Granular Matter (2009) 11:391–401 DOI 10.

1007/s10035-009-0144-4

Cyclic loading of railway ballast under triaxial conditions and in a railway test facility
B. Aursudkij · G. R. McDowell · A. C. Collop

Received: 14 May 2008 / Published online: 19 May 2009 © Springer-Verlag 2009

Abstract A recently developed large-scale triaxial test apparatus for railway ballast testing comprises a double-cell arrangement for measuring volume change by differential pressure. Monotonic and cyclic tests were performed on limestone ballast samples. Axial and volumetric strains and breakage were determined from both types of test. Resilient modulus and Poisson’s ratio were obtained only from the cyclic tests. The permanent axial strain and breakage results from the cyclic tests are compared with the simulated traffic loading in the railway test facility (RTF) which comprises three sleepers embedded in ballast over a subgrade. The traffic loading in the RTF was applied by hydraulic actuators with built-in displacement transducers. A column of painted ballast was placed under a rail seat of the middle sleeper to ease sample collection for sieve analysis at the end of the test. The stress condition in the RTF is predicted by a simple calculation based on findings of previous literature. It was found that the results from the cyclic triaxial test with conditions similar to the predicted conditions in the RTF were comparable to those results from the RTF tests. Keywords Railway ballast · Cyclic loading · Laboratory testing

1 Introduction The rail network is one of the most important transportation systems in everyday life. It provides a fast means of transportation by a durable and economical system. In the past, the train and track superstructure, such as rails and sleepers were
B. Aursudkij · G. R. McDowell (B) · A. C. Collop University of Nottingham, Nottingham, UK e-mail: glenn.mcdowell@nottingham.ac.uk

the focus of attention of railway engineers. Less attention was given to the substructure such as ballast, subballast and subgrade even though they are as important as the superstructure. While the superstructure provides the main function of the railway, the substructure provides the foundation to support the superstructure and to help the superstructure to reach its optimum performance. Track settlement occurs after long-term service. Excessive settlement can cause poor passenger comfort, speed restriction, and potential derailment. According to Selig and Waters [17], most of track settlement occurs in the ballast layer. It is also important to study the degradation of ballast to increase and predict ballast life on the track, reduce waste ballast, minimise the frequency and cost of ballast replacement, and lead to further developments in the railway industry. Performing experiments on ballast in railway track is desirable since results can be obtained for real site conditions. However, it is very difficult to control test variables and to collect data on site. Hence, the RTF introduced by Brown et al. [4] was developed to simulate the conditions in a real track in a proper controlled way. It is a full-scale railway track model housed in a concrete pit filled with subgrade material and ballast. However, a test in the RTF takes a long time to prepare and perform and a lot of labour and effort is involved. Therefore, a large-scale triaxial test apparatus, in which testing is less time-consuming and requires less labour and effort, was also developed. The triaxial test is performed on a large cylindrical ballast sample. Unlike the conventional triaxial testing equipment, the test apparatus measures sample volume change of a dry sample by differential pressure instead of measuring the volume of water entering or leaving a saturated sample or sample radial expansion by conventional on-sample transducers. However, secondary methods for volume change measurement were used when the designed equipment for measuring volume change was

123

392

B. Aursudkij et al.

Fig. 1 Vertical stress at sleeper base contact [18]

not available. The aims of this paper are to introduce the newly developed triaxial test apparatus and to compare the results from the triaxial tests with the results from the railway test facility (RTF).

2 Typical loading on ballast
Fig. 2 Diagram of a box test [17]

To perform ballast testing in a laboratory, the typical loading conditions for ballast should be known so that the tests simulate the real site condition of ballast. According to Shenton [18], the maximum vertical stress in the ballast at the sleeper contact varied between 200 and 250 kPa under a 100 kN load on the sleeper as shown in Fig. 1. This is comparable to Raymond and Bathurst [15] who reported that the average vertical stress at the sleeper-ballast interface was 140 kPa. Key [9] varied the deviatoric stress between 12.5 and 250 kPa in his triaxial tests on ballast. Selig and Waters [17] performed a box test which was a test that simulated ballast behaviour and performance under field conditions. Ballast is placed in a box with a sleeper segment shown in Fig. 2. The test can apply cyclic loading to simulate traffic loading on the rail section as illustrated by Lim [11] and shown in Fig. 3. Horizontal stress sensors were installed on the wall of the box to measure the horizontal stress in ballast. The results from one of the sensors are shown in Fig. 4. It can be seen from the figure that horizontal stress in the ballast in the loaded and unloaded state (at maximum and minimum load of the cyclic loading, respectively) fluctuates between 20 and 60 kPa and eventually reaches 30 kPa. Furthermore, according to Selig and Alva-Hurtado [16], the in-situ confining pressure of self standing ballast perpendicular to the rail was approximately 5–40 kPa. The typical loading frequency of traffic loading in the track is normally around 8–10 Hz for a normal train (by assuming axle wheel spacing of 2.6 m and train speed of 75–94 km/h) and may reach 30 Hz for a high speed train. However, loading frequencies for different laboratory ballast testing vary

