NDT&E International 36 (2003) 145–156

www.elsevier.com/locate/ndteint

Evaluation of railway trackbed and formation: a case study

M. Brougha,*, A. Stirlingb, G. Ghataorab, K. Madelinb
a
Scott Wilson Pavement Engineering, 9/10 Faraday Building, Nottingham Science and Technology Park, University of Boulevard,
Nottingham NG7 2QP, UK
b
School of Civil Engineering, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK

Abstract
To determine the causes of poor track geometry, a trackbed investigation should comprise a full assessment of ballast, sub-ballast and
formation condition. If information about any of these layers is omitted, the true cause of poor formation sites cannot be ascertained
satisfactorily and consequently appropriate ground improvement schemes may not be implemented. This has been observed in the UK rail
network with examples of a lack of detailed site investigation and subsequent expensive remedial work being carried out without eliminating
the initial cause of the problem. This paper summarises the trackbed investigation performed at a site requiring high maintenance, suspected
as being caused by poor formation. The paper concludes with an analysis of ground conditions and possible choice of remedial schemes.
q 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Ballast; Formation; Ground penetrating radar; Falling weight deflectometer; Railway trackbed

1. Introduction substructure problems based on investigation methods and

economic evaluations [4].
Trackbed investigation is considered to be crucial for Hunt [5] has confirmed track quality,1 continued track
assessing potential maintenance requirement and sub- component performance and subsequent maintenance are
sequent design of any remedial measures. In reality, track highly dependent upon the magnitude and variation of
components are often maintained or replaced with little formation stiffness. This is exacerbated by consolidation
regard for properties of the formation and their subsequent and settlement of formation and the subsequent differential
effects on continued total asset life. Examples of expensive track geometry. The achievable track geometry can in turn
remedial work being carried out without eliminating the influence the total asset life of elements of the track
initial cause of the problem have been reported by Cope [1]. infrastructure, such as rails or sleepers, as suggested by
Ideally, site investigation data should be used not only for Cope [1]. This situation has been compounded by an
design of remedial ground improvement schemes, but to increase in rail traffic, pressure for higher line speeds and
predict perceived track performance and more importantly heavier axle loads. Consequently, there is an increasing
schedule necessary work proposals. Sharpe [2] has proposed importance in obtaining a consistent relative measure of
the ‘total route evaluation’ approach, a comprehensive trackbed stiffness in the physical and time restrictions
methodology yielding sufficient information to enable applicable on live track.
optimisation of renewals and maintenance strategy for a In addition to stiffness assessment, any trackbed
ten-year period. This methodology was first suggested by investigation should include an assessment of ballast, sub-
Sharpe and Collop [3], comprising staged levels of ballast and formation. Cope [1] suggests that if this layer
investigation, identifying formation anomalies and enabling information is omitted, the cause of poor formation sites
requirements for formation treatment to be identified many cannot be ascertained and consequently applicable remedial
years in advance. The Association of American Railroads is schemes cannot be implemented. Shahu et al. [6] suggested
also developing an expert system approach to diagnose and sub-ballast depth, after subgrade modulus, was the most
recommend the most cost-effective remedy for various track influential track parameter on overall track response. Sharpe
[2] recognised that the condition of the ballast would have
* Corresponding author. Tel.: þ44-115-922-9098; fax: þ 44-115-943-
1
1302. A track condition index based upon standard deviations of track
E-mail address: [email protected] (M. Brough). geometry.

0963-8695/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 6 3 - 8 6 9 5 ( 0 2 ) 0 0 0 5 3 - 1
146 M. Brough et al. / NDT&E International 36 (2003) 145–156

