Proc. R. Soc. Lond. A-2003-Budd-2097-119

Downloaded from rspa.royalsocietypublishing.
org on June 6, 2013
A nonlinear model for parallel folding with friction

C.J Budd, R Edmunds and G.W Hunt Proc. R. Soc. Lond. A 2003 459, doi: 10.1098/rspa.2003.1139, published 8 August 2003
References Email alerting service
Article cited in: http://rspa.royalsocietypublishing.org/content/459/2036/2097 #related-urls Receive free email alerts when new articles cite this article sign up in the box at the top right-hand corner of the article or click here
To subscribe to Proc. R. Soc. Lond. A go to: http://rspa.royalsocietypublishing.org/subscriptions
Downloaded from rspa.royalsocietypublishing.org on June 6, 2013
10.1098/ rspa.2003.1139

By C. J. B u d d, R. E d m u n d s a n d G. W. H u n t Centre for Nonlinear Mechanics, University of Bath, Bath BA2 7AY, UK
Received 11 November 2002; accepted 17 February 2003; published online 30 June 2003
A basic model is presented for parallel folding of two exible layers under compression, obliged by the presence of overburden pressure to remain in contact along their length. The nonlinear eect of friction is fully explored via a quasi-energy formulation and the calculus of variations, leading to representation as an ordinary dierential equation. The outcomes of both theoretical and numerical modelling are compared with a simple Galerkin approximation. The signi cance of friction-induced jamming is explored. Comparisons between the linearized dierential equation under the nonlinear boundary conditions and the full nonlinear formulation indicate that the linearization captures most of the signi cant behaviour.
Keywords: friction; nonlinear stability; structural geology; jamming; Galerkin mo del; parallel folding
1. Introduction
Geological folding due to tectonic compression is likely to occur under conditions of high overburden pressure (Hobbs et al. 1976; Price & Cosgrove 1990). In multilayered structures as formed from sedimentary rocks, the appearance of voids between layers is then either denied, or at the very least subject to severe energy penalties. If identical layers of nite thickness fold in identical fashion, without voids they can t together only by choosing straight limbs and sharp corners; this leads to the phenomenon known as kink banding (Hunt et al. 2000b). Alternatively, layers can bend with nite curvature about the same centre, implying that each layer diers from its neighbour and leading to so-called parallel folding. A simple compression test on layers of paper constrained by clamps between foam faces that deforms in such a way is shown in gure 1. Related experiments showed kink bands forming when the foam foundations were replaced with stier materials, and overburden pressures were elevated (Hunt et al. 2000 b). Whether kink bands or parallel folds are formed, one key characteristic of folding in the absence of voids is that layers will slip relative to one another. The simultaneous presence of overburden pressure means that friction then becomes of primary importance. If the deected shape is periodic, the direction of friction reverses as the slope changes sign, and this introduces a strongly nonlinear eect to what otherwise might be a linear model. This paper examines fully solutions of a simpli ed two-layer bending model in the presence of such a nonlinearity.
Proc. R. Soc. Lond. A (2003) 459 , 2097{2119 c 2003 The Royal Society
2097
2098
C. J. Budd, R. Edmunds and G. W. Hunt
Figure 1. Parallel folding in layers of paper. For visual clarity, approximately every 10th layer has been edged in black.
The paper starts by introducing a two-layer, axially rigid, exural model and summing the contributing energy terms, which comprise strain energy of bending, foundation energy and work done by load. Friction is also included as a quasi-energy contribution. The expected phenomenology is then explored using a single-degree-offreedom Galerkin approximation, which allows description of the important jammed region. The eects of small imperfections are assessed, and similarities with the corresponding unfolding of a bifurcation point in the absence of friction are explored. A dierential equation for the response of the two-layer model is then developed from the calculus of variations, and the subtle dierences between it and the Galerkin model explored using a phase-plane representation. The paper nishes with a few comparisons with a fully nonlinear model produced by consideration of the full set of possible geometric eects.
2. Pseudo-potential energy
(a ) Nonlinear Formulation Consider two axially and transversely incompressible layers of thickness t, formed from a material of bending stiness EI , embedded in a soft foundation of transverse stiness k per unit length, and compressed longitudinally by a load P such that they remain in contact along their length, as shown in gure 2. The coordinate system for this model is taken as the arclength, x, and the vertical deection, w . To t snugly (without voids), the layers must bend about the same centre of curvature. If centrelines are to remain unchanged in length, dierential stretching at the interface generates slip, s, between the layers as illustrated. The work done against friction during slip will be taken into account as a quasi-energy contribution to the total potential energy. (i) Bending energy If the interface bends to a radius of curvature R , the centreline of the inner layer must bend to a radius R t=2 while that of the outer layer takes a radius of R + t=2.
Proc. R. Soc. Lond. A (2003)
2099
x s R dq dx dw
Figure 2. Slip between incompressible layers constrained to remain in contact.
The bending energy from both layers dUB over an incremental length dx is then 1 1 dx: (2.1) + EI dU B = 1 2 (R + t=2)2 (R t=2)2
If t2 4R 2 , this can simply be rewritten as dUB = EI =R2 dx and the total bending energy contribution over a length L, written in terms of deection w, becomes (Thompson & Hunt 1973) Z L w2 dx; (2.2) UB = EI 1 w _2 0
where dots denote dierentiation with respect to x. Subject to the above restriction on t, this expression holds over large deections. (ii) Foundation energy For a linear foundation of stiness k per unit length, the foundation energy is simply Z L k w2 dx: (2.3) UF = 1 2
0
(iii) Work done by load If the compressive load P moves inwards by the shortening of the interface, the corresponding deection E of the load P over large deections is (Thompson & Hunt 1973) Z Lp 1 w _ 2 dx E =L
(1
0
Hence the work done by the load is Z PE = P