Sleeper Rail

300mm

700mm

Simulation Area

Fig. 3 Plan of rail and sleepers showing section represented by the box test [11]

depending on the capacity and constraint of each apparatus. Compared to the magnitudes of vertical and horizontal loading, the range of variation of loading frequency in laboratory testing is significantly wider. This is because in real track, high loading frequency increases the magnitude of dynamic loading to the track caused by defects such as wheel flats or welded joints on the track. Since the defects are not present in laboratory testing, the loading frequency does not play an important role. Key [9] applied 0.16 Hz for the first 50 cycles and then used 0.5 Hz for the rest of the test. Shenton [18] varied the loading frequency from 0.1 to 30 Hz. However, he used 0.1 Hz for the first eight cycles in all of his tests. The reason for applying low frequency during the beginning of

123

Cyclic loading of railway ballast under triaxial conditions and in a RTF

393

ram against the stationary load cell at the top of the sample. The maximum loading frequency of the apparatus is 10 Hz. However, the loading frequency in the performed cyclic tests is limited to 4 Hz due to the rate of oil flow that drives the machine. 3.2 Volume change measurement The principle of volume change measurement of this apparatus is based on the HKUST system presented in [13]. It uses a differential pressure measurement for volume change instead of on-sample axial and radial displacement measurements. This method was preferred due to the irregular outer surface profile of the sample caused by the large size and angular shape of the ballast. A conventional on-sample measuring system such as strain collars would measure the localised grain movement rather than the overall change in diameter of the sample and would be difficult to attach to the irregular surface of the sample such as the samples of [9,19]. The system uses water which is in direct contact with the sample membrane in an inner cell as shown in Fig. 5. Air at a specified pressure is directed to the outer cell so that the water in the neck is pressurised and consequently a confining pressure is applied to the sample. As the pressures in the inner and outer cells are equal, the inner cell volume remains constant so any change in the volume of the sample will displace an equal volume of water in the inner cell. The volume change is measured by recording the differential pressure between the water in the neck of the inner cell and the water inside a reference tube using a differential pressure transducer. At the beginning of each test, the water levels in both the inner cell and reference tube are equal. If the sample in the inner cell expands, the water level in the inner cell rises while the water level in the reference tube remains at the same level. The differential pressure transducer reading is converted to a volume change based on the known cross-sectional areas of the inner cell neck and the reference tube. However, the volume change from the differential pressure transducer reading is the combination of the sample volume change and the volume change due to the displacement of the bottom ram movement. Therefore, the volume change from the bottom ram must be subtracted to obtain the sample volume change. The differential pressure transducer cannot register the volume change instantaneously. Therefore, the loading speed must be sufficiently low. A loading rate of 1 mm/min and frequency of 0.2 Hz was used in the monotonic and cyclic tests (for measuring resilient properties), respectively. The original HKUST system was originally developed for a smaller and more uniform sample and would probably measure dynamic volume change at a higher frequency. A scale was attached to the neck of the inner cell to double check the accuracy of the volume change measurement. For the large

Fig. 4 Effect of repeated load on horizontal stress in box test [17]

Fig. 5 Schematic diagram of the triaxial apparatus [1]

the test is that the deformations during the first few cycles are generally large and they might exceed the capacity of a testing machine in terms of the hydraulic oil flow required to give the required deformation rate.

3 Large-scale triaxial test 3.1 Loading capacity Figure 5 [1] shows a schematic diagram of the triaxial test apparatus. It can apply a constant confining stress of up to 1 MPa via pressurised air and water. This will be explained in more detail in the later section which also involves the volume change measurement system of this apparatus. The apparatus is also capable of applying both monotonic and cyclic loadings to a sample. The maximum applicable force is 100 kN which is equivalent to a deviatoric stress of 1,415 kPa on the designed sample (with 300 mm diameter). The axial load is applied to the sample by the movement of the bottom

123

394
Volume change reading (mm )
3

B. Aursudkij et al.
60,000 50,000 40,000 30,000 20,000 10,000 0 0 10,000 20,000 30,000 40,000 50,000
3

y = 0.9895x R2 = 0.9999

140mm long

60,000

Volume change by bottom ram (mm )