a strong influence on track quality. Groundwater profiles remediation techniques. The extent and methods of site
and behaviour must also be assessed due to the deleterious investigation used at Leominster are summarised in Fig. 1.
effects on formation condition and track quality. Li and
Selig [7] suggest that if the water table is within 2.3. Desk study
approximately 6 m of the surface, this is the major factor
affecting the moisture content of the formation and Specific poor formation hot spots were identified during a
subsequent problems. visual survey, evidenced by associated wavy top and clean
The nature of any site investigation will depend upon the polished exposed sleeper surfaces. This was supported by
specific nature of the problem and potential methods of track geometry measurements indicating excessive
remediation. In addition, physical and time restrictions in vertical track elevation standard deviations and associated
the UK rail network will limit investigation techniques that poor track quality.
can be employed. An assessment of suitable testing This section of track had previously been identified by
techniques has been described previously by Brough [8]. maintenance history as showing high potential for localised
This paper details a site investigation comprising trial pit formation problems. Historical records of previously
excavation, dynamic cone penetrometer (DCP), ground opened trial pits showed ballast on the up line was
penetrating radar (GPR), automatic ballast sampling (ABS), completely choked with grey clay and the water table was
cone penetration testing (CPT) and falling weight deflect- at or near the formation surface level. On the down line, trial
ometer (FWD) testing. These were used in two distinct pit records showed choked ballast was present but not across
levels of trackbed investigation. the entire trackbed section. This suggested possible ballast
Results of field trials performed at a suspected poor pockets, with choked ballast typically concentrated beneath
formation site and conclusions made are presented. The rails. Water levels were also generally much deeper on the
work forms part of a government funded EPSRC LINK down line. Track maintenance history generally suggested
research project (‘Improving the Stiffness of Existing Rail the potential for and occurrence of formation problems was
Track’—ISERT) in association with Railtrack, GTRM, higher on the up line, possibly due to an increased traffic
Scott Wilson Pavement Engineering, WS Atkins Rail, loading, observed higher water table and associated in situ
Keller Ground Engineering, Fugro and Serco Railtest. moisture. Subsequently, it was decided to locate a
significant proportion of the proposed trial pits on the up
line, with some trial pit locations matched on the down line.
2. Site investigation This desk study highlighted the importance of using
historical data to identify poor formation sites along a route.
2.1. Trial site In addition to trial pit logs, information regarding trackbed
condition can be ascertained from maintenance history and
The site is located near Leominster station in Here- track surveys. For example, performance data from
fordshire in the UK. The initial site choice was based upon automatic ballast cleaning operations could provide infor-
the need for repeated track maintenance, and occurrence of mation on the fines content of ballast, and roughly indicate
contaminated ballast in trial pits excavated as part of formation condition. Unfortunately, potential access to
previous track maintenance contracts. relevant data is not always feasible due to regional
ownership of track maintenance information by contract
2.2. Test schedule operators in a nonstandard format. Consequently, a
consistent data collection procedure is required and
The investigation, performed during a T22 and T33 subsequent data made accessible to supplement trackbed
possession, had several objectives. The preliminary assess- investigations.
ment, carried out under a T2 possession, was to identify soft In addition to track maintenance history, track com-
formation beneath track components and to highlight ponent type and location should be logged due to their
possible causes of differential track settlement. Conse- obvious effects on track behaviour. Subsequent correlation
quently, omissions in data were highlighted along with of track component logs with identified poor formation sites
recommendations for future site work that had to be done in may highlight possible relationships.
a night-time possession. The subsequent more-detailed
investigation confirmed the nature and cause of soft
formation and provided more detailed geotechnical infor- 3. Ground investigation: stage 1
mation that could be used for design and assessment of
2
Trial pits, DCP and GPR were implemented to assess the
Typically day-time work with intermittent access to track that is
primarily governed by railway traffic and subsequent safety considerations.
overall conditions of the entire length of track, and identify
3
Typically night-time work when there is no scheduled rail traffic with sites with existing and potential problems. The site
continuous access to track. investigation performed has been summarised in Fig. 1.
 M. Brough et al. / NDT&E International 36 (2003) 145–156 147

Fig. 1. Schematic showing site investigation performed at Leominster.

3.1. Trial pits assessment of poor formation sites. Third, as a rule

geotechnical engineers did not log trial pits, with a good
Trial pits would not normally be excavated in the possibility that important information may not be recorded
formation layers beneath the trackbed. However, due to or recorded incorrectly.
their perceived relevance to track geometry deterioration Sharpe [2] has also highlighted these limitations in his
they were included and logged as in a traditional site assessment of current state of the art methods of trackbed
investigation. Open trial pits suggested that the poor investigation. As shown in this study, there is the potential
formation sections were more prevalent on the up line, for a more consistent method, with logging and sample
due to contaminated ballast encountered at shallower excavation performed using traditional geotechnical prin-
depths. In all trial pits on the up and down line, ballast ciples and improved quality control. Samples can be
below depths of 0.3 – 0.6 m was completely choked and analysed off-site for classification testing and results used
contaminated with a soft grey clay/silt. In some cases this for design of any relevant remedial schemes. Although
layer was saturated or groundwater seepage was observed depth of trial pit excavation is limited, geotechnical test
below the ballast. This was particularly evident on the up devices such as the shear vane and DCP could be used in
line, with all trial pits opened showing a typically saturated trial pits and improve the current investigation practices in
soft mid layer, consisting of ballast and fines. Although it the UK railways industry. Needless to say, trial pit
would appear that excavated material from the up and down excavation, with improved quality control, still represents
line was of a similar composition, the contaminated fines a useful accepted tool for site investigation of trackbed.
layers on the down line did not exhibit the same high in situ Trial pit logs were used to calibrate the performance of other
moisture content. Excavation of bottom layers (. 0.7 m) test devices used in this investigation.
showed the formation ranged from a firm to stiff intermixed
mid grey and red brown slightly sandy clay/silt with 3.2. Dynamic cone penetrometer
occasional gravel, to a firm to stiff brown slightly sandy
clay/silt. As with other layers, these bottom layers exhibited Due to the improved access to the formation in excavated
higher moisture content on the up line. trial pits, there are substantial benefits of using simple hand
Trial pit logs confirmed initial observations regarding the held devices. Due to the relatively easy operating principles,
identification of poor formation at Leominster station, semi-nondestructive nature, low cost and simple results
particularly on the up line. However, although trial pit interpretation of the DCP, tests were performed at the
logs identified poor formation at discrete points along the maximum depth of open trial pits. The DCP also offered the
track, the extent, depth and nature of the poor formation potential for formation assessment without the removal of
section could not be assessed. This was due to current ballast.
limitations in the trial pit method currently being employed The DCP relies upon a penetration cone attached to a rod,
on the UK railway network. First, depth of trial pit being driven into the soil by means of a manual drop
excavation was limited, not including formation layers hammer, and the resulting penetration and number of blows
where problems are often sourced. Second, trial pit recorded. Following testing, results can be used to obtain
excavation was often infrequent resulting in an incomplete relative measures of bearing capacity in terms of depth of
148 M. Brough et al. / NDT&E International 36 (2003) 145–156

