i
ARODLEMS IN WAVE PROPAGATION IN SOIL AUD ROCK
: by
Nathan M. Newmark, Head EERO LIBRARY
Department of Civil Engineering sor Shae
University of Ilinois Richmond Caliornie Suene
for
SYMPOSIUM ON WAVE PROPAGATION AND OYNAMIC PROPERTIES
‘OF EARTH MATERIALS
23 august 1967
1. INTRODUCTION
The transmission of waves in soil and rock is important in connection
with the propagation of disturbances arising from machinery vibration, blasting,
and earthquakes. ty interest as an engineer is primarily in the effects trans-
mitted to relatively short distances from the source of the disturbance, in
contrast to the interest of the seismologist in disturbances transmitted to
very large distances fron the source. | shall further limit the discussion
of this paper to only those disturbances that are transmitted through the “—~
ground, eliminating fron the discussion the air shock effects from blasts.
Whatever the source of the disturbance, i-e,, vibration, high
explosive or nuclear contained detonation, or earthauake, waves of several
different types are generated and propagated through the aediun and along the
surface. Such waves generally are attenuated as they travel, and change in
shape and in duration with distance. High intensities of stress, either near
the source or at sone distances from the source, may cause failures In the
soil or rock, particularly when these high intensities of stress encounter
pockets or strata of material with low resistance.
Qur knowledge of the characteristics of wave propagation in actual
garth moterials depends primarily on measurements, Such measurements areinfluenced by the characteristics of the instruments used to make the meosure-
ments. For most engineering purposes, particularly in the design of structures
or of cavities in the ground, we are interested in the maximum intensities of
motion as measured by the maximum particle displacements in the ground, the
maximum particle velocities, and the maximum particle accelerations. We are
interested in these maximum values primarily in the free field, although our
concern is also with alterations or modifications in the free field phenomena
caused by the placement of a structure on or in the soil, or the placement of
2 cavity or tunnel in the soil.
Consideration must be given to the effect of layering or stratifi-
cation of the soil or rock, and of possible channeling of the transmitted
energy in particular strata.
Dams and embankments have particular susceptibility to dynamic motions
transmitted through their foundation; failures of embankments can be caused by
the velocities and accelerations transmitted to the material composing them.
In the course of the discussion, | shal! have occasion to discuss
certain bounds and limits on the relative values of maximum acceleration,
velocity, and displacement For ground motions, and | should like to present
some views on so-called "scaling! of the relations for these motions in terms
of the energy of the source, the properties of the material, and the distance.
Finally, | will consider some of the aspects of numerical calculations
of wave propaga
n, with particular attention to problems having to do with the
effects of boundaries, either natural or artificially implied in the process of
calculation.2, NATURE OF PHENOMENA
The various types of disturbance produce somewhat different wave
propagation effects. The nearly steady state vibration from machinery, for
@ particular maximum intensity of motion, may be more serious in its effects
itself than the
on adjacent structures or on the possible failure of the s
same maximum intensity of disturbance froma transient source such as blast
or earthquake. In general, blast generated waves arise from e relatively
compact source, as contrasted with earthquake generated waves, which arise
fron a relatively diffuse source. Hence the phenomena at relatively close
distances for these two effects may be very much different, but for relatively
large distances they may be almost indistinguishable in their nature.
In either case, waves of dilatation (compression and tension),
shear, or surface waves are generated by all of the sources mentioned. i late-
tional waves intersecting @ boundary, particularly a free boundary, generate
shear waves upon reflection, and surface waves or waves similar to surface waves
are generated at boundaries between strata of different properties. These
various types of waves travel with different velocities; hence the combined
motions change markedly with distance from the source, even where there is
no effect of attenuation or energy absorption because of the properties of the
material.
If the shearing stress transmitted through the medium reaches 2 point
where the shearing strength is lower than the stress, failure by slip can occur,
nfluence
lar to fault motions. Such failures, of course,
causing motions
the propagation of stress beyond the point considered and in general absorb
energy much more effectively than do other types of deformation of the soil
or rock.as
Tensile stresses are propagated through sol! or rock, perhaps
arising from the nature of the disturbance, but also arising from reflection
of compressive stresses at a free boundary. Tensile stresses exceeding the
h also
tensile strength of the soil or rock cause fractures or failures wi
absorb energy and change the pattern of distribution beyond the point
considered.
