1 , t 2 ), from which it follows that |f(t)| = t 0 df dt (t) dt ≤ t 0 df dt dt ≤ εt. A1.2
Systems of equations with constant coefficients In a neighbourhood of a singular point x 0 one can obtain useful information about the solutions by considering the linearised equations, following a procedure analogous to the one described in Section 4.10. Setting x = x 0 + y, substituting the latter into (A1.1) and expanding the result in Taylor series in X(x 0 + y) to first order, neglecting the remainder term one arrives at the system ˙ y = Ay, (A1.4) where A = ∂X i
j (x 0 ) . The system of ordinary differential equations with constant coefficients (A1.4) can immediately be integrated: the solution corresponding to the initial condition y(0) = y
0 is given by y(t) = e tA y
, where e tA := ∞ n =0 t n n! A n (A1.5) (cf. e.g. Arnol’d (1978b, section 14)). Example A1.3 The matrix A = ⎛ ⎝
1 3 0 2 0 1 0 0 ⎞ ⎠ has eigenvalues {−1, 2, 3} to which there correspond the eigenvectors ⎛ ⎝
0 −1 ⎞ ⎠, ⎛ ⎝ 2 −3 1 ⎞ ⎠, ⎛ ⎝ 3 0 1 ⎞ ⎠. Setting M = ⎛ ⎝ 1 2 3 0 −3 0 −1 1 1 ⎞ ⎠ one has M −1 =
12 ⎛ ⎝ 3 −1 −9
0 −4 0 3 3 3 ⎞ ⎠ and
A = M ⎛ ⎝ −1 0 0 0 2 0 0 0 3 ⎞ ⎠ M −1 , and therefore e tA
∞ n =0 t n n! A n = ∞ n =0 t n n! M ⎛ ⎝ −1 0 0 0 2 0 0 0 3 ⎞ ⎠ n M −1 = M ⎛ ⎝ e −t 0 0 0 e 2t 0 0 0 e 3t ⎞ ⎠ M
−1 .
698 Some basic results on ordinary differential equations A1.2 saddle
(repulsive) mode
(repulsive) Jordan mode (attractive) star mode (attractive) focus
centre Fig. A1.1 We briefly summarise the behaviour of the solutions of (A1.4) in the case y ∈ R 2 , with the help of Fig. A1.1. Let λ 1
Download 10.87 Mb. Do'stlaringiz bilan baham: |
ma'muriyatiga murojaat qiling