cone penetration per blow. Livneh et al. [9] suggests that a There is normally a strong reflection from the base of the
universal correlation exists between the DCP and California ballast layer in the trackbed. If the ballast has deteriorated or
bearing ratio (CBR) for a wide range of pavement and become contaminated, the radar waves cannot penetrate as
subgrade materials, testing conditions, and technologies. easily, so the strength of reflections from the base of the
In the current study, the DCP was used following trial pit ballast will be low. However, if there is a distinct layer of
excavation. Ideally, the DCP would be better implemented contamination within the ballast (e.g. clay slurry) there may
without prior excavation of the ballast. Initial trials would be a strong reflection from its upper surface. Typically a
suggest that penetration of DCP rods through compacted radar reading is taken between each pair of sleepers, and all
ballast is feasible. The effect of ballast on DCP measure- the results are combined to give an effectively continuous
ments is under investigation at present and will be reported profile. The speed of the survey is typically about 1 km/h.
in the future. The GPR trace is plotted using a grey scale to indicate the
DCP tests on the up line suggested that formation strength of the reflected signal. A constant gain has been
(approximately 1 m deep) underlying the open trial pits applied throughout, which gives an indication of the relative
was very soft, typically comprising three layers with strength of the reflected signal along the track, while
increasing strength with depth. Correlated CBR values filtering out much of the noise which is characteristic of
confirmed the nature of this soft formation layer with most GPR data. The particular gain function used has been
values ranging from 0.8 to 2.4% and an average value of specially developed to bring out the main features of the
1.5%. Similar layers were observed in trial pits opened on ballast layer. Lower layers may also be apparent, but the
the down line. However, results suggested the formation deeper information is less reliable with respect to both depth
was slightly stronger than on the up line. Although DCP and amplitude of reflection.
tests identified the soft formation layer of concern, the For the radar used in this study a 900 MHz antenna was
technique still gave no indication of stiffness variation used, with a typical frequency range between 75 MHz and
with depth. CBR is used extensively in the highways 1 GHz. A 900 MHz antenna was used to give a shorter
industry and has been considered a reliable relative pulse.
measure of combined stiffness and shear strength; This method of assessment can be used for qualitative
however, its application in a dynamic load environment first indication of ballast condition and potential fines
is increasingly subject to scepticism when compared with migration at poor formation sites, and also facilitates better-
other more relevant parameters such as elastic stiffness. targeted intrusive site investigations. Sharpe [3] also
Although the reliability of the DCP as a tool for providing suggests the GPR can be used as an excellent tool for the
geotechnical parameters relevant to the railway engineer quality control of trackbed renewal items, providing
is questionable, it could provide a useful relative indicator continuous ballast depth data.
of formation condition. It is quick, easy to use and At the Leominster site, GPR traces were obtained for the
maintain and could be applied in a T2 possession to the up and down line, to identify poor formation hot spots and
measurement of relative formation condition along a track areas with potential for track geometry deterioration (Fig. 2).
length. This should provide an insight into the heterogen- In agreement with results of trial pits previously opened
eity of the formation in the direction of traffic, now (marked as £ in Fig. 2), there was a section on the up line
known to be a major influence on track quality that exhibited a significant upward migration of fines, and
deterioration. Application rates of six tests per hour in insufficient variable nonslurried ballast depth. This was
trial pits excavated in the trackbed were observed on site. confirmed in automatic ballast samples, with nonslurried
ballast depth (below ballast surface) ranging from 220 to
3.3. Ground penetrating radar 680 mm. As can be seen in Fig. 2, the interface highlighted
by the GPR on the up and down line correlated very well
The inspection of ballast using the ground probing radar with the noncohesive or nonslurried ballast depth observed
has already been described in detail by Sharpe and Collop in ABS (this interface has been taken from ABS sections
[3], Hugenschmidt [10] and Jack and Jackson [11]. and is discussed later).
GPR was used to give an indication of ballast depth, as GPR analysis also suggested that the slurried ballast
well as condition of ballast and the formation. GPR is a layer extended over further track length than that investi-
nondestructive technique that uses electromagnetic radi- gated with trial pits, demonstrating the potential benefits of
ation to identify the presence of layer interfaces between the using GPR for highlighting the limits of any suspected
different materials comprising the trackbed construction. trackbed anomalies. The slurried layer was confirmed
Short pulses of electromagnetic radiation are directed into following ABS (discussed later), and highlighted the
the ground where they are reflected back at interfaces previously discussed limitations of trial pit excavation.
between different layers. The equipment records the Other sections on the up line were also identified that could
strength of reflected radiation as well as travel time for exhibit an upward migration of fines. In summary, on the up
the waves. Depths of layers can be estimated from the travel line the GPR identified three distinct sections of trackbed
time assuming the velocity of wave propagation. which required further investigation to assess existing and
 M. Brough et al. / NDT&E International 36 (2003) 145–156 149