Shear failures can occur on any plane on which the ghearing stress
exceeds the shearing resistance. An illustration of 2 possible type of shear
failure corresponding to a vertical faulting is indicated in Fig. 1. Such
motions may also be accompanied by failures along horizontal planes at the
sane tine.
Tensile failures in soil or rock can arise in various ways. Tensile
stresses can be propagated in regions where the gravity stress produces initial
Conpressions which, when combined with the natural strength of the medium,
exceed the magnitude of the tensile stress that is propagated. tiowever, where
such tensile stresses are propagated into a region where the combined strength
and gravity stress is less than the propagated stress, @ fracture or a parting
can occur with consequent absorption of energy and sudden attenuation of the
stress,
Tensile stresses also occur owing to the reflection of compression
waves from a free surface. Compression waves of stress can be propagated
toward @ horizontal or inclined surface either directly from the source, or
because of reflections from deep stiff layers. When such compressive stresses
approach the surface, as in Fig. 2(a), they are reflected virtually completely,
as indicated in Figs. 2(b) and 2(c). The reflection introduces a tensile stress
equal and opposite to the compressive stress wave. Where the combined stress isoe
tension, and if the combined tension exceeds the strength of the material plus
the gravity compressive stress, parting occurs, as indicated schematically
in Fig. 2(c).
Another type of failure may arise because of liquefaction of soil
under either transient or sustained vibration. Foundation failures under
buildings or dams, and slides in earth or rock-Fill dams ond enbanknents,
may be caused by liquefaction of soils or of sand lenses. (Refs. 1, 2, 3)
An illustration of a possible slide mechanism due to liquefaction of a sand
layer in an earthquake is indicated schematically in Fig. 3.
TYPICAL WAVE FORMS FOR EARTHQUAKE HoT! ONS
Our knowledge of the wave Forms of motion in earth and rock cones
From measurements of various kinds with instruments having a range of charac-
teristics. Unfortunately, a measuring instrument reports its response to the
excitation Tt undergoes, which may be somewhat different from the actual motion
towhich it is subjected. However, in spite of the difficulties, fairly
accurate instrumentation has been developed and a qumber of records have been
obtained of mation in earthquakes, in blasts of various types, and of motions
transmitted by machinery vibration.
The most intense sustained earthquake motion that has been recorded
jn the United States is that for the £) Centro California earthquake of Hay 18,
1980. A record of the motions for the North-South horizontal component in this
earthquake is shown in Fig. 4. The upper part of the figure shows the measured
ground acceleration-time record, which reached a maximun value of 0.329, where
g is the acceleration of gravity. It can be noted that the largest accelera-
tions occurred in the first six seconds of the earthquake, although substantial
accelerations occurred during the entire motion which lasted nearly 39 seconds.See
An integration of these accelerations to give the ground velocity
variation with time is shown also in the figure. The number of oscillations
appear to be less. The maximum ground velocity determined from the integrated
record is 13.7 in/sec, and so far as we are aware it is the maximum that has
been determined for a major sustained earthquake. The Parkfield earthquake
‘of June 28, 1966 had somewhat larger accelerations (approaching 0.59) and
velocities, but apparently corresponded only to a single displacement pulse
out a long duration vibration.
The Integration of acceleration records to obtain ground velocity
theoretically should be reasonably accurate, but because of the large nunber
of oscillations involved, and because of the uncertainty of the location of
the base line, slight errors are introduced in the integration. These are
generally not of serious consequence in the determination of ground velocity.
The errors are cumulative, however, in their effect on further integrations,
and the determination of the ground displacement by integration can be seriously
in error. An integration of the velocity record in Fig. 4 to obtain the groung
n with time is shown also in the lower part of that Figure.
displacement varia
ct to some question but the general
The exact values of displacement are ub
shape of the curve |s yeasonablé, The integration gaye e maximm ¢/splacenent
of 8.3 in. Certainly, some points nearer the fault sustained greater displace-
ments, but possibly no greater accelerations or velocities. It appears that
there were only four or five major oscillations in displacement, as compared
with a substantially greater number of oscillations in velocity or in acceleration.
