D Y N A M I C S OF P I L E D R I V I N G
BY THE FINITE ELEMENT METHOD
Ronaldo I. Borja
Department of Civil Engineering
Stanford University
Stanford, C A 94305
ABSTRACT
It is proposed that the dynamics of pile driving based on the one-dimensional
theory of wave propagation be analyzed by the finiteelement method. This approach enables one to continuously interpolatethe displacement, velocity and acceleration profilesthroughout the pile length. In contrast to the finitedifference
technique presented by F,. A. L. Smith [1] and the finiteelement procedure presented by I. M. Smith [2] in which initialconditions are defined on the basis of a
prescribed hammer velocity,the method presented herein defines initialconditions
on the basis of a prescribedimpact force versus time curve at the pile/hammer point
of contact. Applications of the proposed technique to typical pile driving problems
on an elastoplasticsoiland using an implicit time-integrationscheme are discussed
using a numerical example.
INTRODUCTION
4O
criteria (e.g. see [3-5]).
The procedure of lumping masses and springs as implemented in [1,3-5] restricts the user to explicit integration in time (central difference explicit scheme).
In order to avoid numerical instabilities that afflict all explicit time-stepping schemes
when large time steps are used, Smith [2] presented a finite element solution that
employs an implicit time-stepping scheme for solving pile driving problems. Both
methods presented in [1] and [2], however, simulate 'impact' by prescribing a hammet velocity as determined from the properties of the pile driver and using this
velocity to get the numerical solution started.
It has been shown that the pile driving process can be better described by
prescribing an impact force versus time curve which incorporates the interaction
between the pile and hammer at the point of contact, rather than by prescribing
the hammer velocity alone [6]. This paper presents a finite element solution to pile
driving problems in which the impact forces at the pile/hammer point of contact
are prescribed in the solution, rather than the hammer velocity.
The resulting semi-discrete equation of motion is solved by Newmark's (implicit) predictor-multicorrector method for nonlinear equations [7]. With this implicit method, numerical instabilities are avoided albeit large time steps are employed in the solution. The additional cost engendered by the implicit method over
the explicit method used in [1,3-5] is insignificant in one-dimensional wave propagation analyses since the resulting coefficient matrix in the implicit method is usually
symmetric and banded (mean half-bandwidth = 2 for one-dimensional problems).
The method presented in this paper is applicable to any type of soil loaddisplacement behaviour such as linear, nonlinear, or elasto-plastic. As an example,
a numerical problem in which the prescribed impact forces are computed from the
method presented by Fairhurst [6] is discussed to demonstrate the applicability of
the proposed method to typical pile driving problems.
GOVERNING EQUATIONS
Consider the pile setup and the free body diagram sketched in Figure 1. For
dynamic equilibrium of a typical slice of pile in Figure lb, the following equations
should be satisfied:
P2+f(t)=O
in fl
(1)
41
",(/)
P
..,.
~f'Adz
T
,.,~,t;
P+P,, az
i -PILE
(b)
e(e')
AXIS
(o)
(c)
at z = z h
= Vp at z = zp
(2)
(3)
(4)
u = uo at t = to = 0 (initial condition)
(5)
(6)
where z is the vertical coordinate (depth) over the domain n representing the entire
length L of the pile including its accessories (capblock, pile cap, cushion block, etc);
t is time; u = u(z, t) is the displacement profile at any time instant t; ~ and ~ are
the first and second time derivatives of u, respectively; u0 and ~o are prescribed
initial displacement and velocity, profiles, respectively; P -- P ( z , t) is the axial force
in the pile at any time (compressive force assumed positive); Ph = Pa(t) is the
prescribed impact force-time curve at elevation z = zs corresponding to the pile's
point of contact with the driving hammer (usually at the pile butt); P~ = P~(u, ~)
is the soft bearing reaction against the pile tip at depth z = zp; p is mass density
of the pile material; "y is the unit weight of the pile material (= gp, where g is the
42
gravity acceleration constant); 0 is the pile perimeter; A is the pile cross-sectional
area (both 0 and A may be functions of z); r is the soil's shear resistance per unit
surface area in contact with the pile, and the comma denotes a differentiation.
If the pile tip is fixed on a firm bedrock, equation (4) is replaced with the
following boundary condition:
at z = z p .
u=~=fi=0
(7)
Equations (1)-(7) do not strictly satisfy the equation of motion for longitudinal
wave propagation problems in three dimensions [8]. However, they represent a much
simpler approximation theory of wave propagation in one dimension in which each
slice of the pile is regarded as either in simple tension or compression.