w _ 2 ) dx:
(2.4)
(1
0
w _ 2 ) dx:
(2.5)
2100
(iv) Work done against friction Over the incremental length dx seen in the inset to gure 2, if the centreline of each layer is to remain unchanged in length, there must be a dierence in length between the outermost bre of the inner layer and the innermost bre of the outer layer. From simple bending theory, it follows that, over a small change in angle d , this dierence is Rt R R dx: (2.6) dx = 2 ds = R (t=2)2 R t=2 R + t=2
If, as for the bending energy, it is assumed that t2 4R 2 , this reduces to ds = t dx = t d : R
(2.7)
As slip at position x is cumulative, the total slip s at x is given by Z x Z s= ds = t jd j = t j j:

0 0
(2.8)
Here a modulus sign is included to ensure that for positive or negative the work done is maintained as positive. Integrating as before over the range 0 < x < L the total work done against friction is then Z L Z L U = qt j j dx = qt jsin1 w _ j dx; (2.9)
0 0
where is the coe cient of friction and q is the overburden pressure. (b ) Potential energy function The total potential energy function comprises the bending energy minus the work done by the load (Thompson & Hunt 1973), together with the contribution from the work done against friction. The sign of this friction term can be either positive or negative, depending on whether the friction acts to resist the release of strain energy or in the opposite sense. To allow for all possibilities we therefore introduce a friction indicator (Brogliato 1999) = 1 on the friction energy. Here, = +1 implies that friction opposes the external force and gives a positive friction energy contribution; alternatively, = 1 implies that friction acts in the same sense as the external force, giving a negative friction energy contribution. We can then write the total potential energy function, valid over large deections, as follows: V = UB + UF P E + U Z L Z L w2 1 k w 2 dx d x + = EI 2 2 1 w _ 0 0 Z Z L p 2 1 w _ ) dx + qt +P (1
0
L 0
jsin1 w _ j dx:
(2.10)
The solutions of the system are then stationary points of this energy functional.

P w P x P
2101
P w x P
Figure 3. Slip directions for = +1. Top: Q positive. Bottom: Q negative.
(i) Small-de ection potential energy function If w is assumed to be small, the energy function (2.10) reduces to Z L kw2 + qtjw _ j) dx: Pw _2 + 1 V = (EI w2 1 2 2
0
(2.11)
3. A useful Galerkin approximation

The behaviour of the system can usefully be seen in the context of an unfolded bifurcation plot. To develop such a description, let us assume a deected shape, x ; (3.1) w (x) = Q cos L
(w (0) w(L)). We then consider the energy V (Q; L) of the solutions so that Q = 1 2 of this form and look for stationary solutions in this restricted class. Later we extend the analysis to the full class of solutions, but this simple analysis leads to considerable insights. The deection is illustrated in gure 3 over the range L 6 x 6 L, for Q positive at the top and Q negative at the bottom. When Q > 0, the function w (x) is decreasing on the interval [0; L], so on this interval we could write jw _j = w _ , where = 1. Similarly, when Q < 0, the function w(x) is increasing on the interval [0; L] so we could again write jw _j = w _ but with = 1. This alternative notation will prove useful when we look at the more general solutions by means of the calculus of variations. When the ansatz (3.1) is substituted into the linearized potential (2.11), the resulting energy is given by 2 4 2 1 1 kLQ2 + 2 qtjQj: (3.2) Q2 + 1 Q 4PL V = 2 EI L 4 L L
Here the modulus on w _ , and hence on Q, ensures that for positive , the energy contributions from all four quadrants of gure 3 are always positive as illustrated.
2102