The response of a simple oscillator with viscous damping, of the type
shown in Fig. 5, when subjected to a motion of its base corresponding to a
nt
transient disturbance such as that in Fig. 4, produces relative disploa
of the mass compared with the ground, or strain In the spring and consequent
force in the spring of the oscillator. Other quantities of interest in the
response are the ma
wm energy stored in the spring, which can be measured
by the maximum pseudo relative velocity of the mass, and the maximum force
in the spring, which can be measured in terms of the maximum pseudo acceleration
of the mass. The relations between these quantities are as follows:
v= wd (ly
Ag =u
In Ec. (1) the quantity u_ is the maximum relative displacement in the spring.
The quantity V is the maximum pseudo relative velocity, and the quantity Ag is
the maximum pseudo acceleration. The symbol w is the circular frequency of
vibration of the system.
A response spectrum, showing the three responses, D, V, and A,
plotted as functions of frequency, can be computed for any excitation.
A typical response spectrum, for a moderate amount of damping, for an earthquake
excitation similar to that of El Centro, is shown in Fig. 6. In this figure,
the nunbered points indicate for specific oscillators, representing particular
structures, simultaneous values of D, V, and A, which give a simple approxi-
mation of the maximum response of the structure represented by the osci} lator.
The typical trapezoidal form of the response spectrum shown in Fig. 6 is
obtained for other types of dynamic excitation, ranging from more complex
earthquake motions to motions corresponding to blast excitation, or to other
jotion is often
types of transient disturbance. The response to a steady state
similar in shape except for the fact that for low amounts of damping, @ much morees
peaked response spectrum at the frequency of the steady state motion is
obtained. A response spectrum for the El centro earthquake motions of Fig. 4
is shown in Fig. 7, plotted in a slightly different way to show the amplifi-
cations of the respanse for each leg of the trapezoid. These amplifications
are due to the partially sustained oscillatory components of the motion.
Longer durations of shaking would give higher amplifications. However, the
amplifications in the El Centro earthquake are typically fairly high, especially
for low values of the relative damping factor, 3.
The form of Fig. 7 has sone implications regarding the response of
simple instruments to transient motions. For example, for very low frequency
oscillators, the maximum displacement is quite accurately determined from the
instrument, for small or even moderate anounts of damping, as shown by the
tendency for the left-hand side of the response spectrum in Fig. 7 to approach
asymptotically the ground displacement line. On the other hand, for very high
Frequencies, the maximum acceleration indicated by the oscillator approaches
asymptotically the maximum acceleration of the graund. However, the flatness
of the response over a wide range of frequencies is greater for high degrees
of damping. Hence a highly damped high-frequency instrunent gives a reasonable
measure of maximum acceleration or of acceleration in general.
On the other hand, a moderately damped velocity gage is required to
avoid large amp ons in the response to maximum ground velocity. In any
ease, sone difference in phasing and shape of curves can be expected because
of the response of the oscillator, which can aot completely correspond with
the motions it is intended to measure. A means of overcoming this is available,
by using 2 simple oscillatory system to make the measurements, and computing
From the response the input that produced such a response. This can be doneee
by use of 2 computer (and determinations can be made more precisely of the
input motions under these conditions than can be determined by direct
measurement).
As an indication of the nature of the ground motions measured in
explosions, Fig. 8 is an example of the surface motion in Gnome, taken from
Ref. 4. It is readily apparent that the number of oscillations in acceleration
is considerably less than for the El Centro earthquake motion. However, at
greater distances, the characteristics of the motions from explosions become
more nearly like those from earthquakes, The high intensity motions near the
source are attenuated quite rapidly, and the high frequency components of the
transient oscillations are damped more rapidly than the low frequency components.
Moreover, because of multiple reflections fran deep layers or from discontinuities
in the soil, further oscillations are introduced and the entire duration of the
motion is increased. Hence, the ground motions at very large distances from
the source of an explosion become quite similar to those for an earthquaxe.
Similarly, the ground motions very close to a fault for an earthquake may be
much more like those for an explosion, end probably last for a much shorter
time than the earthquake motions at larger distances.
The rates of attenuation of intensity of motion, and the distances
at which large intensities are felt, are dependent on the amount of eneray
involved at the source, whether the source be an earthquake or an explosive
blast. Later in this paper there are considered some of the relations which
govern the attenuation mechanisris and rates, as indicated by a crude dimensional
analysis. It is sufficient to state here that these attenuation effects are
dependent not only on the energy involved but also on the properties of the
material, and especially the density and seismic velocity or the stiffness ofthe material. In addition, attenuation is caused by energy losses corresponding
to both inelastic or viscous characteristics of the material. And finally,
major energy losses occur because of such mechanisms as slip or tensile frac
tures, as discussed in foregoing sections of this paper.