Fig. 6 Calibration results of the differential pressure transducer

Wave transmitter located at the tip of the transducer

triaxial system used here, although the resilient volumetric strain may not be accurate over one cycle at 0.2 Hz, it is considered that the accuracy is sufficient to measure permanent volumetric strain over many such cycles. The differential pressure transducer can be calibrated by displacing a known volume in the system by moving the bottom ram upward in increments. However, this has to be done without a sample to ensure that the only volume change that the transducer reads is from the bottom ram movement. Figure 6 shows the average calibration results from ten calibrations. It can be seen from the figure that the differential pressure transducer is satisfactory. However, the differential pressure transducer was not available during some tests. Therefore, some other secondary volume change measurement methods that did not require a major modification to the existing system were used. For monotonic tests, the scale, attached to the neck of the inner cell, whose main purpose is to double check the accuracy of the differential pressure transducer was used to measure sample volume change instead. But this cannot be used with cyclic tests as the movement of the water level was too fast to be read with the scale. For a cyclic test, the ultrasonic proximity transducer was used to measure volume change. The ultrasonic proximity transducer shown in Fig. 7 uses sound waves above the audible limit, to measure the distance from a transmitter to a surface. It measures the time lag between the transmitted sound waves and the return sound waves and converts this to voltage. It is ideally suited to provide a noncontact switching device or an analogue device to measure fluid level. The only way of fitting it without a significant modification was to carry out a test with the outer cell removed. The transducer was clamped to a shaft supporting the load cell so that it could look directly down to the water in the inner cell. A test was then carried out with a sample under vacuum in the inner cell as the external pressure could not be applied with the outer cell removed. This system appeared to be working well for the low frequency up to 1 cycle per minute. However, at 4 Hz the upper shaft supporting the load cell could be

Fig. 7 Ultrasonic proximity transducer (UPT) [1]

seen to be moving laterally, because the top platen was tilted due to uneven settlement of the ballast. This would not happen when the outer cell is in place as it acts as a support for the upper shaft and keeps the top platen level. The test was stopped at this stage and only test results up to 2,000 cycles were obtained. However, the differential pressure transducer had now been returned from the manufacturer and was used in the remaining part of this test and the other cyclic tests. 3.3 Test sample According to Skoglund [19], the diameter of a cylindrical triaxial test sample should be approximately five to seven times the maximum particle size of the sample for meaningful test results. Since a typical maximum particle size of ballast is 50 mm, the apparatus was designed for a sample of 300 mm diameter, i.e. six times the maximum ballast particle. The height of the sample is 1.5 times the diameter, i.e. 450 mm. However, Bishop and Green [3] suggested the height should be twice the diameter to eliminate the effect of friction at both ends of the sample. Nevertheless, the ratio of 1.5 enabled easier and more economical design of the cells in this project. The governing factor was the rate of oil flow; for typical resilient strains in ballast, the required displacement rate at a maximum desirable frequency of 10 Hz was not feasible—hence a shorter sample was required. Duncan and Dunlop [5] concluded that end friction caused an insignificant increase in the angle of shearing resistance in their drained triaxial tests on sand. Furthermore, they added that lubrication at both ends was necessary when volumetric strain needed to be calculated. This was because end friction usually caused the triaxial sample to bulge into a barrel shape. This means the diameter of the sample was not uniform through the whole height of the sample which affects the calculation of volumetric strain. However, this is not a problem in this

123

Cyclic loading of railway ballast under triaxial conditions and in a RTF
100 90 80 70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70

395

% Passing

Limestone Specification

Particle size (mm)

Fig. 8 Grading of limestone ballast sample and ballast grading specification according to RT/CE/S/006 Issue 3 [14]

case with the double-cell arrangement since the volumetric strain can be measured directly. Even though the overall volumetric strain is measured by the differential pressure transducer, the volumetric strain is still non-uniform. Additionally, an attempt was made to reduce sample end friction by placing a pair of 300 mm diameter circular latex sheets sandwiching a film of silicone grease at each end of the sample. Ballast which follows the UK ballast grading specification RT/CE/S/006 Issue 3 [14] shown in Fig. 8 is used to make up a sample. The sample is prepared in a split aluminium mould on a vibrating table. The bottom latex sheets, used to reduce sample end friction, are placed at the bottom of the sample. It should be noted that they had 20 mm diameter holes at their centres for drainage purposes. Ballast is then filled in three layers with equal height. Each layer is vibrated for thirty seconds with a 20 kg surcharge on top of each layer. The sample mass ranges between 48 and 52 kg, corresponding to a bulk density of 1,500–1,630 kg/m3 and void ratio of 0.6 to 0.7. The sample is covered by two layers of latex membrane. The thicknesses of the inner and outer membranes are 2 and 1 mm, respectively. It was found that a bicycle repair kit could be used to fix the holes in the membranes if there were any. Another pair of latex sheets is placed at the top of the sample. This pair has no hole because drainage is not possible through the top of the sample. After that, the silicone grease is applied in the top and bottom grooves located at both sample ends. An o-ring is used to seal the membranes into the groove at each end. Also, insulating tape is put around both o-rings and two jubilee clips are used to cover the o-rings. The jubilee clips hold the position of the o-rings during a test while the tape helps to maintain an even pressure on the o-rings. 3.4 Test procedure and programme Both monotonic and cyclic tests were performed on limestone samples. After each test, the sample is sieved to obtain