P minimum instability load
c = +1
c = +1
jammed region
c = -1
c = -1
Q
Figure 4. Bifurcation diagram indicating jammed region for constant .
(a ) Bifurcation diagram The combination of the coe cient of friction and the friction indicator can now be seen in exactly the role of an imperfection, `unfolding the bifurcation point at Q = 0 and P = P C (see x 5 c) as shown in gure 4. To derive this gure we look at the stationary values of V with respect to variations in Q, so that we solve the equation @V =@Q = 0. This gives Q=
2 =L
4 qt ; 2EI 4 =L3 kL
(3.3)
where the ` term in this expression arises from the modulus sign in the expression (3.2) for V , so that we take the `+ sign if Q is positive and the ` sign if Q is negative. Indeed, the presence of the modulus sign jQj gives the unfolding a rich and novel structure, extending the more usual classical picture. (b ) Critical loads and wavelength selection There is a critical value of P = P C at which jQj = 1. From (3.3) this is given by PC = 2EI L2
2
kL2 : 2
(3.4)
Conventionally, this is obtained by minimizing P C over all possible values of L. It is easy to see that this occurs when r 4 2EI (3.5) L= k and p (3.6) P C = 2 2EIk:
This expression corresponds to the classical result for a strut on a linear elastic foundation of bending stiness 2EI (Hunt et al. 1989), and allows us to determine the critical wavelength at which we expect to see the unfolded bifurcation structure.
2103
Alternatively, we may determine L from the condition that the energy per unit length should be stationary with respect to changes in L (Peletier 2001), so that @ (V =L) = 0: @L
This leads to the condition 2 4 Q2 P Q2 + 1 2EI 2 L L
2 qt
jQj = 0: L
(3.7)
Combining the expressions (3.3) and (3.7) it is easy to see that the critical wavelength is again given by (3.5). Thus the alternative conditions on L, for the minimum critical load and for the energy to be stationary with respect to changes in L, lead to identical values for the wavelength. (c ) Stability of the solution paths Stability of equilibrium under dead load is governed in the rst instance by the second derivative (Thompson & Hunt 1973):
EI 4 d2 V = 2 L3 dQ
1 P 2 kL: +1 2 2 L
(3.8)
If d2 V =dQ2 > 0, there is a relative minimum of V and the solution is stable. Conversely, if d2 V =dQ2 < 0, there is a relative maximum and it is unstable. Clearly, with P C given by expression (3.4), if P < P C , we have stability, and if P > P C , we have instability. (d ) Jamming At constant load P , points in the region between the two curves de ned by = 1 (or +1) are stationary positions where the system can be considered to be `trapped or `jammed between the two critical slip conditions with Q positive and negative and the frictional forces acting in opposite senses (as in the top and bottom of gure 3). Anywhere within the jammed region the system sits in equilibrium. The nite area of the region therefore replaces the in nitesimally thin equilibrium paths of a perfect (bifurcating) system without friction (Thompson & Hunt 1973). Placed outside the jammed region, at constant load the system would move horizontally towards it until being brought to a halt at its boundary. A typical sequence under changing load might then be as follows. At zero load, the system can rest in equilibrium anywhere along the Q-axis within the jammed region. To get to the intersection with a = 1 line for example, a jQj value greater than that at the intersection can be input, whereupon the system will slip back to the boundary. This reverses the direction of the frictional arrows from those shown in gure 3. Application of positive load P then moves the system into the jammed region, following the dashed line, until the equivalent critical state for = +1 is met. Instability then occurs, i.e. the system will suddenly deect with jQj increasing, and continue to deect. In the presence of a suitable nonlinearity there may be restabilization at some other larger value of jQj, but for the linear system this would never happen.
2104
C Note that symmetry of the curves about the value p P = P suggests that the C minimum instability load is always at P = 2P = 4 2EI k , independently of the value of (6= 0). This suggests that, as vanishes, there would be a sudden drop in the load to cause instability, from 2P C to P C . The eect is lost with the introduction of a small imperfection, as described in the following section.
(e ) Initial imperfections An imperfection in the sense of Q, implying that the system has a natural bias towards buckling in the positive Q direction, would be expected to appear in the energy function as a term in Q with a negative coe cient (Thompson & Hunt 1973). Hence, from (3.2), 2 4 2 1 1 kLQ2 + 2 qtjQj Q Q2 + 1 Q 4PL V = 2 EI L 4 L L
(A
P B )Q2 + 2 qtjQj
Q:
(3.9)
The imperfection has the eect of shifting the jammed region as shown in gure 5, which can be seen by minimizing (3.9) over all Q. If Q > 0, then @V = 2(A @Q
P B )Q + (2 qt (2 qt ) : 2(A P B )
) = 0;
giving Q=
(3.10)
If Q < 0, then
@V = 2(A @Q
P B )Q
(2 qt + ) = 0;
leading to Q=
2 qt + : 2(A P B )
(3.11)
Let threshold value th = j2 qtj. Then if 0 < < th , we get gure 5a and if > th , gure 5b. If the magnitude of is greater than that of th , the bias is strong enough to push the critical slip curve for = +1 and Q > 0 into the region where P < P C , as shown in gure 5b. The jammed region then divides into two and misses the bifurcation point altogether. As th depends linearly on , th ! 0 as ! 0. The sudden drop in minimum instability load described in the previous section is therefore destroyed by the presence of a vanishingly small imperfection.
4. Calculus of variations
The Galerkin analysis of the previous section has constructed a stationary solution of V over the class of cosine functions. We now extend this analysis to nd the stationary solutions of V over the class of all admissible functions. We will continue to consider solutions which have a half wavelength of L so that they are monotone over the interval [0; L] and w _ (0) = w _ (L) = 0: (4.1)

