Sei sulla pagina 1di 7

Ahlvin

SOME PAVEMENT TESTING CONCERNS TO RECALL

By:
Richard G. Ahlvin, PE
Consulting Civil Engineer

PRESENTED FOR THE 1999 FEDERAL AVIATION ADMINISTRATION


TECHNOLOGY TRANSFER CONFERENCE

04/99
Ahlvin 1

SOME PAVEMENT TESTING CONCERNS TO RECALL

By R.G. Ahlvin, PE 1

The need for methods by which pavements could be designed to support heavier aircraft
began, essentially, with WW-II. At the time our heaviest aircraft was the 30,000 lb. weight range
Douglas DC-3, or, militarily, the C-47 transport. Wheel loads were a nominal 15 kips. This was
in the range of heavy truck loads and highway pavement design methods served for airfield
pavement design.
The war brought the B-17 and B-24 bombers, over twice as heavy as the C-47, with wheel
loads twice the heaviest truck loads. No established design methods existed to provide for these
greater loads. Accordingly studies were begun to fill this need for heavier pavement design.
Highway design methods were greatly extrapolated by the best available pavement experts,
and the resulting criteria were verified by accelerated traffic tests of full-scale pavements.
Subsequently pavement loads were doubled again and again, but the need for large increases
in structure was avoided by doubling supporting wheels:

Aircraft Strut Load Range, kips Wheels per Strut


C-47 15 1
B-17 37 1
B-29 75 2
B-36 150 4
Boeing B-747 300 8

1
Consulting Civil Engineer, 4 Shadow Wood Drive, Vicksburg, MS 39180
Ahlvin 2

Tire pressures experienced a somewhat parallel increase:

Aircraft Tire Pressure, psi


C-47 80
Current Heavy 230
Fighter 250
Navy Fighter 320

Structural design of pavements for the large increase in load support over highway needs
dictated a need to model pavements mathematically and determine limiting stresses induced by
the loads. These induced stresses would be compared to resisting strength as a basis for design
methods.
Unfortunately, technology at that time did not permit determination of either the stresses
induced by a load or the resisting strength of the pavement structure. It was decided that means
for developing a capability to evaluate induced stresses or the capacity of a pavement structure to
resist such stresses would be unlikely to be devised within the time of need for pavement design.
It was, thus, necessary to proceed with development of largely empirical, experience based,
pavement design methods. The resulting methods came into common use and to a large extent
continue in use today.
Despite the need to develop empirical pavement methods, efforts continued toward devising
more rational means. Concepts for more rational methods, while nominally similar for flexible
and rigid pavements, were initially not the same. These comments will refer to considerations for
flexible pavements.
Initially we had only the single-layer (half-space) model to analytically represent a
pavement. And mathematically we had only the Boussinesq point-load solution for determining
induced stresses or strains at points within the pavement model. We did, by then, have the
cleverly devised Newmark Charts for determining stresses induced by a uniform pressure load
on the surface. These charts were devised using a theory of reciprocity to invert the Boussinesq

Figure 1 Basis of Newmark charts.


Ahlvin 3

point-load solution to provide influence elements at the surface inducing an increment of stress at
a point below. There was no basis for evaluating a stress resistance capacity of pavement layer
materials.
The stresses from Newmark Charts involved the assumption of elastic behavior. Before long
closed form solutions were developed for circular areas having the load uniformly distributed
over the area. This was a large improvement over the need to count squares on Newmark Charts.
In following years research examined elliptic tire areas, studied inelastic behavior, and
nonuniform load distributions, but no widely accepted improvements emerged from these.
Initially we had only slide rules and mechanical calculators, and only in fairly recent times
has the computer capacity to freely work with the elastic layered models emerged. We now also
have the capability to work with 3-D finite element models. Using these, work is underway with
stress-dependent stiffness (E-modulus) response of individual elements, but the work has not yet
matured to effective methods ready for wide application. These efforts toward improved models,
better mathematical methods, and the needed computer capabilities need to continue, and we
need to take every advantage of the truly great accomplishments already or soon to be made.
Unfortunately these tremendous advances over early capability seem to have bread the
feeling that we are close to a total functional solution to the pavement behavior problems. The
ease with which we can now formulate computer models of layered pavements and calculate
stress, strain, and deflection does contribute to the final solution idea. This, however, is really
only one major part of the overall problem. Even the layered pavement model we can now treat
has need of refinements relative to E-modulus, layer interface treatments, empirical relation
between some computable strain (stress or deflection) to pavement distress, etc.
The intent of this presentation is consideration some significant needs of pavement behavior
technology which remain as a challenge.
An essential to load support type pavement design is a representation (likely by model) of
the distress or failure actually involved and the providing of strength sufficient to resist.
Pavement distress is represented by cracking and formation of rut-like depressions; commonly
considered to result from internal shearing or some type of accumulating fatigue. No means of
modeling or representing this distress has so far been presented or suggested by research. There
has also been no means for evaluating the probable shear forces on possible shear paths
(surfaces) or to determination of the effective shear resistance.
We do now have the long sought elastic layered pavement model. Using the model we can
compute stress, strain, and deflection induced by a surface load and for any location within the
pavement. There has been some success in relating the maximum computed vertical strain at the
subgrade level to pavement performance. This correlation of vertical strain to pavement behavior
is empirical, experience based. There is no direct relation of calculated strain to observable
aspects of pavement distress. Note that this process involves accepting that a single magnitude of
a calculated parameter using a layered pavement model can project the total performance of a
real pavement subject to mixed traffic for an entire use-life. Note also that theory indicates the
induced vertical strain at subgrade level should be very different for thin structures on strong
subgrades than for thick structures on weak subgrades. Some work attempts to relate
performance to a variety of magnitudes of vertical strain depending on subgrade strength or total
Ahlvin 4