Fig. 2. GPR trace at Leominster trial site.

potential future formation problems (Fig. 2). As a rule, the and higher line speeds, there may be the potential for future
extent and nature of trackbed anomalies highlighted in the problems as observed on the up line.
GPR trace correlates well with trackbed layers observed in The current study confirmed the suitability of GPR for
this additional site investigation (discussed later). initial identification and characterisation of existing and
On the GPR trace for the down line, three sections are potential poor formation sites. Limitations with the current
again evident. However, even though similar soils are methods of trackbed investigation, such as trial pit
present on the down line, the current formation problems excavation, were also highlighted. Although GPR can be a
observed on the up line are not apparent. This is in useful tool for assessment of trackbed condition, reducing
agreement with conclusions from trial pit excavation, the frequency of sampling or trial pit excavation required, a
sampling and subsequent laboratory testing. As similar minimal supplementary disturbed sampling should be
soils are present on the down line, with increased loading carried out. Suitable application to quality control of ballast
150 M. Brough et al. / NDT&E International 36 (2003) 145–156

depth during renewal works was also confirmed in the field controlled conditions by a trained geologist or engineer, and
tests. a subsequent accurate record of the formation and trackbed
layers [2]. The technique gives an idea of the condition and
3.4. Summary residual life of ballast directly beneath track components,
and overcomes many of the limitations of the trial pit
The trackbed from 38 ml 1108 yd4 to 38 ml 1197 yd, to method previously discussed. Not only is a full assessment
an approximate depth of 1.8 m, comprised a four layer of ballast, sub-ballast and formation condition feasible, the
system typically consisting of clean ballast (0 –0.3 m), technique is easy to use and does not require ballast
choked ballast (0.3 –0.6 m), a low bearing capacity, high in excavation. Sharpe [2] suggests depth of application is also
situ moisture content slightly sandy clay silt (0.6 –1 m) and much greater than is feasible with trial pits opened beneath
an underlying soft clay (1 – 1.8 m) exhibiting increasing the trackbed, with depths of up to 4 m achieved in 1 h.
strength with depth. This underlying soft clay exhibited low Concern has been raised about the ABS breaching the
bearing capacity on the up and down line, however, this was integrity of filtering or blanket layers, or even geomem-
slightly lower on the up line due to an observed higher in branes installed at cess heave sites. Ayres [12] has suggested
situ moisture content. Although bearing capacity was low, a fine to medium sand backfill should be used to maintain
analysis suggested sufficient existing or potential strength to filter properties. Alternatively, a cationic bituminous
resist the applied dynamic loading, with many of the emulsion could be used in the case of geomembrane
problems as a result of the slurried ballast and high in situ removal.
moisture content. Following analysis of data from the stage 1 evaluation,
Trial pit excavation suggested a water table extending up targeted ABS samples were taken at a total of 16 locations,
into the ballast layer with the trackbed typically saturated to a typical maximum depth of 2 m below the underside of
below 0.5 m. The observed increasing strength with depth rails. Following retrieval, samples were returned for logging
(to a maximum depth of 1.8 m) in the identified clay layer, and moisture content determination. A typical photograph
and shallow water table, could be due to surface ponding in and sample log for the up line can be found in Fig. 4.
subgrade depressions. In combination with repeated load- The trackbed typically consisted of a clean ballast layer,
ing, this could cause the observed slurried soft surface clay underlain with contaminated ballast and soft to firm clay. On
layer, with gradual in situ moisture content reduction with the up line, the contaminated ballast layer consisted of high
depth and associated increased strength. This layer arrange- moisture content slurried ballast. On the down line, the
ment has been presented in Fig. 3. GPR suggested the clean ballast layer was shallower and the contaminated
formation problem was more extensive on the up line than ballast layer consisted of stiffer clayey gravel. This was also
was targeted with trial pits, with further more detailed observed in the GPR trace, with an obvious section on the up
evaluation required. line demonstrating upward formation migration as
suggested by a shallow variable depth interface. On the
adjacent down line, there was a reasonably obvious more
4. Ground investigation: stage 2 uniform depth interface, possible evidence of the stiffer
contaminated ballast layer observed. Both GPR and
Following the stage 1 evaluation, omissions in data were subsequent targeted ABS highlighted the difference in
highlighted along with recommendations for future site moisture content and subsequent nature of fines migration
work that had to be done in a night-time possession. This observed in the up and down line trackbed.
detailed site investigation focussed on the trial pitted section Observations in ABS agreed with trial pit logs previously
of track and further along the track to investigate other discussed, demonstrating suitable application of the ABS
potential localised formation problems. The ABS, CPT and instead of hand dug trial pits. The slurried ballast on the up
FWD were used to confirm the major causes of the problem line was also evidenced by moisture content values recorded
and provide geotechnical parameters for design and in ABS samples ranging from 25 to 30%, in good agreement
assessment of any proposed remedial measures. The site with moisture contents of disturbed samples taken from trial
investigation performed has been summarised in Fig. 1. pits. A relatively shallow water table was also evident in the
slurried ballast layer in the up line ABS samples that was not
4.1. Automatic ballast sampler evident in the down line ABS samples.
ABS samples further supported conclusions from GPR
Sample tubes are driven dynamically into the trackbed, analysis that the poor formation area extended beyond the
through ballast cribs, to obtain a disturbed continuous core trial pitted section of track, covering a maximum track
sample of the ballast and underlying formation. Samples are length of 230 m. Trial pits had previously identified only a
sealed in a plastic tube, allowing examination off-site under 100 m section of track with poor formation. If trial pit logs
had been relied upon alone, the poor formation section of
4
In the UK railway industry track locations are referenced in miles (ml) trackbed would not have been completely identified,
and yards (yd). Metric conversion 1 ml ¼ 1760 yd ¼ 1609 m. possibly resulting in insufficient remedial treatments. This
 M. Brough et al. / NDT&E International 36 (2003) 145–156 151