(a) P (b) P
2105
c = +1
c = +1
c = +1
c = -1
c = -1
c = -1
c = +1
c = -1
Q Figure 5. Jamming in the presence of a positive initial imperfection . (a ) 0 < < th . (b ) > th .
We initially assume that L is known. Later we consider two alternative mechanisms for nding L, either that the critical load P C is minimized (equivalent to the calculation in (3.6)) or such that V =L is stationary with respect to changes in the length-scale. (Both mechanisms have been seen to be equivalent for the space of cosine functions.) Over a halfwave of length L, the total potential energy of the system has the form Z L V = F (w; w; _ w ) dx;
0
where F , the total potential energy per unit length, is given by F = EI w2 1 w _2
P (1
where = 1. As L is half a wavelength, w _ does not change sign and thus is xed at either +1 or 1 and F is a smooth function of w , w _ and w . Here, as before, = +1 if w is increasing over the half interval [0; L] and = 1 if w is decreasing over the same half interval. Applying the calculus of variations by taking a small increment w of the function w, we nd that the perturbation to the energy V is given to leading order by the expression (Thompson & Hunt 1973) Z L @F @F @F w dx: (4.3) w _+ w+ V = @w @w _ @w 0
kw2 + qt sin1 (w _ ); w _ 2) + 1 2
(4.2)
The order of dierentiation in this expression may be reversed so the second term may be written as Z L Z L @F d @F w dx: w _ dx = _ dx @ w _ 0 @w 0
2106
Integration by parts of this expression yields L Z L 2 L L d @F d @F @F @F w + w _ w + V = d x2 @ w @ w d x @ w @w _ 0 0 0 0
d @F @F w dx: + @w _ dx @ w (4.4) For V = 0 for all w we require both that the ordinary dierential equation
@F d @F d2 @F =0 (4.5) + 2 @w _ dx @ w dx @ w should be satis ed and that each of the expressions in the square brackets should be zero. This leads to a new set of boundary conditions for the function w. The boundary condition (4.1) forces all admissible perturbations to satisfy
w _ (0) = w _ (L) = 0; and hence terms in the second square bracket vanish identically. The rst and third terms in the square brackets also vanish for all w if, at x = 0 and L, we have @F @w _
d @F = 0: dx @ w
(4.6)
Substituting from the explicit form (4.2) for F , and making use of the boundary conditions (4.1), this reduces to ... qt 2EI w = 0; at x = 0 and L. We thus nd the additional boundary conditions qt ... ... w (0) = w (L) = : 2EI
(4.7)
The new boundary condition (4.7) represents a step change in the shear force at the boundary, and requires some explanation. Frictional traction on a layer sums to a force of magnitude qL, acting along an outermost bre as shown in gure 3. Each is reacted at one end of the region in question by an equal and opposite force acting at the neutral axis of the layer. This sets up an out-of-balance moment in each layer, of magnitude qLt=2, which in turn is resisted by a couple comprising lateral point loads at the ends of the region. It is these reactive point loads that appear as the boundary conditions (4.7). Performing the necessary dierentiations on the explicit form of F , from the Euler{ Lagrange equation (4.5), we obtain the nonlinear dierential equation .... 2EI [ w (1 + P w(1 ... w _ 2 ) 1 + f 4 w w w _ + w3 g(1 w _ 2 )1=2 (1 + w _ 2 (1 w _ 2 ) 2 + 4 w 3 w _ 2 (1 qtwjw _ j(1 w _ 2 )3 ]
w _ 2 ) 1 )
w _ 2 )3=2 + kw = 0: (4.8)
The associated linear dierential equation for small displacements is given by .... 2EI w + P w + kw = 0; (4.9) with the same boundary conditions as before. This equation corresponds to taking the stationary values of the energy resulting from the `linearized energy density given by kw 2 + qtjw _ j: (4.10) Pw _2 + 1 F = EI w 2 1 2 2
2107
It is interesting that the only way that friction enters the linear dierential equation (4.9) is as the boundary term (4.7). Although friction acts along the length, bending theory is based on the assumption that plane sections remain plane, and thus at rst order it is unable to account for a distributed shearing eect like that being imposed here. The linearity of (4.9) simpli es the analysis and allows us to deduce some results immediately without solving the dierential equation explicitly. In particular we have the following. Lemma 4.1. If k > 0, then Z
L
w d x = 0:
0
(4.11)
Proof . From (4.9) it follows that w= 1 .... (2EI w + P w ): k
Thus
1 ... .... [2EI w + P w _ ]L (2EI w + P w) dx = 0 = 0; k 0 a RL using (4.1) and (4.7). For consistency, when k = 0, we insist that 0 w dx = 0.
1 w ( x ) dx = k
Lemma 4.2. The function w depends linearly on the friction.
Proof . Suppose that the function u(x) satis es the equation plus boundary conditions .... 2EI u + P u + ku = 0; u _ (0) = u _ (L ) = 0 ; qt ... ... u (0) = u (L) = : 2EI
Let w = u, then
du d( u) dw ; = = dx dx dx etc. Substitution leads to an identical equation, .... 2EI w + P w + kw = 0;
with the boundary conditions w _ (0) = w _ (L ) = 0 and qt ... ... w (0) = w (L) = : 2EI
2108
5. An analysis of the linear equation

