Deformation characteristics of fresh and fouled ballasts subjected to tamping maintenance

Soils and Foundations 2016;56(4):652–663

HOSTED BY

The Japanese Geotechnical Society

Soils and Foundations www.sciencedirect.com journal homepage: www.elsevier.com/locate/sandf

Technical Paper

Deformation characteristics of fresh and fouled ballasts subjected to tamping maintenance Janaka J. Kumaraa,n, Kimitoshi Hayanob a b

Department of Civil Engineering, Yokohama National University, 79-5, Tokiwadai, Hodogaya-ku, Yokohama 240-8501, Japan Department of Urban Innovation, Yokohama National University, 79-5, Tokiwadai, Hodogaya-ku, Yokohama 240-8501, Japan Received 8 June 2015; received in revised form 3 March 2016; accepted 25 March 2016 Available online 10 August 2016

Abstract A number of train passes over time induces large settlement on ballasted railway tracks. Tamping maintenance application, which is practised worldwide, can lead to additional subsequent track settlement from the disturbances caused by the tamping tools. In this study, a series of model tests on a scaled-down ballasted railway track was conducted to examine the settlement characteristics of ballast subjected to tamping maintenance application. The particle movements during cyclic loading were tracked using a particle image velocimetry (PIV) approach to study the local deformations induced by the tamping tools. The results revealed that the intrusion of fouling material and subsequent maintenance application altered the settlement characteristics significantly. The PIV results revealed that the top ballast is loosened by the tamping tools. It was also found that the strength properties of ballast deteriorate with the fouling of the material. Notably, the strain hardening behaviour of ballast is weakened when the material undergoes 30% fouling or more according to the fouling index, FIp (i.e., an indication of degree of ballast fouling). The results of this study suggest that track maintenance should involve the tamping of fouled ballast before FIp reaches 30%. & 2016 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Keywords: Ballasted railway track; Fouled ballast; Fouling index; Model test; Particle image velocimetry; Settlement characteristic; Shear strain; Tamping application; Triaxial compression test

1. Introduction Most of the railway tracks throughout the world are ballasted railway tracks, which are preferred over concrete railway tracks. Many aspects, including simplicity in construction and maintenance works, have led to the wide use of the ballasted railway tracks. Track settlement occurs gradually over time with long-term service due to the large number of n Correspondence to: Department of Civil Engineering, Tokyo University of Science, 2641, Yamazaki, Noda, Chiba, 278-8510, Japan. Fax: þ 81 04 7124 2150. E-mail addresses: [email protected] (J.J. Kumara), [email protected] (K. Hayano). Peer review under responsibility of The Japanese Geotechnical Society.

train passes. Excessive settlement can cause poor passenger comfort, speed restriction and potential derailments. As reported in Selig and Waters (1994), ballast contributes to most of the substructure settlement (see Fig. 1) although the main function of the ballast layer is to restrain the track geometry. It has also been reported that a major portion of the track maintenance budget is spent on the substructure (Ionescu et al., 1998; Raymond et al., 1978). In ballasted railway tracks, ballast fouling occurs when the finer materials (i.e., the materials smaller than fresh ballast) mix with fresh ballasts due to heavy repeated train loads. It is expected that fresh or clean ballast is used in the construction of a railway track with the fouling components not exceeding 2% (Indraratna and Salim, 2005). Generally, finer materials

http://dx.doi.org/10.1016/j.sandf.2016.07.006 0038-0806/& 2016 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

Fig. 1. Substructure contributions to settlement (modified after Selig and Waters, 1994).