Fig. 3. Typical trackbed profile in poor subgrade section on the up line (38 ml 1108 yd –38 ml 1197 yd).

research highlights the benefits of combined GPR analysis, In summary, testing highlighted the benefits of combined
supported with ABS testing, for the initial identification of GPR analysis, supplemented with ABS, for identification
poor formation sites. and assessment of poor formation sites. Extent of poor
Penetrometer readings were taken throughout the clay formation could be identified with GPR and targeted for
layer in ABS samples. Initial evaluation of the technique more detailed investigation with ABS. ABS gave relatively
suggested the narrow internal diameter of sampling tubes undisturbed continuous profiles, to a typical depth of 2 m, of
could cause sample disturbance and inaccurate shear ballast, sub-ballast, formation, moisture content, shear
strength profiles. However, test results on the whole agreed strength and extent of fines migration. If further depths
with DCP and CPT (discussed later) results. DCP tests need targeting, for sample extraction or soil profiling and
suggested that formation (roughly 1 m depth) underlying the characterisation, the CPT can be implemented. Sharpe [2]
open trial pits was very soft, typically comprising three suggests ABS is a relatively simple technique, allowing up
layers with increasing strength with depth. Shear strength to ten 1 m samples to be taken in 1 h with a four-man team.
measurements of the clay layer in ABS samples also
suggested a soft to firm clay layer with a shear strength 4.2. Static cone penetration test (SCPT)
ranging from 50 to 140 kPa. Average shear strength of 56
and 63 kPa were recorded in the up and down line, Horsnell and Adam [13] suggest the SCPT provides
respectively. This lower shear strength on the up line was detailed information on soil type and stratigraphy, allowing
as expected, due to the higher moisture content observed. on-site assessment of various engineering parameters and
ABS shear strength measurements suggested that the clay can be adapted to allow detection of hydrocarbon contami-
layer comprised alternating layers of stiff and soft clay or nation in ballast and formation. A more detailed assessment
consisted of an upper stiff crust underlain with soft clay. of soil stratigraphy can be obtained using a piezocone
CPT results tended to agree with this alternating stiff–soft penetration test in which pore pressures established in
layer formation. the soil during cone penetration are measured. The CPT is

Fig. 4. ABS log and picture.

152 M. Brough et al. / NDT&E International 36 (2003) 145–156

a useful tool for trackbed investigation and provides buildings, the CPT could not be used to obtain profiles
information on the quality, quantity and in situ character- across the track and confirm the existence of subgrade
istics of the ballast and the underlying subgrade material. depressions. CPT was performed eight and seven times
This data can provide a complete understanding of the along the track using hydraulic penetrometer equipment on
geotechnical properties which will govern drainage charac- the up and down line, respectively. A 7.5 tonne capacity
teristics, ground stability and ballast/subgrade deformation electric cone was used for each of the tests. Following data
behaviour. A test can be performed in the same time, or less, processing, friction ratios were calculated and soil type,
than it takes to excavate and investigate traditional trial pits. shear strength and coefficients of volume compressibility
Chrismer and Li [4] using CPT test results, have (Mv) determined.
suggested a failure mode similar in nature to the one CPT plots suggested a ballast layer from 0.9 to 1.3 m
suggested as occurring at the Leominster site. They suggest thick (Fig. 5). Although this was classified as ballast,
due to a distorted subgrade surface profile, water becomes with no evidence of contaminated ballast reported, a
trapped within the granular layer, further weakening the more detailed analysis of data could be performed to
subgrade and driving its displacement. Simply increasing highlight these distinct layers. This was proposed for any
the granular layer thickness has proven ineffective. This future CPT analysis to provide a more detailed assess-
situation appears to be similar in nature to problems ment of the trackbed and underlying shallow surface
experienced at the Leominster site, i.e. a very shallow layers. Evidence of ponding in shallow subgrade
water table in the trackbed possibly due to subgrade depressions could not be confirmed from CPT results
depressions with underlying slurried ballast and high without a more detailed evaluation of friction ratio data
moisture content soft clay at the layer interface. Continued at shallow depths.
ballast replacement, effectively increasing ballast depth, has Immediately beneath the ballast was a clay layer (Fig. 5).
not remedied the problem with continued onset of poor track On the down line, this layer typically comprised a firm to
geometry. Automatic ballast cleaning, may have exacer- very soft clay and in some instances contained thin layers of
bated the problem, with remaining fouled ballast shoulders sand. Of particular interest was the progression of strength
creating a channel for water ponding. through this clay layer. Typically, the layer became
For the failure mode discussed above, literature would increasingly soft with depth, up to the underlying sand
suggest the CPT could be applied to assessment for and gravel layer. On the up line, similar conditions were
subsequent design of a remedial scheme. However, due to observed. The clay layer typically comprised a firm to very
a relatively high ballast shoulder and nearby adjacent soft clay with no evidence of sand layers. This clay layer