Let the axial force be given by
-EA
(8a)
OOz '
where E is the pile's modulus of elasticity while the negative sign implies that a
positive P is a compressive force. If the compression waves axe transmitted through
a capblock material whose stress-strain characteristics axe as shown in Fig. 2, equation (8a) may be replaced with
r = f(l + Js~)
a = #(1+4~p)
(9)
(10)
where Js and Jp are, respectively, the soil's surface and point damping constants
in the sense of [1], ~ and @ are the static side and pile tip resistance per unit
contact area obtained from the static load-deformation relationships for the soil,
and ~p = ~(z = zp) is the velomty of the pile tip.
The variational expression equivalent to equations (1)-(10) can then be written
thus:
n w(P,, +
f)dn
= O,
(11)
43
where w is the virtual displacement.
nOw
Ou d[l + wl,ffi,, Ph + wl,=,,Pp
-~zEA~z
+ [ w(pAii + v0 Jt~
7A) dfl = 0,
(12)
or~
(13)
A similar formulation
can be used with equation (85). In equations (12) and (13), the virtual displacement
w must be consistent with constraints imposed on the body. Thus for a pile whose
tip is fixed on a firm base,
(13)
equation (7).
problem:
Man+l + p(vn+l,d~+l) = fn+l,
(14)
M =/nNTpANdfl;
(15)
fn+l is the external force vector representing the contributions of the pile's selfweight and the prescribed impact force evaluated at the same time station tn+l,
thus:
(16)
4-/BT(EA)B
44
the nodal acceleration vector. In equation (15), M may be assumed constant for
a given pile. Selective numerical integration of this matrix at the nodes results in
mass lumping and M becoming diagonal. In equations (16) and (17), ih and ip
are column vectors with unit elements in the rows corresponding to the equation
numbers for nodes h and p, respectively, and zero elements elsewhere. Note that for
a pile whose tip is fixed on a firm base, the term involving ip drops out of equation
(17).
TRANSIENT ALGORITHM
Equation (14) holds for both linear and non-linear (or elasto-plastic) soil stressstrain behaviour and can be discretized with respect to time by employing an implicit version of Newmark's integration scheme for a second order equation [7]. The
algorithm employed herein consists of a predictor phase, followed by a multi-pass
corrector phase defined by
K.A.dlq = ~[i],
(18)
K* = M/(AI2/~) + ,~Cn_]_l/(Al~)
tan
~L xct-n
"~'n-I-I
(19)
where
Ma~l+l
--
"vIil
A[i] ~
v~,
n+l,'n+l/
(20)
is the residual force vector of the ith iteration, and Ad[q is the search direction of
the ith iteration. The corrector phase is repeated until the residual of the matrix
equation is driven close to zero. See Owen and Hinton [9] for a summary of the
main features of this algorithm.
The algorithm shown in equations (18-19) contains Newmark's integration parameters 7 and/~. For stability, 7 >- 1/2 and ~ = i / 4 ; for second-order accuracy
7 = 1/2 [10]. In equation (19) At = tn+l - tn is the time step which governs the
accuracy of the solution.
Defining Pn+l = p(v,~+l, dn+l), the tangential viscosity matrix at future time
station tn+l in equation (19) is evaluated, thus:
ct&n
Opn+:t = / n NT~n+ISJ, N dfl + (@'n+IApJp) i v ,
n + l ----~
(21)
45
where ip is a square matrix containing an element of unity in the diagonal slot
corresponding to the equation number for node p, and zero elements elsewhere.
The tangential stiffness matrix is given by:
gt,n
OPn+l= / NTOfa+lo(1.bj,~n+l)Nda~_/flBT(EA)Bd n
n+l = 0dn+l
+ ~A,(1
Jpvpl,.+1)
(22)
:n the numerical evaluation of equations (21) and (22), the stresses, stress gradients
and velocity profile/~n+l axe evaluated at the Gauss integration points.
The time-stepping scheme of equation (14) may be initiated by observing that
at time t = to = 0, both do and v o are known from (5) and (6). The initial
acceleration vector ao can then be computed from (14) thus:
ao = M -1 (fo - p(Vo, do))
(23)
NUMERICAL
EXAMPLE
In the following example, an elastic pile is partially embedded in an elasticperfectly plastic soil medium. Consider the following data for a typical pile driving
problem (refer to Figure 2):
1. Hammer Type: Vulcan Number 1; rated energy = t5 ft-k (20.33 kN-m);
hammer efficiency (fraction of nominal potential energy effectively transmitted at
the pile butt) = 64%,
2. Pile Type: HP 12x53; total length L = 1,200 in (30.49 m); embedded
length Le = 960 in (24.39 m); cross sectional area A = 15.6 sq.in. (100 sq.cm.);
mass density p = 7.3 x 10 -4 lb-sec2/in 4 (78,000 kg/m3); modulus E = 30 x 10s psi
(2 105 MPa).