come from underlying layers as well as being produced by particle crushing (Hossain et al., 2007; Indraratna et al., 2011a; Selig and Waters, 1994). However, it should be noted that the source of fouling materials may be differ according to the nature of railway tracks and the source of original ballast (Selig and Waters, 1994; Feldman and Nissen, 2002). As reported by Indraratna et al. (2002a) and Indraratna and Salim (2005), a wide variety of material is used as ballast around the world: economic and environmental issues are among the considerations when sourcing ballast. Finer material intrusions alter the original particle size distribution (PSD) of ballast, from uniform gradation to less uniform gradation, depending on the amount of fouling materials mixed with fresh ballast. The altered gradation of fouled ballasts results in settlement characteristics different than those of fresh ballast (Indraratna et al., 2006). Huang and Tutumluer (2011) and Indraratna et al. (2011a) discussed the effects of ballast fouling on geo-grid reinforced ballasts and found that the strength properties of fouled ballasts are affected by the degree of ballast fouling. Cambio and Ge (2007) have also reported that the strength properties of fouled ballasts are significantly affected by the intrusion of fouling materials into fresh ballast. Once the railway track settlement reaches the allowable limit, maintenance is necessary to return the railway track to its original position since differential settlement can result in many problems, including derailment. Tamping application is practised worldwide as the main maintenance application (Indraratna and Salim 2005). One of the problems arising from the use of tamping application is that the tamping tools can loosen the top ballast and also induce particle crushing depending on penetration depth of the tamping tools into the ballast layer. As discussed briefly in this paper, ballast deteriorates due to many reasons, including breakage of angular corners and sharp edges (i.e., resulting into finer materials), an infiltration of fines from the surface and finer materials pumping from the underlying layers under train loading (Indraratna et al., 2011a; Selig and Waters, 1994). As a result of these actions, ballast becomes fouled, less angular and its shear strength is reduced

653

(Indraratna et al., 2005). Angular particles have better load spreading properties due to better interlock than rounded particles (Holtz and Gibbs, 1956; Indraratna et al., 1998; Leps, 1970). Due to the increasing importance of the ballasted railway tracks, a number of studies on the deformation characteristics of fouled ballasts have been conducted in the recent past (Indraratna et al., 2001; Lackenby et al., 2007; Raymond and Bathurst, 1994). However, the effects of ballast fouling or tamping application itself on settlement characteristics have scarcely been studied, resulting in a lack of knowledge on the settlement characteristics of fouled ballasts subjected to maintenance applications. Generally, the degree of ballast fouling is described by a fouling index (Ionescu, 2004; Selig and Waters, 1994). Later, Feldman and Nissen (2002) proposed a volume-based fouling index due to variations in specific gravities of fresh ballast and fouling materials. Recently, Indraratna et al. (2010) and Indraratna et al. (2011c) modified the volume-based fouling index proposed by Feldman and Nissen (2002) due to its practical limitations and introduced slightly different fouling indexes. The fouling indexes proposed by Feldman and Nissen (2002) and Indraratna et al. (2010) are important when the fouling material has a low specific gravity compared to fresh ballast. The volume-based fouling indexes are also very useful when studying the drainage characteristics of ballasted railway tracks since they take the voids in the ballast particles into account more appropriately than mass-based fouling indexes. The volume-based approach proposed by Indraratna et al. (2010) has later been used in laboratory and numerical approaches as well (Indraratna et al., 2013, 2014). Since the specific gravities of the materials used in this study were approximately the same (e.g., 2.75 of fresh ballast compared to 2.65 of fouling material) and because of its simplicity of use, the fouling index reported by Ionescu (2004) is still used to classify the fouled ballasts. The original equation proposed by Ionescu (2004) was modified to fit into the scale-down materials used in this research as given in Eq. (1). In the equation proposed by Ionescu (2004), they were given as P0.075 and P13.2, which were simply modified to P0.015 and P2.64 respectively taking account of the scaled-down ballast (i.e., 1/5th scale) used in this study. According to Ionescu (2004), ballast with FIp o 2 is considered clean (or fresh) ballast, when 10r FIp o 20 the ballast is considered moderately fouled, when 20r FIP o 40 it is consodered fouled ballast, and when FIp 4 45 the ballast is considered highly fouled. FI p ¼ P0:015 þ P2:64

ð1Þ

where FIp is fouling index, P0.015 and P2.64 are percentages passing at 0.015 and 2.64 mm respectively. In this research, settlement characteristics of fouled ballasts with different fouling indexes were studied by conducting model tests on a scaled-down railway track. The field tamping application was simulated on the model trackbed using a simple tool and its effects on subsequent settlement were examined using a particle image velocimetry (PIV) approach. The PIV method was used to track the particle movements and