Fig. 5. Typical CPT results from the up line (soil type profiles).
 M. Brough et al. / NDT&E International 36 (2003) 145–156 153

had a fairly consistent depth on the up and down line, and the upper 1.5 m of the trackbed, there is good correlation
was still evident at maximum depths of 2.9 –3.8 and 2.9 – between the results of the GPR, trial pits, ABS and CPT.
4.4 m, respectively. The stiffer clay at the top of this layer
could be evidence of ballast penetration due to repeated 4.3. Falling weight deflectometer
dynamic loading or of consolidation, with the underlying
clay demonstrating less consolidation and subsequently In the highways industry, the FWD was developed to
being much softer in nature. Calculated coefficients of provide detailed information on the structural response of
volume compressibility support this conclusion, with clay pavements to simulated traffic loading. An impulsive load is
from 2.5 to 3.5 m depth exhibiting high compressibility and applied to the pavement surface and the velocity response is
potential for consolidation. The incidence of very soft clay measured using geophones [2]. These velocity response
seemed greater on the up line, presumably due to the time histories are integrated, and the peak displacements are
increased loading, reduced ballast thickness and higher in used to give a ‘static’ deflection bowl that characterises the
situ moisture contents. quasi-static deflection response of the pavement [14]. With
On the up and down line beneath this clay layer, was a minor modifications, the FWD has been applied to the
medium dense to dense sand and gravel layer. Although measurement of railway track. A pre-determined impulse
other defined thin layers were often sandwiched between load is applied close to the rail seat area of a sleeper, and the
this clay and sand and gravel layer, these intermediate layers transient deflections of the sleeper and ballast accurately
seemed to be transitional, consisting of a combination of measured using geophones [3]. The deflections of the loaded
overlying clay and underlying sand and gravel layers. In sleeper, the ballast in the adjacent crib and the ballast in the
summary, although these minor transitional layers existed next crib, are interpreted as corresponding to the sleeper
(roughly 0.6 m deep), stratigraphy consisted of a firm deflection, the deflection of the loaded ballast and the
becoming soft clay underlain with a medium dense to dense deflection of the formation, respectively [1].
sand and gravel. Currently, the FWD is suggested as the only satisfactory
Although the sand and gravel layer would not be targeted way of obtaining sufficient data to assess trackbed stiffness
magnitude and variation for an individual site [2]. The track
with any ground improvement technique, the layer could be
loading vehicle (TLV) is not applicable, as scheduling for
used as a drainage channel, providing no artesian pressures
and implementation at a targeted section of track in the rail
exist. A more or less continuous sand and gravel layer can
network would not be cost effective. The TLV is currently
be found at a depth of 3.8– 5.3 and 3.9 – 5.4 m on the up and
being developed to measure parameters such as dynamic
down line, respectively. Any drainage system used would
track stiffness and track receptance whilst moving, and
have to be installed to these depths if the sand and gravel
should potentially be more applicable for whole route
layer were to be used effectively. Vibrating wire piezo-
assessments [5]. Hunt suggests that the most effective
meters have been installed to assess the pore water pressure
monitoring should be vehicle mounted and aim to give a
in the formation, and the potential for using the gravel layer continuous track stiffness profile, identifying sites for
as a drainage channel. consideration for remedial treatment. These sites could be
In summary, CPT result suggested that the stratigraphy investigated further by targeted stiffness measurement and
seemed to consist of a firm becoming soft clay underlain site investigation techniques. Research would suggest,
with a medium dense to dense sand and gravel. This clay including Hunt [5], Sharpe [2] and Sharpe and Collop [3],
layer had a fairly consistent depth on the up and down line, that the FWD is the most applicable method for determining
and was still evident at maximum depths of 2.9 –3.8 and quasi-static sleeper stiffness, an important parameter when
2.9 –4.4 m, respectively. Although soils were the same on assessing sites exhibiting track geometry deterioration.
the up and down line, there appeared to be a localised In this trial, FWD was performed on the up line, over a
moisture problem and shallower clean ballast depth on the 130 m section. Six months later, further testing was carried
up line. Beneath the clay layer, a more or less continuous out across the whole site on the up and down line (Fig. 1).
sand and gravel layer could be found at a depth of 3.8 –5.3 Following testing, deflection data retrieved was analysed
and 3.9 –5.4 m on the up and down line, respectively. If the off-site to provide track stiffness profiles. Deflection and
clay is analysed further, it would appear that shear strength stiffness profiles can be found in Fig. 6.
gets lower and compressibility gets higher with depth. This Deflection and stiffness profiles agree with observations
could be evidence of ballast penetration at the top of the clay from previous testing performed. Test results from the
layer or possible consolidation causing the observed shallow initial 130 m section on the up line suggest subgrade
stiffer clay conditions. There seems to be a layer from 2.5 to deflections (d1000) are not ideal for track support, with
3.5 m of particular concern exhibiting very low shear relatively large deflections observed (Fig. 6(a)). However,
strength and high compressibility. problems have not yet manifested in the section tested. This
CPT results could not be correlated with GPR results at may be due to the relatively uniform consistency of stiffness
excessive depth due to the shallow depth of application across this section (from 38 ml 969 yd to the platform end).
of the radar survey. As has been discussed previously, in If FWD profiles are compared with previous test results,
154 M. Brough et al. / NDT&E International 36 (2003) 145–156