The linearity of the dierential equation (4.9) makes analysis straightforward, as seen in this and the next section. This analysis gives considerable insights into the non-classical role played by the modulus term representing the friction. A fuller numerical investigation of the nonlinear equation (4.8) is given later. The case of k = 0 is technically easier and is given rst. (a ) The solution for k = 0 When k = 0 the ordinary dierential equation (4.9) becomes 9 .... 2EI w + P w = 0; > > = w _ (0) = w _ (L ) = 0 ; (5.1) > qt > ... ... ; w (0) = w (L) = : 2EI Integrating once, and using the boundary conditions to determine the constants of integration we have qt ... : (5.2) 2EI w + P w _ = 2EI In the phase space describing (w; _ w) the solutions of this equation give trajectories that are piecewise circular arcs centred on the point qt ;0 : (w; _ w) = P
Hence (5.2) has the solution qt x + A cos w (x ) = P
r P P x + C: x + B sin 2EI 2EI
(5.3)
To nd the constants A and B we apply the boundary conditions to give the solution p r r r P P P 1 2EI qt qt x + C: x sin L cos tan x+ w= 2EI 2EI 2EI 2 P 3=2 P (5.4) The free constant of integration C can be xed by insisting (following lemma 4.1) that the mean of w should be identically zero. For the solution of (5.4) to be consistent with the earlier analysis, we require that w(x) should be monotone over the interval [0; L]. This imposes a restriction on P and it is easy to see that for monotonicity we require that P should lie in the interval 8 2 EI : (5.5) P 2 [0; Pm ax ] 0; L2
Here, Pm
ax
is the critical value at which w(0) = w (L) = 0: (5.6)
Consistent with the earlier Galerkin analysis, we de ne
(w (0) Q= 1 2
w(L));
A nonlinear model for parallel folding with friction so that, after some manipulation, we have r r P 1 2EI qt L ; tan L+2 Q= 2EI 2 P 2P
2109
(5.7)
which is directly comparable with the expression (3.3). We see from (5.7) that jQj becomes in nite at the critical value of the external load given by 2 2 EI : (5.8) PC = L2 This is precisely the value obtained by the Galerkin analysis given earlier in (3.4) when k = 0. However, a consequence of setting k = 0 is that L must be xed a priori. As L increases, P C drops, until for a two-layer specimen of in nite length, P C = 0 and L = 1. Hence while the above analysis gives insight into the form of the solutions of the problem, it highlights the importance of the foundation stiness in selecting the value of L, and we return to this point presently.
(b ) Bifurcation diagrams for k = 0 Using the expression (5.7) we may plot the (Q; P ) bifurcation diagram and the resulting solutions described by (5.4). There are two cases to consider here. (i) Case 1: slipping In this case the friction indicator is xed at one of the two critical values of = 1. The value of Q for each value of P is then given by (5.7). There are four quadrants to the resulting gure, obtained by taking = 1 and = 1, as shown in gure 6. This picture is very similar, both quantitatively and qualitatively, to that given in gure 4, although now the external load P is restricted to be less than Pm ax . If = +1, we have a positive energy contribution due to friction, which in this case acts like strain energy of bending in resisting the external load. If = 1, we have a negative energy contribution due to friction which now acts in the opposite sense, to resist the release of strain energy. Examples of these two extremes are given in the (w; _ w ) phase plane in gure 7, along with a third solution well within the equivalent of the jammed region of gure 4, at = 0:032. Over most of the trajectory these solutions appear very similar, with the only signi cant dierences arising at the ends of the sample. This reects the dierence ... in w , which comes about as the directions of the point load reactions reverse with the change in sign of . (ii) Case 2: jamming Some nonlinear work against friction is experienced in moving between the two solutions given by = 1 and = +1 through the equivalent of the jammed region of gure 4. However, in the linearized equation (4.9), the total work done against friction over one half wavelength is Z L qt w _ dx = 2 qtjQj: (5.9)
0
2110

40 35 30 25 20 0 1 15 10 P
0.1
0.1 0 -0.1 0 1
c = +1 a = +1
0 -0.1
c = +1 a = -1
c = -1 a = +1
-0.5
0.1 -0.1 0 1
5 0
0.1 0 0.1 -0.1
0 -0.4 -0.1
c = -1 a = -1
0.4 0 1 0.5 Q
Figure 6. Plot of load P against amplitude Q, for k = 0, EI = 0:5, L = 1 and qt = 0:1.
a = -1
1.0 0.8 0.6 0.4 0.2
a = +1
-0.3
-0.2
-0.1
-0.2 -0.4 -0.6 -0.8 -1.0
0.1
0.2
0.3
c = -1 c = 0.032 c = +1
Figure 7. Phase-plane trajectories plotted for k = 0, EI = 0:5, L = 1 and qt = 0:1.
Hence, there is no dierence in the work done against friction between any two solutions with the same Q. A path of jammed solutions in the jamming region can then be described by xing Q in the expression (5.7) and letting vary between 1 and +1. This in turn implicitly determines the value of P corresponding to each such value of . Note that if ! 0, then we have P ! P C and the solution approaches the pure sinusoid function w(x) = Q cos( x=L ), as considered in the Galerkin analysis.
A nonlinear model for parallel folding with friction (c ) The solution for k 6= 0
2111
The solution of (4.9) for k > 0 can be obtained along similar, although more complex, lines to that for k = 0. This analysis allows a value of L to be determined. Seeking a solution of (4.9) of the form w(x) = ei x we obtain 1 p 2 P 2 P 8EI k: (5.10) = 4EI 4EI
Observe that if is real, then the minimum value of P over all values of is given when p P : (5.11) P = 2 2EI k and 2 = 4EI 2 2 2 Let !1 = and !2 = + 2 , where are given in (5.10). Then
w(x) = A1 cos(!1 x) + B1 sin(!1 x) + A2 cos(!2 x) + B2 sin(!2 x): From the boundary conditions given in (4.1) and (4.7), at x = 0 we have !1 B 1 + !2 B 2 = 0 ; qt 3 3 ; !1 B 1 !2 B2 = 2EI
(5.12)
and hence, B1 = qt 2 !2) 2EI !1 (!1 2
and
B2 =
qt : 2 !2) 2EI!2 (!1 2
(5.13)
At x = L we have similarly that !1 A1 sin(!1 L) + !1 B1 cos(!1 L)