Either for the purpose of interpreting empirical results, or for the
Purpose of making theoratical and numerical calculations of wave propagation
in earthlike materials, the physical properties of the material must be known
or must be stated in a relatively simple form. The stress-strain-time relations
for soil and rock are complex, and these "constitutive relations" are inade-
quately determined at present, although a great deal of effort is being spent
in studies of the properties of soil and rock under both static and dynamic
loading. The difficulties in determining the constitutive relations arise
because of various factors including especially the following: the large
variation in properties from point to point even in a relatively homogeneous
deposit of material; the change in properties of the materials when removed
From the site for the purpose of testing in a laboratory; the effect of water,
either partially or completely saturating the material; the effect of the complex
stress and constrained deformation states which govern the actual behavior as
compared with the relatively simple stress states which it is possible to intro-
duce in experiments; and possibly, of most importance, the change in properties
of the material with change in intensity and duration of stress as well as
stress or motion history. for example, relatively intense transient motions
can be transmitted through a sand lens, whereas a much less intense sustained
vibration may cause liquefaction of the lens with change in its properties by
orders of magnitude.
(One of the most proaising means of determining soil and rock properties,
that could be most useful in theoretical studies of wave transmission, is use of
}ae
measurements in situ. However, practical difficulties exist in such measurements
especially at even moderate depths, in maintaining the degree of constraint and
the ambient pressures associated with the undisturbed material. Nevertheless,
appropriate measurements in situ, combined with laboratory studies of various
kinds, can lead to a better understanding of the constitutive relations for use
in wave propagation studies.
It must be emphasized that our interest extends beyond the elastic
range to the conditions corresponding to failure, in order that we can have
fan appropriate understanding of the energy absorbing mechanisms that govern
the attenuation of intensity of the motions and stresses transmitted. Moreover,
the small energy absorbing mechanisms that govern the phenomena of viscous
behavior and slight degrees of plastic behavior have an important influence on
the attenuation mechanisms that are effective at relatively large distances from
the source. Even slight energy absorption can affect the results of theoretical
calculations to a major extent. At present, agreement between measurements and
calculations can only be regarded as a fortuitous matter and should not be
interpreted as a confirmation of particular constitutive relations.
CONSISTENT BOUNDS ON HAXIMUM MOTIONS
When a point in a medium moves in a certain direction with time, the
maximum values of its acceleration, velocity, and displacement must be consistent.
A derivation is given here of upper and lower bounds on the maximum velocity
relative to the maximum acceleration and displacenent. These relations are
useful in verifying the accuracy of measured quantities, and in defining the
form of equations relating these maximum values to the various parameters
describing the mediun and the released encroy. The discussion herein is taken
from atia
The notation used in the derivation is defined
Fig. -9, which is
a plot of consistent acceleration~, velocity-, and displacement-time relations
for a given point in the medium. The maximum values of displacement y,
velocity ¥, or acceleration ¥ are denoted by u,, & respectively. The
ne Ee
minimum values are correspondingly denoted by u, U_, andi. The minimum
Weluby pees note neenaddi ky ban epani Walle ue EUeLE sire lap lacananta jth
minimum velocity must be either zero or negative. For initial conditions of
rest, the minimum displacement must be zero (or negative).
We shall designate the greater numerical value of u, or u_ by U,
12x
and the lesser value by U.,., and corresponding designations will be used for
Gey
irs It is noted that U.., etc., may designate the value
Vain? Umin®
of u_
(aa
s larger numerically than u,, ete.
We shall designate the maximum increments in y, J, and ¥ by the
synbol: d,, 4. and G,, and the manioun decroncnts by @_, 4, aod d, the
fs
greater numerical value of d, and d_ is denoted by 0. and the lesser by
qr and sinilarly for B and B.
The time between successive zeros, in the loop where the maximum
numerical value occurs of displacement, velocity, or acceleration, is denated
by H, A, or
y respectively. +
The following relations are obvious, based on the relations defined
in the foregoing:
i
D,
min? = max
For the further derivations, we require the sketch in Fig. 19 of
the loop having the largest value of G, plotted as if U... were positive.