the particle size distribution. For a monotonic test, the sample is initially put under a seating load of 1 kN and then loaded at 1 mm/min until the axial strain reaches approximately 12% to ensure that the sample does not expand radially so much that it touches and might damage the inner cell. Since the value of axial strain is used as a finishing point of each test, the maximum deviatoric stress for each test is unknown before the test. However, the peak strength of the material should be obtained by doing so. It should be noted that the differential pressure transducer was not yet available when conducting the monotonic tests. The sample volume change was obtained by reading the scale on the inner cell neck instead as explained in Sect. 3.2. For the cyclic tests, one hundred thousand cycles of sinusoidal loading are applied to the sample. The sample is also put under a seating load of 1 kN at the beginning of the test. Then, the load is slowly increased to the mean load, i.e. the midpoint between the maximum and minimum load of the cyclic loading, at approximately 0.5 kN per minute to avoid a sudden dilation which might damage the inner cell and make the water in the neck overflow. Subsequently, cycles 1 to 5 and 6 to 30 are loaded with durations of 5 minutes per cycle and 1 minute per cycle, respectively. Then, most parts of the test are performed with a frequency of 4 Hz until cycle 100,000. However, cycles 31–100, 181–200, 481– 500, 981–1,000, 1,981–2,000, 4,981–5,000, 9,984–10,000, 19,981–20,000, 49,981–50,000, and 99,981–100,000 are loaded at 0.2 Hz to obtain the sample volume change. It should be noted that during cyclic loading, the seating load of 1 kN is always kept to ensure that the sample was always in contact with the load cell. Unlike the monotonic tests, the differential pressure transducer was used in the cyclic tests. However, for one cyclic test, the volume change during the first 2,000 cycles was measured by the ultrasonic proximity transducer. After that, the differential pressure transducer was used instead as explained in Sect. 3.2. Three monotonic tests and four cyclic triaxial tests were performed on limestone samples with constant confining stresses of 10, 30, and 60 kPa. The parameters of each test can be found in Table 1. It should be noted that the names of the monotonic tests indicate the confining stresses while the names of the cyclic tests indicate both confining stresses and maximum (q/p )max ratios. 3.5 Monotonic test results Figure 9 shows the plot of deviatoric stress against axial strain. It can be seen that the deviatoric stress eventually becomes stable and does not drop significantly. Figure 10 shows the plot of volumetric against axial strain. It can be seen from the figure that after a short period of volumetric compression, the sample began to dilate. The test with 10 kPa has the largest dilation corresponding to its largest (q/p )max .

123

396 Table 1 List of triaxial tests Test type Test name Confining Max (q/p )max Density stress deviatoric (kg/m3 ) (kPa) stress (kPa) 10 30 60 10 60 60 N/A N/A N/A 60 180 180 360 N/A N/A N/A 2.0 2.0 1.5 2.0 1,511 1,539 1,545 1,539 1,559 1,600 1,592
1.0 0.9 0.8 0.7 M10 (10 kPa) M30 (30 kPa) M60 (60 kPa)

B. Aursudkij et al.

% passing

0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 5 10 15 20 25

Monotonica M10 M30 M60 Cyclic C10/2.0 C60/1.5 C60/2.0
a All

C30/2.0b 30

Size (mm)

Permanent axial strain (%)

three monotonic tests were performed without the differential pressure transducer b The ultrasonic proximity transducer was used only during the first 2,000 cycles. The differential pressure transducer was used afterwards
400

Fig. 11 Particle size distributions after monotonic triaxial tests
7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0 20,000 40,000 60,000 80,000 100,000
C10/2.0, qmax = 60kPa C30/2.0, qmax = 180kPa C60/1.5, qmax = 180kPa C60/2.0, qmax = 360kPa

Deviatoric stress (kPa)