3 !1 A1 sin(!1 L)
!2 A2 sin(!2 L) + !2 B2 cos(!2 L) = 0; qt 3 3 3 : !1 B 1 cos(!1 L) + !2 A2 sin(!2 L) !2 B2 cos(!2 L) = 2EI

3 !1 !2 (!1 3 !2 ) sin(!1 L) sin(!2 L) 6= 0;
Solving this simultaneous system, provided that (5.14)
A1 and A2 are obtained uniquely. Loss of solvability occurs when either !1 L or !2 L is a multiple of . The corresponding value of P = P C thus arises when from (5.10) equals =L, so that kL2 2EI 2 (5.15) + 2 : PC = 2 L This result corresponds with the value (3.4) given by the Galerkin analysis.
(d ) Wavelength selection For a particular foundation stiness k , the length of the halfwave will determine the value of the bifurcation point P C . The question we now ask is how is the wavelength, and hence P C , selected? As in the Galerkin analysis, we may choose L either to minimize the critical load P C , or to seek the value of L such that @ (V =L)=@L = 0. For the Galerkin approximation these two approaches give the same answer, but in this more general case the answers are dierent, although very similar.
2112
10

k = 10 k=5 k=2
PC
4
k=1
k = 0.5
2
k = 0.2 k = 0.1
0
1
Figure 8. Variation of P C with L for di erent values of k .
(i) Minimization of the load To nd the halfwavelength associated with the minimum critical load we can minimize P C over all values of L. Figure 8 shows how P C changes with L for various values of k . Dierentiating (5.15) with respect to L and setting the result to zero leads to r 4 2EI ; (5.16) L= k
corresponding to (3.5) of the Galerkin analysis. Substituting into (5.15) then again gives the classical expression (3.6) for the minimum value of P C for the strut on linear elastic foundation (Hunt et al. 1989). ^ = P k 1=2 . Then the linear dierential equation (4.9) rescales Let y = k 1=4 x and P to ^ wyy + w = 0: 2EIwyyyy + P (5.17) Also, if we have L as given in (5.16), then w satis es the boundary conditions p 4 2EI ) = 0 (5.18) wy (0) = wy (
and wyyy (0) = wyyy ( p 4
2EI ) =
qt 3=4 k : 2EI
Let k 3=4 = ^ . Then we have wyyy (0) = wyyy ( p 4
2EI ) =
^ qt : 2EI
(5.19)
We deduce that all solutions of (4.9) for general k , with the boundary conditions satis ed when L is given by (5.16), are rescalings of the solution when k = 1.
A nonlinear model for parallel folding with friction (ii) Stationary values of energy density
2113
Alternatively, we may consider the problem of nding stationary values of the energy per unit length (Peletier 2001). To do this we set x = L and express all derivatives with respect to . This allows us to rescale the linearized energy so that all calculations are over the length [0; 1]. We then have Z 1 V kw 2 + L1 qtjw0 j) d ; (5.20) L 2 P (w 0 )2 + 1 = (L4 EI (w00 )2 1 2 2 L 0
where primes refer to derivatives with respect to . Dierentiation with respect to L gives Z 1 @ (V =L) = (4L5 EI (w 00 )2 L3 P (w0 )2 + L2 qtjw0 j) d : (5.21) @L 0
For a stationary solution, @ (V =L)=@L = 0. Thus, after a rescaling of the above integral to be an expression in the original variables, it follows that for @ (V =L)=@L = 0 we must augment the ordinary dierential equation (4.9) plus the boundary conditions (4.1), (4.7) with the integral condition Z L (4EI w 2 P w _ 2 + qtjw _ j) d x = 0 : (5.22)
0
Equation (5.22) implicitly de nes the length L. The linearized ordinary dierential equation for w is now given by .... 2EI w + P w + kw = 0; ... ... with w _ (0) = w _ (L) = 0 and w (0) = w (L) = qt=2EI . Multiplying by w , integrating by parts over [0; L], and applying the boundary conditions, we have Z L [ qtw ]L + (2EI w2 P w _ 2 + kw 2 ) dx = 0: (5.23) 0
0
But [ so that Z
L 0
qtw ]L 0
qtjw _ j dx;
0
(2EI w2
Pw _ 2 + kw2 + qtjw _ j) dx = 0:
(5.24)
Combining the two equations (5.22) and (5.24) gives the following two results. Lemma 5.1. If w and L are such that V = w = 0 and @ (V =L)=@L = 0, then Z L Z L 2 2EI w dx = k w 2 dx; (5.25)
0 0
or the strain energy of bending equals the energy stored in the foundation.
Lemma 5.2. The critical value of L dened by lemma 5.1 is independent of the friction .
2114