654

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

the local deformations induced by the tamping tools. The deformation characteristics of the fouled ballasts were also studied by triaxial compression tests. All in all, this study provides a practical solution for fouled ballasts in terms of a fouling index to ensure the effectiveness of tamping maintenance. 2. Ballasted railway tracks and maintenance applications

Sleeper

Rail

3. Materials and testing methods In the initial part of this research, model tests on a scaleddown railway track were conducted to study the settlement characteristics of fresh and fouled ballasts subjected to tamping maintenance. In the next part, triaxial compression tests were conducted to study the deformation characteristics of fresh and fouled ballasts. 3.1. Model tests

Crib

Subgrade

Substructure Superstructure

Due to the deterioration of the ballast, ballasted railway tracks are restored quite regularly compared to their concrete counterparts. Generally, railway tracks are constructed with a 250 mm thick ballast layer, as shown in Fig. 2. Mainly, tamping and stoneblowing applications are practiced to restore the settled railway tracks. According to Wright (1983), both the tamping and stoneblowing applications result in ballast breakage during the process of insertion into the ballast layer. Fig. 3 shows the sequence of the tamping application. As shown in Fig. 3b, the settled sleeper is first lifted to the original position. Then, the inserted tamping tools are tilted laterally to bring the ballast from the sides of the sleeper to the space below it, as shown in Fig. 3c. The final space between the sleeper and ballast surface is filled by adding new ballast (see

Fig. 3d). The tamping tools loosen the top ballasts according to the depth that they penetrate into the ballast layer. As reported by Selig and Waters (1994), the lifts of the sleepers can be large as 30–40 mm. Therefore, tamping tools can penetrate more than 90–100 mm into the ballast layer if they are penetrated into one size of a ballast particle (i.e., around 60 mm). The density of the top ballasts underneath the sleepers cannot be brought to the original design level because the lack of space during maintenance works makes it impossible to use compaction machines. The loosening of the top ballasts, therefore, may be a result of tamping maintenance, while the bottom ballasts are densified due to the number of train passes in the pre-maintenance works. It should be noted that the practical issues of stoneblowing maintenance, notably the development of a two layer system and particle crushing, have been well-documented (Anderson et al., 2001; Wright, 1983).

Ballast

250mm

Sub ballast

100mm

Placed soil Natural ground

Fig. 2. Schematic diagram of ballasted railway track.

A scaled-down (i.e., a one-fifth of the field scale) model trackbed was constructed in a sand box with the interior dimensions of 800 mm length (i.e., front side), 304 mm width and 300 mm height as shown in Fig. 4. The front wall of the sand box is made of glass to facilitate capturing of the images during the loading process (see Fig. 4a). A duralumin footing, 48 mm wide and 290 mm long, was used as the sleeper. In the model tests, the ratio of sleeper width to ballast thickness was

Fig. 3. Sequences of tamping application; (a) settled sleeper, (b) lifting the settled sleeper to the original position, (c) inserting tamping tools, (d) filling the space underneath sleeper by laterally moving ballasts and (d) after tamping application (modified after Selig and Waters, 1994).

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

655

Table 1 Physical properties of ballast and fouling material. Property

Ballast

Fouling material

Maximum particle size, D100 (mm) Mean particle size, D50 (mm) Minimum particle size, D0 (mm) Coefficient of uniformity, Cu Coefficient of curvature, Cc Maximum dry density, ρd,max (kg/m3) Minimum dry density, ρd,max (kg/m3) Maximum void ratio, emax Minimum void ratio, emin

11.2 7.9 4.8 1.51 0.98 1549 1304 1.028 0.712

2.0 1.5 1.2 1.30 0.95 1705 1514 0.751 0.556

Fig. 4. (a) Front and (b) side view of schematic diagram of the model apparatus (not to scale).

