Introduction It has been shown (McDowell et al, 2004a) that the behaviour of different ballasts varies enormously.
McDowell et al (2004a) examined the performance of each of four different ballasts in a box test simulating train loading and rearrangement caused by tamping and were able to correlate the results in terms of settlement and degradation with simple index tests such as the Los Angeles Abrasion value and the Micro-Deval Attrition value (as defi ned in BS EN 13450: 2002 (British Standards Institution, 2002)).
In that paper, one ballast was clearly inferior to the others in terms of performance. In a subsequent paper, McDowell et al (2004b) investigated using the same box test method, as to whether a weak ballast could be mixed with a strong ballast to give a ballast mixture which meets current specifications, and how much weak ballast could be included without adversely affecting the performance of the mixture to a significant extent.
This would allow the selection of an appropriate amount of weaker, cheaper aggregate for inclusion in rail trackbeds, reducing ballast cost.
This paper follows on from McDowell et al (2004b) and examines the performance of ballast reinforced with geogrid.
Two alternative ballasts have been considered. Ballast A (strong) is a granite comprising mainly plagioclase (35%), quartz (30%) and alkali feldspar (20%). Ballast B (weak) is a granodiorite from a different quarry containing 50% plagioclase which has been mainly altered to clay and mica, 30% quartz and 10% hornblende (Large, 2003).
Performance is measured in terms of permanent settlement, stiffness and degradation in box tests. Three alternative geogrids are used. The influence of the location of the geogrid within the box is also studied.
As for the ballast tested by McDowell et al (2004b) in the previous Ground Engineering paper, tamping was simulated using a Kango hammer after certain numbers of cycles of load. The infl nce of the geogrid on increasing the time period between tamping (ie reducing maintenance frequency) was also investigated.
Ballast properties In the UK, until recently, the Wet Attrition Value (WAV) and Aggregate Crushing Value (ACV) have been used to assess the potential performance of ballast (Railtrack, 2000).
The Los Angeles Abrasion (LAA) test and Micro-Deval Attrition (MDA) tests have now been adopted in the new European specification BS EN 13450 (2002) (British Standards Institution, 2002).
McDowell et al (2003, 2004a) showed that because the ACV test is performed on 10mm-14mm particles, this test is of little use because of the size effect on particle strength, and track ballast consists of much larger particles. The WAV, LAA and MDA tests each involve measuring the degradation of ballast in a revolving drum. Table 1 shows the WAV, LAA and MDA values for ballasts A and B. The LAA values and MDA values were determined according to the Standards BS EN 1097-1 (BSI, 1998) and BS 1097-2 (BSI, 1996) respectively.
In addition, the flakiness index and particle length index are given according to the BS EN 13450 (2002) specifi cation. A full description of the tests is given in McDowell et al (2004a).
It can be noted that Appendix E of the Railtrack Line Specification (Railtrack 2000) states that from 1 April 2005, the LAA value must not exceed 20, and the MDA value must not exceed 7. However, this specification has been adopted early and is already now in use. Ballast B therefore clearly does not meet the new specification, but the question arises as to whether a geogrid may be able to improve its performance.
Box tests McDowell et al (2004a) describe a box test for simulating train loading on ballast and rearrangement by tamping. The box, shown in Figure 1, has a length of 700mm, width of 300mm and height of 450mm, and can be envisaged as representing a section of ballast underneath the rail seat as shown in Figure 2. It is made mainly of case-hardened steel with one side (a longer side) made of reinforced Perspex, so that degradation can be observed during the test.
The base of the box is made of wood and a 10mm thick rubber sheet was placed between the ballast and the wood to replicate a typical stiffness of sub-ballast and subgrade.
To compare performance consistently in the box tests, ballast samples were sieved to check that the initial gradings were consistent and conformed to the new BS EN 13450 (2002) specifi cation for track ballast grading (Table 2).
The box tests were conducted with wet ballast because track ballast is often in the wet condition, and ballast in the wet condition is considered to be more critical (Selig and Waters, 1994). Thus, each ballast sample was soaked in water for 48 hours to ensure that all ballast particles were fully saturated before pouring into the box. A controlled amount of water was also added at various stages during the testing, as described later.
Tests on ballast A First, the ballast was tested unreinforced. It was found by Lim (2003) that placing the ballast in three layers of 100mm was a suitable method of sample preparation. The general procedure used for this investigation was therefore the same, checking that no large voids were present in each layer.
Once the first 300mm of ballast had been placed the sleeper was then positioned on top ensuring it was placed in the centre of the box.
The areas between the sleeper and the edge of the box were then filled to bring the level of the ballast to the top of the box (ie crib ballast).
It was very difficult to ensure that the sleeper level was exactly level with the top of the box, so the initial sleeper level was recorded in each test.