c = -1 c = +1
3.5
3.4 L 3.3
3.2
3.1
0.2
0.4
0.6 Q
0.8
1.0
1.2
Figure 9. Di erence in wavelength prediction between equations (5.16) and (5.26) plotted for EI = 0:5, k = 1, = 1 and qt = 0:1.
It should be noted that if we approximate w (x) by Q cos( x=L ) and w (x) by ( =L)2 Q cos( x=L ), then (5.25) gives that L (2EI =k )1=4 , which is precisely the expression obtained earlier. However, for 6= 0, the result for L diers slightly from that of (5.16), as shown in gure 9. If as before we rescale the identity (5.25) by setting x = L , we have Z Z 1 2EI 1 002 w d = k w 2 d : L4 0 0
Thus
R1 But, from lemma 4.1, 0 w d = 0, and also w (0) = w (1) = 0. It therefore follows from repeated applications of the Poincar e inequality (Adams 1975) that Z 1 Z 1 002 4 w d > w 2 d : (5.27)
0 0
R1 2EI 0 w002 d : L = R1 2 d k w 0
4
(5.26)
Combining results we nd that for all solutions r 4 2EI ; L> k
(5.28)
as seen in gure 9.
6. Numerical results
Numerical solutions to the governing dierential equations (4.9) and (4.8) subject to boundary conditions (4.1), (4.7) have been obtained using the numerical continuation code AUTO (Doedel et al. 1997). In accordance with the convention of this package, the output that follows is given over the rescaled length = [0; 1] discussed earlier.

P
2115
c = +1 a = +1
c = +1 a = -1
c = -1 a = +1
c = -1 a = -1
-0.5
-0.4
-0.3
-0.2
-0.1
0.1
0.2
0.3
0.4
Q 0.5
Figure 10. Bifurcation plot for the linear di erential equation (4.9) plotted for EI = 0:5, k = 1, qt = 0:1 and L = .
0.08
P = 0.52 P = 3.48 w 1 -0.2 -1 0.2 w w 0.5 -0.2 0.2 -0.5 0 0.2 0.4 0.6 0.8 1.0 w
0.04
-0.04
-0.08 x
Figure 11. Wave-shape and corresponding phase portraits for the case of gure 10 at Q = 0:082.
(a ) Linear equation Numerical solutions to the linear equation (4.9) are presented in gure 10. Corresponding wave shapes and phase portraits are given in gures 11 and 12 for the two dierent values of Q = 0:082 and Q = 0:42, respectively, at the two extremes of slip represented by = 1.
2116
0.4

P = 1.85 P = 2.16 w 5
0.2
-1 -5
w w 5
-0.2 -1 -0.4 0 0.2 0.4 x -5 0.6 1
0.8
1.0
c = +1 a = +1
c = +1 a = -1
c = -1 a = +1
c = -1 a = -1
Q
-0.8
-0.4
0.4
0.8
Figure 13. Bifurcation plot for the nonlinear di erential equation (4.8) plotted for EI = 0:5, k = 1, qt = 0:1 and L = .
(b ) Nonlinear equation Numerical solution of the full nonlinear equation (4.8) for a typical set of parameter values is given in gure 13. Wave shapes and the corresponding phase portraits are given in gures 14 and 15 for the two dierent values of Q = 0:082 and Q = 0:42, respectively, at the two extremes of slip represented by = 1.

0.08 P = 0.41 P = 3.50
2117
0.04
w 1 w w -1
-0.2
0.2
-0.04 -0.2 -0.08 0 0.2 0.4 x
0.5 0.2 -0.5 0.6 0.8 1.0 w
0.4
P = 1.65 P = 2.26
0.2
w 4 w
0 -1 1 -4
w 4
-0.2
-1 -0.4 0 0.2 0.4 x -4 0.6
0.8
1.0
7. Concluding remarks
Comparisons between the solutions of the linear and nonlinear equations are presented in gure 16 for two dierent values of , one small and one large. These, along with the plots of gures 10{15, demonstrate that little of phenomenological signi cance is added by including the full set of geometric nonlinearities. Figure 13 indicates a slight overall downwards curvature to the = 0 response in comparison
2118
5 4 P 3