designed to replicate field conditions (i.e., approximately, 1.0). Therefore, the stress distribution zone of the model tests and field conditions were maintained equally. The axial displacement was measured using two external displacement transducers placed at the front and the back of the sleeper as shown in Fig. 4b. 3.2. Sample preparation Crushed stone, approximately 1/5th the size of the field ballast, was used as railway ballast. The field ballast and the crushed stone tested in this study are generally produced from the same source. At the crushing plants, large size aggregates (e.g., used as field railway ballast) and small size aggregates (e.g., used in concrete mixtures or similar sizes used in our experiments) are produced using different crushing rates. As reported by Marschi et al. (1972) and Indraratna et al. (1998), scaled-down particles may produce a slightly larger internal frictional angle, ϕ (e.g., an increase of 2–3 degree of ϕ when the particle size is decreased from 39 to 7 mm). Janardhanam and Desai (1983) reported that particle size does not significantly influence ballast strains at various stress levels. Vallerga et al. (1957) indicated that internal friction angle heavily depends on the particle shapes. As reported by Vallerga et al. (1957) and Indraratna and Salim (2005), the most importance factors of the ballast are the parent rock and the particle shapes. Recently, Pen et al. (2013) tested railway ballasts with different particle sizes varying from 9.5 to 62.5 mm to study the angularity of particles. As reported in Pen et al. (2013), due to a negligible variation in angularity, scaled-down materials are appropriate substitutes for testing purposes. Cyclic triaxial testing conducted using three parallel gradations of ballast (with the maximum size of 63.5, 38.0 and 19.0 mm) by Sevi and Ge (2012) reported that the particle shape and relative density are the most important parameters. As discussed by many researchers, the most important factors of ballast are the parent rock and angularities of the particles. The increase in the

Fig. 5. Particle size distribution of fresh and fouled ballasts.

frictional angle due to a possible reduction in size is quite negligible (e.g., an increase of 2–3 degree from 50–60 degree of fresh ballast). Since we used a reduced-size crushed stone of similar angularities to field ballast, it is our belief that the results match the behaviour of field ballast appropriately. The physical properties of ballast (i.e., crushed stone) and fouling material (i.e., medium sized sand, referred as M sand hereafter) are given in Table 1. Medium sized sand was selected as the fouling material due to the inability of the particle image velocimetry (PIV) method adopted in this study to track particle movements of smaller particles than medium sized sand. Particle size distribution (PSD) curves of fresh ballast (i.e., crushed stone), fouling material (i.e., M sand), fouled ballasts (i.e., sand-crushed stone mixtures) along with the Japanese standard ballast gradation (i.e., 1/5th of the field scale) are shown in Fig. 5. The ballast thickness was selected as 50 mm (i.e., 1/5th of the 250 mm thickness used as field ballast layer). The model trackbed was prepared in five layers, each 10 mm in thickness, to achieve uniform compaction with depth. The model trackbed was prepared with 80% relative density to simulate the field railway track conditions. Also, the model trackbed was prepared using dry ballast to simulate the field conditions. A hand vibrator was used to compact the model trackbeds (see Fig. 6a and b). The hand vibrator is attached with a thick rubber plate as shown in Fig. 6a. The rubber

656

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

Curved bottom surface

Wooden plate Ballast layer

Fig. 6. (a) Photograph of the hand vibrator and (b) schematic diagram of compaction process.

Table 2 Fouling index, percentage of fouling materials and dry density of the test samples. Test sample Fouling index, FIp

Percentage of fouling materials

Dry density, ρd (kg/m3)

Case Case Case Case Case Case

0 15 30 50 70 100

1519 1698 1829 1929 1788 1684

1 2 3 4 5 6

0 15 30 50 70 100

plate prevents particle crushing during the trackbed preparation. A wooden plate placed on the ballast surface was subjected to vibration by the hand vibrator to achieve the required relative density uniformly. The percentages of fouling materials in the fouled ballast samples were selected based on the fouling index given in Eq. (1). It has been reported that intrusion of fouling materials into ballast is not significant until the percentage of fouling materials reaches 10% or more (Selig, 1985). Therefore, we selected fouled ballasts with over 15% of fouling materials. It should be noted that, coincidentally, the fouling index, FIp adopted in this research (i.e., Eq. (1)) is equal to the weight-based percentage of fouling materials in the fouled ballast samples. The values of FIp, percentage of fouling materials and dry density of the test samples are given in Table 2. In Table 2, Case 1, Case 6 and Cases 2–5 represent fresh ballast, fouling material and fouled ballasts respectively. 3.3. Void ratio determination The extreme void ratios, emax and emin (emax and emin are maximum and minimum void ratios respectively) were calculated using the respective dry densities (ρd,min and ρd,max respectively) as given in Eq. (2). The density tests were conducted according to the Japanese Industrial Standards (JIS 2009). In the minimum density test, the materials were poured into a mould maintaining a zero height using a small container. The materials were placed in ten equal layers. The test was conducted five times and the average value was taken as the final result. In the maximum density test, the materials were poured into the mould maintaining a zero height. The materials were placed in five equal layers and each layer was vibrated for three minutes. The test was conducted five times