structure thickness. It might easily be argued that some correlation of pavement behavior with
maximum shear would be more satisfying in concept, but no such relations have been put forth.
The elastic layered pavement model and correlations between calculated maximum induced
vertical strain and performance is a powerful tool. We should continue to make full use of it. But
we must also not assume unproven aspects of the correlation. The correlation has been
established between computed vertical strain and total performance, but the correlation has not
been shown to relate to other elements of behavior. For instance, to use vertical strain as an
indicator of overlapping effects of adjacent wheels on performance involves an unproven
assumption. It might be true, and some argue it is likely true, but it remains unproven. Vertical
deflection somewhat akin to vertical strain has been employed to account for effects of
adjacent wheel interaction. This practice was based on research analysis concerning two to four
wheels, but with warnings of some overconservatism which appeared likely to become too great
when larger groups of wheels are involved. And this has since been found to be true.
While we can now model layered pavements and calculate the indicated induced stresses,
strains, and deflections, past research studies have shown that the calculated stresses and
deflections do not well represent equivalent values measured in-situ. Here again, our tremendous
progress toward layered pavement modeling and stress, strain calculation ability seems to have
led to acceptance that the computed values determined do represent those actually induced.
Studies comparing calculated stresses with measured stresses in actual pavements or full-
scale test sections are quite limited. Also the majority of these only undertake to measure vertical
stresses. Vertical stresses respond directly as distributed load. They are not (much) impacted by
effects of Poissons Ratio, layer interface disruptions, or other lateral irregularities. Some

Figure 2 Examples of computed versus measured stresses (USAE-WES-TM3-323 #4).


Ahlvin 5

computed to measured vertical stress comparisons have been considered to show reasonable
agreement. Where lateral stresses have been measured quite large differences between calculated
and measured values have been found. Such discrepancies, of course, impact also on shear
stresses, major principal stresses, and the equivalent strains.
Applications of elastic layer modeling of pavements requires determining a stiffness
modulus (E-modulus) for use in each layer. Determination of suitable modulii for use in
computations remains a not well established procedure. There are three methods commonly
employed to arrive at E-modulus but each has some short coming. First is use of resilient
modulus testing of laboratory specimens. Representative specimen characteristics, confining
stresses, and loading patterns permit a variety of results from which one must be employed. A
second method is use of an empirical correlation between CBR and E-modulus. The most
common value results from 1500xCBR, psi (about 10xCBR metric, MPa), but the research which
established this relation found a fairly wide range in the constant. The 1500 is commonly used as
a reasonably representative value. A third means for determining the modulus of pavement layers
is by back-calculation using an elastic layered model. It is necessary for this purpose to have
measured surface deflections defining a deflection basin for an existing pavement subject to a
surface load. Common practice is to employ NDT (non-destructive test) methods to load and
measure deflections in the pavement. Trial modulii for the layers are adjusted until the measured
deflections (basin) pattern is reproduced. It is, unfortunately, fairly common for this process not
to lead to a set of reasonable E-modulii, but when it does the values appear to be preferable to
others. When more than one of these three methods are employed the indicated layer modulii are
not found to well agree.
There is a fourth method which has been devised for stiffness response of pavement layers.
This involves determination of wave propagation rates within the layer. The wave velocity is
simply related in theory to the elastic modulus. Apparently because strains are very small the
determined modulii represent a too steep stress-strain slope (modulus) which does not provide
reasonable modulii for computations. Some research has attempted to provide a reduction factor
of some kind to yield more reasonable values, but no standard methods have resulted.
A very significant further aspect of rational projection of expected pavement performance is
the treatment of the load repetitions resulting from user traffic. It has become well established
that a pavement structure must be designed to sustain a combination of load magnitude and load
repetitions. A large load applying few repetitions can have pavement support requirements the
same as a smaller load at many repetitions. In contemplating mixed load magnitudes for a variety
of aircraft and loading conditions, wheel load position varying due to aircraft wander and to tread
spacing, load spread widths atop various layers, overlapping effects of adjacent wheels, induced
and relieved residual stresses due to load applications, it is obvious that the load repetitions
aspect of pavement behavior is a very complex problem. The redeeming feature is that effects of
repeated loads impacts as the log of the number of repetitions, and thus great precision has not
been required.
The methods we have devised for treating load repetitions from mixed traffic are very
simplistic and crude. Because great precision has not been necessary the methods in use are
serving, but before a quite rational means for airfield pavement design or evaluation can mature
some much better ways of treating accumulating load repetitions is needed.
Ahlvin 6

These shortcomings in relation to our flexible pavement design capability should be


recognized and continued to be studied and treated in the near future. We do need to combat the
tendency to consider that our significant past progress has brought us to a near total solution to
the pavement behavior problem.

Potrebbero piacerti anche