350 300 250 200 150 100 50 0 0 5 10 15

M10 (10 kPa), q/p'max = 2.3 M30 (30 kPa), q/p'max = 2.2 M60 (60 kPa), q/p'max = 2.0

No. of cycles

Fig. 12 Permanent axial strains from cyclic triaxial tests

Axial strain (%)

Fig. 9 Deviatoric stress versus axial strain from monotonic tests
Axial strain (%)
0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 -5.0 -6.0 -7.0 5 10 Compression 15

Volumetric strain (%)

Dilation

M10 (10 kPa), q/p'max = 2.3 M30 (30 kPa), q/p'max = 2.2

approximately the same after each test and no particles smaller than 22.4 mm were put in the sample before the test. It can be seen from the figure that even though the breakage is small, it increases with the decreasing level of dilation. This finding agrees with [7] who stated that the breakage increased when dilation was suppressed. 3.6 Cyclic triaxial test results

M60 (60 kPa), q/p'max = 2.0

Fig. 10 Volumetric strain versus axial strain from monotonic tests

The volumetric strain data shows that the samples were more or less dilating at a constant rate after about 12% strain, suggesting that a critical state was not imminent, but rather that each sample had reached and sustained its peak strength over quite a large range of strain. Figure 11 shows the particle size distributions of the samples after the monotonic tests. It should be noted that only particle sizes smaller than 22.4 mm are shown in the figure because the number of particles larger than 22.4 mm remained

According to Fair [6], the permanent axial strains from cyclic triaxial tests on ballast from similarly prepared tests are not consistent. The cause of this discrepancy is the bedding errors occurring during the first cycle. Following that, the permanent axial strains of the first cycles from all cyclic tests were removed. The permanent axial strains from the tests are shown in Fig. 12. It can be seen from the figure that with the same confining stress, permanent axial strain increases with increasing deviatoric stress or increasing (q/p )max . With the same (q/p )max , the sample contracts more with increasing confining stress. Also, with the same maximum deviatoric stress, permanent axial strain increases with increasing (q/p )max . The permanent volumetric strains are shown in Fig. 13. According to the figure, only the test with 10 kPa shows dilative behaviour (negative permanent volumetric strain).

123

Cyclic loading of railway ballast under triaxial conditions and in a RTF
Permanent volumetric strain (%)
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 0 20,000 40,000 60,000 80,000 100,000 Dilation Compression C10/2.0 C30/2.0 C60/1.5 C60/2.0
300 Mr = 6.7θ0.6

397

Resilient modulus (MPa)

250 C30/2.0 200 150 100 50 0 0 100 200 300 400 500 600 C10/2.0 C60/1.5 C60/2.0

No. of cycles

Sum of principle stresses (kPa)

Fig. 13 Permanent volumetric strains from cyclic triaxial tests
300

Fig. 15 Resilient moduli versus sum of principal stresses
0.50

Resilient modulus (MPa)

Poisson's ratio

250 200 150 100 50 0 0 20,000 40,000 60,000 80,000 100,000

C10/2.0 C30/2.0 C60/1.5 C60/2.0

0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0 20,000 40,000 60,000 80,000 100,000

C10/2.0 C30/2.0 C60/1.5 C60/2.0

No. of cycles

No. of cycles

Fig. 14 Resilient moduli from cyclic triaxial tests

Fig. 16 Poisson’s ratios from cyclic triaxial tests
1.0

However, test C60/2.0 dilates at the beginning of the tests and then contracts after a few cycles. This might be because the maximum deviatoric stress in this test (360 kPa) was equal to the peak strength of the sample under the same confining stress in the monotonic test as shown in Fig. 9. Furthermore, with the same (q/p )max , one might expect the permanent volumetric strain to increase with increasing confining stress. However, the strain from C60/2.0 is less than C30/2.0 probably because of the dilation at the beginning of C60/2.0. The plot of resilient modulus (Mr ) against number of cycles is shown in Fig. 14. It was found that the final resilient moduli followed the K − θ model (Mr = k1 θ k2 , where θ is the sum of principle stresses and k1 and k2 are empirical constants; [10]) as shown in Fig. 15. According to the figure, k1 and k2 were found to be 6.7 and 0.6, respectively. The Poisson’s ratios are shown in Fig. 16. It was found from the figure that definite trend of Poisson’s ratio cannot be found from the tests with the same (q/p )max . The fluctuations in the resilient modulus and Poisson’s ratio results may be due to the electrical drift that can affect the differential pressure transducer as it is very sensitive. Figure 17 shows the particle size distribution of the fines after each cyclic test. The breakage behaviour matches the findings of [8]. According to them, the behaviour of ballast