Lines of slope +U,., and -U_,, fron the maximum point will intersect theSee
time axis conpletely within the loop, because no slope of the curve of U is
greater than U,,,. On one side of the triangle formed by these two stones,
a line with slope
Inin Will also Tie completely within the loop, since one
of the sloping sides has no slope greater than U..
The quantities t, and t) In Fig. 10 are given by the relations:
(3)
It follows immediately that
(4)
Because U. SU, it is also true that
min = “nox
or
age 20d A (5)
Similarly one can derive the relation
Soe et (6)
from a diagram similar to Fig. 10, applicable to the maximum displacement
ordinate, Equations (5) and (6) give upper and lower bounds for the maximum
yelocity, but these bounds involve a time interval.
Another bound on velocity can be derived from the fact that the
entire area of the loop in Fig. 19 must be greater than the area of thetriangle within the loop, and must be less than the maximum value of the
displacement increment, since successive zeros on the velocity-time curve
correspond to successive maxima and minima on the displacement curve
7)
since U SU
Inin = Umax
By combining Eqs. (2) and (7) a series of inequal
ies Is obtained.
where Un. and Unjq are nearly equal, the middle term in Eq. (7) can be
considered to be so nearly equal to the right-hand term that it is omitted
from the summary expression following
Zu) (8)
From the end expressions in Eq- (8) one finds final
de <2u : (2)
Imax = ? Ymax Ynax
This expression holds for all cases, and in independent of time interval, etc.
The upper limit in Eq, (9) can actually be reached, but only under
somewhat unusual conditions. Consider, for example, the situation portrayed
in Fig. 11, where constant rectangular pulses of acceleration are indefinitely
Cones (te pate 2 ons ieera co bare as slow) th an arrropr (ate
negative velocity, the maximum and minimum quantities are all numerically
equal. As indicated in the figure, if for o tine Interval of h, infinitely
repeatedones
For repeated sinusoidal motions, the coefficient becomes unity. For use in
blast motion problems, it is suggested that the factor ranges from about 1/2
to about 1/5 for close-in and for intermediate distances, and for very large
distances, Eq. (11) might be used.
5. RELATIVE MOTI ONS
It is often of importance to have information about the relative
motions of two points either close together or some moderate distance apart,
in connection with the design of structures supported on or in soil or rock.
Information concerning relative motions, as determined from measurements, is
Often difficult to interpret or to assess. However, where it is fairly clear
that 9 wave motion is propagated in one direction without interference with
other waves in other directions, and where the change in shape of the wave fron
point to point is relatively small, one can make inferences about the relative
motions between nearby points in a fairly simple manner, as outlined here. For
example, consider two points, point | and point 2, at a distance bM apart, as
shown in Fis. 12. Consider a displacement P at point | and P plus an increment
at point 2, as indicated, where the second term of the increment involving the
second derivative of ® with respect to x is important only if b is relatively
large. Now let us consider a situation where a wave is propagated from point |
towards point 2, with a displacement in the form given by:
p= F(x-ct) (2)
in which ¢ is the velocity of this particular wave propagation and t is the
time. Then the various derivatives of the displacement P with respect to x
and t are given by the fallawing relations48 «
These relations, in Eqs- (18) and (19), are often of use in deter-
mining the maximum strain that must be experienced by an element extending
over sone distance, or the maximum curvature in an element, as for example
a tunnel, pipeline, or other structure.
Where the distance b is relatively long, the change in length Sb of
the origins] distance b between points | and Z is given by substituting the
appropriate expressions fron Eos. (17) into the increment in displacement in
Fig. 12. When ® is in the direction of x, one derives:
2
toe (20)
In most instances, only the first term on the right-hand side of Eq. (20) is
important, but this is not necessarily true in general.
Other relationships are of importance in the case where the mations
are caused by more general disturbances than a wave of nearly constant shane
transmitted in one direction. for example, it 15 apparent that the maximum
change in the distance between paints 1 and 2, Bb,,, is related to the maximum
displacements at points 2 and } in the following way:
Peat = mak 2” ‘nex 1 te
in many instances, this relation may be trivial because the maximum displacements
may be nearly equal, but since they do not occur at the same time, it is obvious
that the maximum transient change in length must be greater than the difference
in the maximum displacements. It is, however, true that the maximum change in
length is less than the: difference between the maximum displacement at either
point | or point 2, less the minimum displacement, or the displacement in the |
opposite direction, at the othe? point. The minimum displacement would of course