(a)
linear nonlinear
4.8 4.4 P 4.0 3.6
(b)
2 0 0.2 0.4 0.6 Q 0.8 1.0 1.2
3.2 0 0.2 0.4 0.6 Q 0.8 1.0 1.2
Figure 16. Comparisons of linear and nonlinear solutions: (a ) qt = 0:1; (b) qt = 1.
with the linear view of gure 10 for example, but the related 6= 0 curves show little dierence of signi cance. By comparison, the nonlinearity that is brought in by the modulus change at the boundaries has a profound phenomenological eect on the system, as demonstrated most clearly in the Galerkin plots of gures 4 and 5. Analysis throughout has concentrated on a single halfwave of buckling, no indication being given of how the pattern continues into further waves. The unstable nature of the nal buckle (see gure 4) suggests that homoclinic (Budd & Peletier 2000; Budd et al. 1999) or localized behaviour is to be expected, which in this case might manifest itself as the formation of a single halfwave, with the remainder of the layers remaining jammed in the straight con guration. Then, if a tendency to restabilize is found over large deections, a form of cellular buckling would be expected (Hunt et al. 2000a), with halfwaves forming and then locking-up in sequential fashion. In such circumstances, a heteroclinic orbit that connects the at state to a periodic con guration de ned by Maxwell considerations (Budd et al. 2001) becomes the dominant characteristic.
References
Adams, R. A. 1975 Sobolev spaces. Academic. Brogliato, B. 1999 Nonsmooth mechanics, 2nd edn. Springer. Budd, C. J. & Peletier, M. A. 2000 Approximate self-similarity in models of geological folding. SIAM J. Appl. Math. 60, 990{1016. Budd, C. J., Hunt, G. W. & Peletier, M. A. 1999 Self-similar fold evolution under prescribed end-shortening. Math. Geol. 31, 989{1005. Budd, C. J., Hunt, G. W. & Kuske, R. 2001 Asymptotics of cellular buckling close to the Maxwell load. Proc. R. Soc. Lond. A 457, 2935{2964. Doedel, E. J., Champneys, A. R., Fairgrieve, T. F., Kuznetsov, Y. A., Sandstede, B. & Wang, X. J. 1997 AUTO97: continuation and bifurcation software for ordinary di erential equations. Technical Report, Department of Computer Science, Concordia University, Montreal, Canada. (Available at ftp://ftp.cs.concordia.ca/pub/doedel/auto.) Hobbs, B. E., Means, W. D. & Williams, P. F. 1976 An outline of structural geology. Wiley. Hunt, G. W., Bolt, H. M. & Thompson, J. M. T. 1989 Structural localization phenomena and the dynamical phase-space analogy. Proc. R. Soc. Lond. A 425, 245{267. Hunt, G. W., Peletier, M. A., Champneys, A. R., Woods, P. D., Ahmer, W. M., Budd, C. J. & Lord, G. J. 2000a Cellular buckling in long structures. Nonlin. Dynam. 21, 3{29.
2119
Hunt, G. W., Peletier, M. A. & Ahmer, W. M. 2000b The Maxwell stability criterion in pseudoenergy models of kink banding. J. Struct. Geol. 22, 669{681. Peletier, M. A. 2001 Generalized monotonicity from global minimization in fourth-order ordinary di erential equations. Nonlinearity 14, 1221{1238. Price, N. J. & Cosgrove, J. W. 1990 Analysis of geological structures. Cambridge University Press. Thompson, J. M. T. & Hunt, G. W. 1973 A general theory of elastic stability. Wiley.

Proc. R. Soc. Lond. A-2003-Budd-2097-119

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Proc. R. Soc. Lond. A-2003-Budd-2097-119

Transféré par

Droits d'auteur :

Formats disponibles

Downloaded from rspa.royalsocietypublishing.

org on June 6, 2013

A nonlinear model for parallel folding with friction

References Email alerting service

To subscribe to Proc. R. Soc. Lond. A go to: http://rspa.royalsocietypublishing.org/subscriptions

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

A nonlinear model for parallel folding with friction

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

C. J. Budd, R. Edmunds and G. W. Hunt

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

A nonlinear model for parallel folding with friction

Figure 2. Slip between incompressible layers constrained to remain in contact.

Hence the work done by the load is Z PE = P

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

C. J. Budd, R. Edmunds and G. W. Hunt

If, as for the bending energy, it is assumed that t2 4R 2 , this reduces to ds = t dx = t d : R

As slip at position x is cumulative, the total slip s at x is given by Z x Z s= ds = t jd j = t j j:

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

A nonlinear model for parallel folding with friction

Figure 3. Slip directions for = +1. Top: Q positive. Bottom: Q negative.

3. A useful Galerkin approximation

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

C. J. Budd, R. Edmunds and G. W. Hunt

Figure 4. Bifurcation diagram indicating jammed region for constant .

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

A nonlinear model for parallel folding with friction

This leads to the condition 2 4 Q2 P Q2 + 1 2EI 2 L L

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

C. J. Budd, R. Edmunds and G. W. Hunt

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

A nonlinear model for parallel folding with friction

where F , the total potential energy per unit length, is given by F = EI w2 1 w _2

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

C. J. Budd, R. Edmunds and G. W. Hunt

Integration by parts of this expression yields L Z L 2 L L d @F d @F @F @F w + w _ w + V = d x2 @ w @ w d x @ w @w _ 0 0 0 0

Downloaded from rspa.royalsocietypublishing.org on June 6, 2013

A nonlinear model for parallel folding with friction

Proof . From (4.9) it follows that w= 1 .... (2EI w + P w ): k

Lemma 4.2. The function w depends linearly on the friction.

du d( u) dw ; = = dx dx dx etc. Substitution leads to an identical equation, .... 2EI w + P w + kw = 0;