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 5 10 15 20 25

C10/2.0, max q = 60 kPa C30/2.0, max q = 180 kPa C60/1.5, max q = 180 kPa C60/2.0, max q = 360 kPa

% passing

Size (mm)

Fig. 17 Particle size distributions from cyclic triaxial tests

breakage resulting from cyclic triaxial tests with the same maximum deviatoric stress can be divided into three zones as shown in Fig. 18. The ballast breakage index in the figure indicates the breakage level based on calculation of area under the particle size distribution before and after each test as shown in Fig. 19. The first zone is called the dilatant unstable degradation zone (DUDZ). In this zone, excessive radial expansion occurs due to low confining stress (about 10– 30 kPa) and results in shearing and attrition of particle which causes degradation. The second zone is called the optimum degradation zone (ODZ) where the degree of degradation

123

398

B. Aursudkij et al.

I-beam Loading frame Load actuators Sleepers Geogrid (optional)

3

2 Ballast

1

0.20 m 0.30 m

Subgrade

0.90 m

Concrete slabs I-beam 2.10 m

0.15 m

Fig. 20 End-view diagram of the facility [1]

Fig. 18 Effect of confining pressure on particle degradation [8]

zone, increasing confining stress level increases the stress at particle contact and thus increases the breakage level. Similar to the findings of [8], the breakages from C30/2.0 and C60/1.5 which have the same maximum deviatoric stress are similar as they both are in the ODZ (the second zone). The breakage from C60/2.0 is larger than C60/1.5 due to larger deviatoric stress. Furthermore, the breakage from C10/2.0 is comparable to C30/2.0 and C60/1.5. If the test with 10 kPa confining stress had the same maximum deviatoric stress with C30/2.0 and C60/1.5, i.e. 180 kPa, the breakage from that test would have been larger than the two. However, the test with 10 kPa confining stress presented here had the maximum deviatoric stress of 60 kPa. Therefore, the level of breakage was reduced so that it was comparable with C30/2.0 and C60/1.5. 4 Railway test facility 4.1 Apparatus A detailed presentation of the RTF has been given by Brown et al. [4]. The RTF is housed in a concrete pit with dimensions of 2.1 m(width) × 4.1 m(length) × 1.9 m(depth). A schematic end-view of the facility and a photo are shown in Figs. 20 and 21 [1], respectively. Figure 20 [1] shows that a geogrid can be placed at the bottom or at any point in the ballast layer but it was not used in the test presented here. The pit was filled with subgrade material and railway ballast. Silt was chosen as the subgrade material because of its availability and ease of placing and compaction. It was compacted in 180-mm layers using a plate vibrator to a depth of 900 mm. After compaction, it had a density of 1,770 kg/m3 , surface stiffness of approximately 18.42 MPa, and a moisture content of 15.5%. The ballast used in the RTF is the same

Fig. 19 Ballast breakage index in [8]

is smallest among the three zones and the confining stress is about 30–75 kPa. This level of confining stress is able to help the sample achieving the optimum particle arrangement where the average number of contact per particle is maximised and hence the probability of particle breakage is reduced [12]. The confining stress in the third zone or the Compressive stable degradation zone (CSDZ) is larger than 75 kPa. In this zone, the breakage index increases with increasing confining stress. Even though the maximum average number of contact per particle can also be achieved in this

123

Cyclic loading of railway ballast under triaxial conditions and in a RTF Fig. 21 The railway test facility [1]

399

Sleeper number 2 1 LVDT (front end)

3

100 80

Actuator 1 Actuator 2 Actuator 3

60 40 20 0

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Time (s)

Fig. 22 Loading pattern in the railway test facility [1]

type of ballast used in the triaxial tests, i.e. limestone. The ballast was placed and compacted in 100-mm layers by the same plate vibrator to a depth of 300 mm. Three concrete sleepers were embedded in the ballast. Loading is transmitted from three servo hydraulic actuators to the sleepers by means of the spreader beams (three steel beams on top of the sleepers in Fig. 21 [1]) located on rollers on the rail seatings. Each actuator has a built-in vertical displacement transducer for recording the settlement. In addition to the vertical displacement transducer in the middle actuator, the settlement of the middle sleeper was also measured from two LVDTs: one at each end of the middle sleeper (as shown in Fig. 21 [1]) to double check the settlement reading from the middle actuator. 4.2 Test procedure Simulated traffic loading is achieved by applying sinusoidal loading with maximum magnitude of 94 kN and 90 degree phase lag between each actuator as shown in Fig. 22 [1]. This loading pattern was suggested by Awoleye [2]. It is intended to simulate a train running over three sleepers with 50% of