3. Capblock Type: Micarta; height /)c = 12 in (0.30 m); area = Ac = 100
sq.in. (645 sq.cm.); mass density Pc = 9 x 10 -5 lb-sec2/in 4 (9,700 kg/mS); unit
weight 7c = 0.035 lb/in s (95 kN/mS); modulus
Ec =
46
CAPII.OCKI/RAil
I F~rM.~...I.~j
\ [~ILECAP
ITRI[II
ITREII
*"VI
-T
PILE~.~.~o WT(I.IOI*)
8TItA|N
I,
14)O
PT
PILE
ITItAIN
I
CAPBLOCK
lINEARITRI[I$
~L_~IUAKE
I
Ol[ABlltITIt[ll
(I :i,,~T)
/ IIIPL.
81DERESISTANCE POINTRI-rIIITANCE
PROTOTYPE
F.E. MESH
(b)
(o)
FIGURE
SOIL
a. Side Resistance: typical value per inch (meter) length of bounding area
of pile for skin friction [11] f0 -- 100 Ib/in (17.49 kN/meter); total ultimate side
resistance assuming a uniform distribution, Tu -- 100,000 Ib (445 kN); assumed
displacement to cause initialyielding ('quake' in the sense of [1]), Q, -- 0.10 in
(2.54 ram); assumed viscous damping constant, J, -- 0.05.
b. Point Resistance: ultimate resistance assumed to be 50% of ultimate side
resistance, Pu -- 50,000 Ib (223 kN); assumed displacement to cause initialyielding
(quake), Q~ = 0.10 in (2.54 ram); assumed viscous damping constant, Jr = 0.15.
5. Prescribed Stress Pulse at the Pile Butt: The (assumed) instantaneous
impact versus time curve as the hammer strikes the capblock to drive the pile is
shown in Fig. 3 [6]. The (compremive) impact stressgenerated as shown in Fig. 3 is
the result of wave transmiuions and reflectionsalong both the rod and the hammer.
A piece-wiN linear finitedement input function with a tension cut-offis also shown
in this figure. The tension cut-offrepresents the instant at which the rod starts to
47
PILE TYPE: HPIEX6B
86
b~
w
~,.
F(1
Io
~-0
HAMMERT~cANNOI
8O
16
l0
TEHIOn
C
UT/ -OFF
'T~ES.*UETo
, . . , , .YATC wE,..T~
__ ~ - ~ .
o
o
IO
tS
TIME, MILLISEC.
CONCLUDING REMARKS
A finite element numerical procedure for analyzing the dynamics of pile driving
by the one-dimensional theory of wave propagation is presented. It is proposed that
the impact force-thne curve be prescribed in the solution, rather than the ram velocity, to better describe the pile/hammer interaction during the pile driving process
[6]. Without sigu~cantly engendering additional CPU costs over the method based
48
TOTAL
DISPL.
(*.lo m~ I
/Z.~
8 0 ~T
( t . n m)
II iROONO
IC
II
--
'i
POINT R E | J I l T ,
Pew llO,O00 I.II
(ttt KtI
/u
4 MSEC.
4' s IIMSEC.
d a ISMSEC.
~'- 2S MSEC.
DISPLACEMENT
SCALE :
49
Program, U. S. Dept. of Transportation, Vols. 1-4 (1976).
[6] Fairhurst, C., Wave mechanics of percussive drilling, Mine ~ Quarry Engineering (1961) 122-130.
[7] Newmark, N. M., A method of computation for structural dyvamlcs, J. Engrg.
Mech. Div., A.S.C.E. (1959) 67-94.
[8] Timoshenko, S. P. & Goodier, J. N., Theory of Elasticity, 3rd edn., McGrawHill, New York (1970).
[9] Owen, D. R. J. and Hinton, E., Finite Elements in Plasticity: Theory and
Practice, Pineridge Press Ltd., Swansea, U. K. (1980).
[10] Hughes, T. J. R., Analysis of transient algorithms with particular reference
to stability behavior. Ch. 2: Computational Methods in Mechanics 1: Computational Methods for Transient Analysis, Ed. by T. Belytschko & T. J. R.
Hughes, North Holland Press (1983) 67-155.
[11] CheUis, It. D., Pile Foundations, 2nd ed., McGraw-Hill, New York (1961).
Received 23 October 1987; revised version received 5 March 1 988; accepted 7 March
1988