Fig. 6. Deflection and trackbed stiffness profiles for (a) up line (b) down line.
 M. Brough et al. / NDT&E International 36 (2003) 145–156 155

then the consistent stiffness across the 130 m section ponding in subgrade depressions seemed greater on the
initially tested correlates with the well-defined interface up line with a relatively shallow water table evident.
suggested by the GPR trace. Initial test results suggested a This was also highlighted by penetrometer readings, with
well-defined interface, constant adequate ballast depth and shear strength measurements in up line ABS samples
generally consistent track support stiffness. Subsequently, suggesting a soft becoming stiff clay layer with values
no apparent track geometry deterioration problems are ranging from 50 to 140 kPa. The softening of clay at the
evident on this section of track. This consistent stiffness was top of this layer could be evidence of the ponding
still apparent across the 130 m section following FWD discussed previously. The incidence of very soft clay
testing six months later. Even though stone blowing had seemed greater on the up line, presumably due to the
been performed in the interim, there were no significant increased loading, reduced ballast thickness and higher in
changes in the deflection characteristics of the sleeper, situ moisture contents. On the down line, the clean
ballast and formation (Fig. 6(a)). ballast layer was shallower and the contaminated ballast
FWD testing across the whole site on the up and layer consisted of stiffer clayey gravel.
down line further defined the formation problem area of Ponding in shallow subgrade depressions could not be
concern previously highlighted (Fig. 6(a) and (b)). There confirmed from CPT results without a more detailed
appeared to be a length of track on the up and down line evaluation of friction ratio data at shallow depths. CPT
that was exhibiting a less consistent deflection and result suggested that the stratigraphy beneath the ballast
trackbed stiffness profile than was observed across the seemed to consist of a firm becoming soft clay underlain
whole site. From 38 ml 1276 yd to 38 ml 1014 yd, the with a medium dense to dense sand and gravel. If the
sleeper, ballast and formation deflections appear more clay was analysed further, it appeared that shear strength
variable than further along the track. This agrees with became lower and compressibility became higher with
previous testing that suggested the poor formation section depth. This could be evidence of ballast penetration or
stretched from the highways over bridge (38 ml 1276 yd) consolidation at the top of the clay layer. A layer from
to 38 ml 1024 yd.
2.5 to 3.5 m of particular concern exhibiting very low
If FWD data are analysed further the source of the track
shear strength and high compressibility could be
geometry problem can be further understood. Trackbed
evidence of the clay prior to any consolidation. With
stiffness magnitude does not appear to be the main cause for
increased loading or line speeds, this layer could cause
concern, with the range of measured stiffness similar along
more severe long-term settlement and associated track
the site. However, increased trackbed stiffness variation
geometry deterioration.
does appear to be correlated with the poor formation section
In agreement with ABS and CPT results, FWD
(38 ml 1276 yd – 38 ml 1014 yd). Previous results suggested
subgrade deflections are indicative of an underlying soft
ponding of water in subgrade depressions on the up line,
clay. However, trackbed stiffness magnitude does not
further evidenced by nonuniform subgrade deflections
across this section of track. As can be seen in Fig. 6, appear to be the main cause for concern, with the range
deflections of the formation are less consistent and relatively of measured stiffness similar along the up line. However,
high in this section on the up and down line. This can be increased trackbed stiffness variation does appear to be
seen when compared with other FWD data from other sites correlated with the poor subgrade (38 ml 1276 yd– 38 ml
as presented by Sharpe [1]. Even though track geometry 1014 yd). Deflections of the formation are less consistent
problems manifest on the up line, no significant differences and relatively high in this section on the up and down
can be observed in trends of formation deflection data on the line. If data are examined further, it would appear that
up and down line. If deflection data are examined further, it ballast deflections are typically higher on the up line,
would appear that ballast deflections are slightly higher on possibly suggesting the problem is sourced in this layer.
the up line, possibly suggesting the problem is sourced in This would suggest that the problem is shallow seated
this layer. Alternatively, this could suggest a problem at the and could be remedied by treatment of the contaminated
ballast/subgrade interface, such as fines migration and ballast and upper formation layers.
ponding. In summary, there appeared to be several suggested
causes for the observed track geometry deterioration on the
4.4. Summary up line. Ponding of water in shallow subgrade depressions
causing localised softening, fines migration into the ballast,
The poor subgrade extended beyond the trial pitted heterogeneous stiffness and consequently nonuniform track
section of track, stretching from the cross drain (38 ml settlement. This wavy top in turn has exacerbated the rate of
1276 yd) to roughly 38 ml 1024 yd on the up line, a track geometry deterioration. Although there is underlying
maximum length of 230 m. The trackbed typically soft clay exhibiting a high potential for consolidation, this
consisted of a clean ballast layer, underlain with layer should not need targeting in any remedial scheme,
contaminated ballast and soft to firm clay. The incidence providing load transfer issues can be addressed in shallow
of high moisture content slurried ballast and potential subgrade layers.
156 M. Brough et al. / NDT&E International 36 (2003) 145–156