the wheel load on the middle sleeper and 25% of the wheel load on the outer sleepers. With this loading pattern, the RTF simulates an axle load of approximately 20 tonnes which is comparable to a typical heavy axle load on top of the middle sleeper. Due to the pressure and flow capacity of the hydraulic pump in the laboratory, the loading frequency in the RTF was limited to 3 Hz. Simulated traffic loading of one million cycles is applied to the ballast in the RTF. Before a test, a 300-mm diameter column of painted ballast is placed under the front-end rail seat of the middle sleeper to aid sample collection after the test. To obtain the settlement, the position readings from the three actuators and two LVDTs located at both front and rear ends of the middle sleeper are recorded when all sleepers are held at 1 kN load to ensure that they are in contact with the actuator. The readings are manually taken before the cyclic loading begins and after cycles 100, 1,000, 2,000, 5,000, 10,000, 20,000, 50,000, 100,000, 150,000, 200,000, 250,000, 300,000, 400,000, 500,000, . . ., and 1,000,000. After the test, the painted ballast column is collected for sieve analysis. 5 Result comparison: triaxial and railway test facility tests A simple analysis can be performed to estimate the stress conditions under the loading area of the RTF and to compare the results with the triaxial tests. This analysis is based on the findings from [17,18]. From Fig. 1, Shenton [18] found that the maximum contact pressure at sleeper base was approximately 200–250 kPa. According to Selig and Waters [17], the observed horizontal stress in ballast fluctuated but eventually reached 30 kPa as shown in Fig. 4. In other words, this is a typical residual horizontal stress in ballasted tracks. From both findings, it is reasonable to simulate the condition of ballast under traffic loading in the RTF by a cyclic triaxial test with constant confining stress of 30 kPa and maximum axial stress of 200–250 kPa. The values of the confining

Load (kN)

123

400
2.0 C10/2.0 C30/2.0 C60/1.5 C60/2.0 RTF

B. Aursudkij et al.

1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0 20,000 40,000 60,000 80,000 100,000

No. of cycles

Fig. 23 Permanent axial strains from cyclic triaxial tests and railway test facility

0.9 0.8 0.7

% passing

0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 5 10 15 Size (mm) 20 25

C10/2.0 C30/2.0 C60/1.5 C60/2.0 RTF

Fig. 24 Particles smaller than 22.4 mm from cyclic triaxial tests and railway test facility

stress and maximum axial stresses result in (q/p )max from 1.96 to 2.13. This stress condition matches the condition of test C30/2.0 (30 kPa confining stress and (q/p )max of 2.0). The permanent axial strains and particle size distribution from the cyclic triaxial tests are compared to the permanent axial strain of the middle sleeper and the particle size distribution of the painted ballast from the RTF as shown in Figs. 23 and 24. The permanent axial strain from the RTF was obtained by dividing the settlement by the ballast height prior to the test (300 mm). It should be noted that since the first data point of strain starts at cycle 100 in the RTF, the strains of the first 100 cycles from all cyclic triaxial tests were removed to compare the results on the same basis. Also, the strain from the RTF after cycle 100,000 is not included in the plot as the triaxial tests were performed for only 100,000 cycles. It can be seen from both figures that the results from test C30/2.0 are very close to the RTF results as expected. 6 Conclusions A new large-scale triaxial test apparatus for testing a cylindrical sample of railway ballast has been developed. The apparatus records the sample volume change by mea-