To restrain the sleeper from moving horizontally or tilting, a steel piston was attached to it and a guide plate for the loading piston attached to the box frame to guide the sleeper during cyclic loading.
This sleeper guiding mechanism is shown in Figure 3.
Ballast settlement was measured by measuring the displacement of the top side of the bottom flange of the sleeper using an LVDT. The ballast was loaded cyclically with a sinusoidal load pulse with minimum load of approximately 3kN and maximum load of approximately 40kN (equivalent to an axle load of 20t to 25t) for 100,000 cycles, at a frequency of 3Hz.
The ballast tamping process was simulated in the box test by inserting a one inch wide chisel into the ballast using a Kango hammer (see McDowell et al, 2004a).
Before the chisel was inserted into the ballast, the sleeper was lifted until the top of the sleeper was level with the top of the box. At this level, the bottom of the sleeper was 300mm from the bottom of the ballast layer. Thus, the ballast could be tamped to regain (approximately) its original thickness. The chisel was then inserted towards the ballast underneath the sleeper through a guide hole, 160mm from the sleeper edge, at an angle of approximately 10º to the vertical. Figure 3 shows the guide holes: one on each side of the sleeper.
Three 'tamps' were applied at each side of the sleeper in a number of different locations (95mm, 150mm and 205mm from the Perspex wall).
Each insertion took approximately 2s and tamping was conducted at 100; 500; 1,000; 5,000; 10,000 and 50,000 cycles.
Prior to tamping, two litres of water were poured evenly on each side of the sleeper to ensure that the ballast remained wet during the test. Water and fines which drained out of the box were retained on an aluminium tray underneath the box.
To ensure that the same amount of ballast was available to be 'pushed' underneath the sleeper, and to maintain the correct amount of crib ballast, additional ballast was added to the top of the box after tamping.
Figure 4 shows the settlement of the unreinforced ballast as a function of number of cycles. The spikes indicate the tamping process when the sleeper was lifted to be level with the top of the box. Two tests were performed on different samples of ballast A to check repeatability, shown in Figure 4. The results are also consistent with results of box tests by Lim on the same ballast (Lim, 2003). Figure 5 shows the ballast stiffness (sleeper bearing stress/defl ection) measur ed for each of the two samples - these results are also reasonably repeatable.
Geogrid-reinforced ballast Reinforced samples were created using alternative geogrids. Figure 6 shows a small section of a typical punched and stretched geogrid.
The properties of the three grids are given in Table 3. Grid 1 has an aperture size of 39mm. Grid 3 has the same aperture size (65mm) as Grid 2, but is thicker and therefore stiffer and stronger. The geogrid must remain outside of the tamping area. Thus, tests were initially performed with the grid at 100mm from the bottom of the ballast layer.
Figure 7 shows the sleeper settlements for the three different geogrids. It is clear that Grids 2 and 3 give better performance than Grid 1, with Grid 3 performing slightly better than Grid 2 after a large number of cycles.
It should be noted that for the test using Grid 1, after the test and removal of the ballast, there was signifi cant damage to the geogrid.
The geogrid had split down the central rib measuring a distance of about 150mm. This breakage had occurred directly beneath the sleeper where the highest load was applied. Grid 2 was also damaged with slight splitting present. No damage had occurred to Grid 3, which gave a similar performance to Grid 2. Figure 8 shows the sleeper settlement of the unreinforced ballast, together with the results from Figure 7(a) and (b).
Figure 9 shows stiffness as a function of number of cycles for the unreinforced ballast A compared to the ballast reinforced with each of the three grids placed 100mm from the base of the ballast layer. Grid 3 gives the highest stiffness (but only just higher than Grid 2), though the differences between the results are only of the same order as the variability evident in Figure 5 for the unreinforced ballast.
Figure 10 shows the result of placing Grid 3 at 200mm from the base, compared with placing the grid at 100mm from the base. The figure also shows the result of using grids at both locations.
The sample with the grid at 100mm from the base performs best, showing that the function of the geogrid is to limit tensile strains, as the ballast-geogrid composite system is acting as a beam in bending. It was also found that the use of two grids gave a higher stiffness than using Grid 3 at 200mm from the base, or the unreinforced ballast, but not as high a stiffness as using Grid 2 or Grid 3 alone placed at 100mm from the base.
Effect on required maintenance An attempt was made to establish whether the use of a geogrid could reduce the required tamping frequency for ballast A. Figure 11 shows settlement as function of number of cycles for the ballast reinforced with Grid 3 at 100mm from the base, compared with the unreinforced case. The frequency of tamping for the reinforced ballast is half that for the unreinforced ballast. It can be seen that after 50,000 cycles, the performance of both ballasts is very similar. This suggests that geogrids might be useful in reducing the number of tamping maintenance operations required in track.