5. Remedial ground improvement References

Results would suggest that track geometry deterioration [1] Cope D. Track technology for the future. Second course on vehicle/
on the up line at Leominster is due to a combination of track interaction, programme and notes, Cambridge: Emmanuel
insufficient clean ballast thickness, high in situ moisture College; 2000.
[2] Sharpe PS. Trackbed investigation. PWI J Rep Proc 2000;118(3):
content and increased loading from freight traffic. Con-
238 –55.
tinued maintenance in the form of repeated addition of [3] Sharpe P, Collop AC. Trackbed investigation—a modern approach.
ballast and tamping has resulted in excessive ballast Proceedings of the First International Conference on Maintenance and
thickness, much of which extends into the underlying clay Renewal of Permanent Way and Structures, London, UK: Engineering
layer, forming ballast pockets along the trackbed. Ponding Technics Press, Brunel University; 1998.
[4] Chrismer SM, Li D. Cone penetrometer testing for track substructure
of water in resultant formation depressions has caused a
design and assessment. Sixth International Heavy Haul Railway
continual softening of the upper clay formation layers, Conference—Strategies Beyond 2000, Cape Town, South Africa;
extensive fines migration into the clean ballast layer and a 1997.
nonconsistent trackbed stiffness. This differential stiffness [5] Hunt GA. EUROBALT optimises ballasted track. Railway Gazette
and localised ponding has exacerbated the problem, International; December 2000.
minimising potential benefits of repeated ballast addition. [6] Shahu JT, Kameswara Rao NSV, Yudhbir. Parametric study of
resilient response of tracks with a sub-ballast layer. Can Geotech J
As stiffness or shear strength magnitude appears to be
1999;1137– 50.
sufficient across the site, this should not be the main focus of [7] Li D, Selig BT. Evaluation of railway subgrade problems. Transpn
any remedial programme. To overcome the problems at Res Rec 1995;1489:17 –25.
Leominster, either consistency of the formation condition [8] Brough MJ, Ghataora GS, Stirling AB, Madelin KB, Rogeus CDF,
should be improved and subsequently any formation Chapman DN. Investigation of railway track subgrade. I. In situ
depressions disrupted, or the ponding water and slurried assessment. Institute of Curl Engineering, Transport Journal. (2003).
[9] Livneh MI, Ishai I, Livneh AL. Effect of vertical confinement on
ballast layer should be removed. At present, shallow soil dynamic cone penetrometer strength values in pavement and subgrade
mixing treatments or novel drainage technologies are being evaluations. Transpn Res Rec 1995;1473:1–8.
considered. A full assessment of all available ground [10] Hugenschmidt J. Ballast inspection using ground penetrating radar.
improvement technologies for remediation of trackbed Proceedings of the Second International Conference on Railway
formation is reported in Ref. [15]. Engineering, London, UK; 1999.
[11] Jack R, Jackson P. Imaging attributes of railway track formation using
ground probing radar (GPR). Railway Engineering—98, First
International Conference on Maintenance and Renewal of Permanent
Acknowledgements Way and Structures, London, UK: Engineering Technics Press,
Brunel University; 1998.
The authors gratefully acknowledge the combined [12] Ayres DJ. Trackbed investigation. J Rep Proc Permanent Way
financial and technical support of the EPSRC and the Institution 2000;118(4):356–60. Comments on paper of same title by
Phil Sharpe.
industrial partners Railtrack, GTRM, Scott Wilson Pave- [13] Horsnell MR, Adam CH. Geotechnical investigations for railway
ment Engineering, WS Atkins Rail, Keller Ground Engin- maintenance and renewal. Proceedings of the Third International
eering, Fugro and Serco Railtest. The authors also gratefully Conference on Railway Engineering, London, UK; 2000.
acknowledge the specific technical advice and discussion [14] De Almedia JR, Brown SF, Thom NH. A pavement evaluation
provided by Scott Wilson Pavement Engineering regarding procedure incorporating material non-linearity. ASTM STP 1198
1994;.
the GPR, ABS and FWD. Similar thanks go to Fugro for
[15] Brough MJ, Stirling A, Ghataora G, Madelin K. Improving railway
advice and discussion regarding the CPT and rig appli- subgrade stiffness—assessment of traditional in situ ground improve-
cations, and to WS Atkins Rail for supervising and logging ment techniques. Proceedings of the Third International Conference
the trial pits. on Railway Engineering, London, UK; 2000.