suring the difference between the level of pressurised water surrounding the sample and the fixed level of water under the same pressure by the differential pressure transducer. As the system cannot detect the volume change instantaneously, the loading speed must be sufficiently low. However, when the differential pressure transducer was not available, the volume change was measured by recording the change in level of pressurised water surrounding the sample instead. Both monotonic and cyclic triaxial tests were performed on limestone samples with grading within the standard specification. Three monotonic tests with constant confining stresses of 10, 30, and 60 kPa were performed. It was found that the samples reached their peaks strength when they were loaded until approximately 12% axial strain. The sample with lowest confining stress gave the largest amount of dilation but the smallest amount of breakage. Four cyclic tests were performed. Three were performed with (q/p )max of 2.0 and each with confining stresses of 10, 30, and 60 kPa. The other test was performed under confining stress of 60 kPa and (q/p )max of 1.5. It was found in the cyclic tests with the same (q/p )max that the permanent axial and volumetric strains increased with increasing confining stress. However, in the test with 60 kPa confining stress and (q/p )max of 2.0, the sample dilated for the first few cycles and then contracted. This might be because the maximum axial stress equalled the peak strength of the sample under monotonic loading, when large amounts of dilation occurred. The resilient modulus was found to follow the K − θ model. Also, even with small deviatoric stress, the breakage from the test with 10 kPa and (q/p )max of 2.0 was comparable to the other tests with much larger deviatoric stress except for the test with 60 kPa and (q/p )max of 2.0. This is because the test with 10 kPa and (q/p )max of 2.0 showed dilative behaviour. A RTF has been built in a concrete pit and comprises subgrade material, ballast, and three sleepers. The sleepers are loaded with out of phase sinusoidal loading to simulate traffic loading. The settlement can be recorded by the displacement transducers in the loading hydraulic actuators. A column of painted ballast with 300-mm diameter is placed under the front-end rail seat of the middle sleeper to aid ballast collection for sieve analysis after the test. A simple analysis based on typical observed residual horizontal stresses and bearing stresses beneath sleepers gives rise to a stress regime consistent with that used in the triaxial tests. According to the analysis, the stress conditions under the loading area of the RTF are similar to a triaxial test with 30 kPa confining stress and (q/p )max ratio from 1.96 to 2.13. The permanent axial strain and particle size distribution from the triaxial test with confining stress of 30 kPa and (q/p )max ratio of 2.0 were found to be very close to the RTF. This calculation gives some confidence that the stresses applied in the triaxial tests were appropriate to in-situ conditions.

123

Permanent axial strain (%)

Cyclic loading of railway ballast under triaxial conditions and in a RTF

401 10. Lekarp, F., Isacsson, U., Dawson, A.: State of the art I: resilient response of unbound aggregates. J. Transp. Eng. ASCE 126(1), 66–75 (2000) 11. Lim, W.L.: Mechanics of railway ballast behaviour. Ph.D. dissertation, University of Nottingham (2004) 12. McDowell, G.R., Bolton, M.D., Robertson, D.: The fractal crushing of granular materials. J. Mech. Phys. Solids 44(12), 2079–2102 (1996) 13. Ng, C.W.W., Zhan, L.T., Cui, Y.J.: A new simple system for measuring volume changes in unsaturated soils. Can. Geotech. J. 39(3), 757–764 (2002) 14. Railtrack Line Specification: RT/CE/S/006 Issue 3. Track Ballast (2000) 15. Raymond, G.P., Bathurst, R.J.: Repeated-load response of aggregates in relation to track quality index. Can. Geotech. J. 31(4), 547–554 (1994) 16. Selig, E.T., Alva-Hurtado, J.E.: Predicting effects of repeated wheel loading on track settlement. In: Proceeding of the 2nd International Heavy Haul Conference, Colorado Springs, pp. 476–487 (1982) 17. Selig, E.T., Waters, J.M.: Track Geotechnology and Substructure Management. Thomas Telford, London (1994) 18. Shenton, M.J.: Deformation of railway ballast under repeated loading conditions. British Railways Research and Development Division, UK (1974) 19. Skoglund, K.A.: A study of some factors in mechanistic railway track design. Ph.D. dissertation, Norwegian University of Science and Technology (2002)

References
1. Aursudkij, B.: A laboratory study of railway ballast behaviour under traffic loading and tamping maintenance. Ph.D. dissertation, University of Nottingham (2007) 2. Awoleye, E.O.A.: Ballast type—ballast life predictions. Derby, British rail research LR CES 122 (1993) 3. Bishop, A.W., Green, G.E.: The influence of end restraint on the compression strength of a cohesionless soil. Geotechnique 15, 243–266 (1965) 4. Brown, S.F., Brodrick, B.V., Thom, N.H., McDowell, G.R.: The Nottingham railway test facility. In: Proceedings of the Institution of Civil Engineers: Transport, vol. 160, No. TR2, pp. 59–65 (2007) 5. Duncan, J.M., Dunlop, P.: The significance of cap and base restraint. J. Soil Mech. Found. Div. Proc. ASCE 94(SM1), 271–290 (1968) 6. Fair, P.: The geotechnical behaviour of ballast materials for railway track maintenance. Ph.D. dissertation, University of Sheffield (2003) 7. Indraratna, B., Ionescu, D., Christie, H.D.: Shear behaviour of railway ballast based on large-scale triaxial tests. J. Geotech. Geoenviron. Eng. 124(5), 439–449 (1998) 8. Indraratna, B., Lackenby, J., Christie, D.: Effect of confining pressure on the degradation of ballast under cyclic loading. Geotechnique 55(4), 325–328 (2005) 9. Key, A.: Behaviour of two layer railway track ballast under cyclic and monotonic loading. Ph.D. dissertation, University of Sheffield (1998)

123