Box tests on ballast B A smaller number of tests were also conducted on ballast B. Figure 12 shows the settlement of the unreinforced ballast, compared to the unreinforced ballast A. Figure 13 shows the settlement of unreinforced ballast B compared with the ballast reinforced with Grid 3 placed 100mm from the base.
The use of the geogrid improves the performance, but the reduction is not as marked as that for ballast A (see Figures 7 and 8). Figure 14 shows the stiffness of unreinforced ballast B, together with that for the ballast reinforced with Grid 3 at 100mm from the base, and unreinforced ballast A.
The stiffness of ballast B is much lower than ballast A, and the use of a geogrid only increases the stiffness marginally. McDowell et al (2004a) noted that ballast B was much more susceptible to degradation than ballast A, so it appears that for much more crushable material, the use of a geogrid will be less effective. To confirm this, the degree of particle breakage was studied for the two ballasts.
Particle breakage Figure 15(a) shows the initial particle size distribution for ballast A, together with the changes in the grading in the box as a whole, and in a column 100mm deep directly underneath the sleeper. Figure 15(b) shows the results for ballast A reinforced with Grid 3 placed 100mm from the base. Similar results were obtained for the other geogrids.
Figure 16(a) shows the corresponding particle size distributions for unreinforced ballast B and the results for ballast B reinforced with Grid 3 placed 100mm from the base are shown in Figure 16(b). Clearly more breakage occurs for ballast B.
Table 4 shows Hardin's Total Breakage Bt (Hardin, 1985), which measures the area swept out by the particle size distribution (with particle size plotted on a logarithmic scale), for each of the tests. For ballast A, the inclusion of Grid 2 or Grid 3 reduces the amount of breakage observed directly under the sleeper, but not for the box as a whole (and more breakage was observed when Grid 3 was placed at both 100mm and 200mm from the base simultaneously). The opposite effect is observed for ballast B.
The differences in the measured values, however, between the reinforced and unreinforced samples, are small. It seems that the geogrid does not influence the degree of particle breakage much, so this is difficult to quantify.
However, the accumulation of permanent strains will surely depend on particle breakage and only small increases in particle damage will permit further permanent strains to develop. For a very weak ballast such as ballast B, it appears that the use of a geogrid will do little to combat particle breakage; hence the reduction in permanent settlement caused by the introduction of a geogrid will be limited.
Conclusions Box tests have been performed on each of two ballasts: unreinforced and reinforced, with the use of geogrids.
It appears that for much more crushable ballast, the use of a geogrid gives a minor improvement in performance. For the ballast of much better quality, the improvement is much more marked.
Three alternative geogrids were used; it was found that the grid with aperture size 65mm performed much better than that with a 39mm aperture size, for ballast graded according to the current specification. An increase in thickness of the grid also improved the performance slightly for the larger aperture size.
The effect of the geogrid was to increase stiffness only marginally, and with only a marginal effect on particle breakage, but to reduce permanent settlement signifi cantly. It was found that in the box tests, the ballast only needed to be tamped half as often if a geogrid had been installed at the appropriate location. It was found that placing the geogrid 100mm from the base gave a better performance than at 200mm from the base, or using geogrids at both locations.
This shows that the function of the geogrid is strain-limiting, as the ballast-geogrid composite system acts as a beam in bending. The paper has therefore highlighted the use of geogrids to improve ballast performance, with the possible benefi t of reducing the frequency of maintenance operations.
British Standards Institution (1996). BS EN 1097-1: Tests for mechanical and physical properties of aggregates - determination of the resistance to wear (micro-Deval).
British Standards Institution (1998). BS 1097-2:
Tests for mechanical and physical properties of aggregates - methods for the determination of resistance to fragmentation.
British Standards Institution (2002). BS EN 13450 (2002): Aggregates for railway ballast.
Hardin BO (1985). Crushing of soil particles.
Journal of Geotechnical Engineering, ASCE 111, No. 10, pp1177-1192.
Large D (2003). Petrographic description of railway ballast samples. University of Nottingham (confidential report).
Lim WL (2004). Mechanics of railway ballast behaviour. PhD dissertation, University of Nottingham.
McDowell GR, Lim WL and Collop AC (2003).
Measuring the strength of railway ballast.
Ground Engineering 36, No1, pp25-28.
McDowell GR, Lim WL, Collop AC, Armitage R and Thom NH (2004a). Comparison of ballast index tests with performance under simulated train loading and tamping. Proceedings of the Institution of Civil Engineers - Geotechnical Engineering 157 GE(3), 151-161.
McDowell GR, Buchanan J and Lim WL (2004b). Performance of ballast mixtures.
Ground Engineering 37, No10, 28-31.
Railtrack (2000). Railtrack Line Specification RT/CE/S/006 Issue 3: Track ballast.
Selig ET and Waters JM (1994). Track geotechnology and substructure management. Thomas Telford: London.