and the average value was taken as the final result. Gs ρw 1 e¼ ρd

ð2Þ

where e is void ratio, Gs is specific gravity, ρw is water density and ρd is dry density. 3.4. Cyclic loading Cyclic loading in vertical direction was applied to simulate the train passes. One hundred loading cycles each were applied in the pre- and post-tamping applications. Due to the manual process of loading control system, we had to limit the number of loading cycles to one hundred, although a higher number of loading cycles would need to be applied to simulate the field conditions ideally. In a field railway track, a uniform densification of ballasts occurs due to the stress rotation by a moving load. In contrast, the model tests included a point load (or a non-moving load). Therefore, it should be noted that a larger number of loading cycles would consolidate fouling materials under the sleeper more than the field conditions. For this reason a relatively smaller number of loading cycles was applied. There have been reports that the initial settlement of ballasted railway track is reached within a small number of loading cycles, after which the residual settlement continues (Hayano et al., 2013). Generally, the settlement after the first loading cycle has been considered as the initial settlement (Shenton, 1975; Chrismer and Selig, 1993; Indraratna et al., 2001). Therefore, one hundred loading cycles was considered sufficient to study the settlement characteristics of ballasted railway tracks. As Barksdale (1972) reported, the permanent deformation is a linear function of a logarithm of the number of load cycles. Therefore, it should be noted that a higher number of loading cycles would result in a larger settlement. Cyclic loading was applied to the model trackbed through the sleeper at a constant displacement rate of 0.05 mm/s. The maximum amplitude of the applied stress, sv was maintained at 120 kN/m2 (i.e., approximately 70% of the failure stress, sv, max of fouling material under monotonic loading). The failure of ballast under cyclic loading is progressive and occurs at a smaller stress than static loading (Selig and Waters, 1994; Sevi and Ge, 2012). The vertical stress for the field ballasted railway tracks is slightly higher (e.g., around 200 kPa) than the vertical stress applied in this study (Indraratna et al., 2011b; Raymond and Bathurst, 1994). Since the model trackbed prepared of fouling material fails at a relatively smaller vertical stress than

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

657

Fig. 7. (a) Photograph of the tamping tool, (b) cyclic loading in pre-tamping application (TA), (c) TA and (d) cyclic loading in post-TA.

fresh and fouled ballasts, an equal and relatively smaller vertical stress was applied for all the trackbeds to avoid unnecessary data interpretation. We also noted that some researchers, including Brown et al. (2007), applied a similar vertical stress (i.e., 114 kPa) and found consistent results with the field conditions. Since the applied vertical stress is still around 60– 70% of the field conditions, the model tests should simulate the field conditions appropriately. However, it should be noted that a higher stress would induce a higher settlement (Brown and Hyde, 1975; Stewart, 1986).

Load cell Ballast layer Sleeper Camera

3.5. Tamping application Tamping was simulated by the tool shown in Fig. 7a. First, one hundred loading cycles were applied to the model trackbed through the sleeper (Fig. 7b). Then, the sleeper was lifted to the initial position (Fig. 7c). Next, the tamping tool was inserted about 15–20 mm (from the sleeper) into the model trackbed by the sides of the sleeper. After the tamping tool had reached the required depth, it was tilted horizontally several times to permit the ballast to move laterally. This procedure was followed at several locations until the voids between the sleeper and the ground surface were completely filled by ballast. Tamping was carried out each side, alternatively (i.e., after tamping was done for a small length on one side, the tamping tool was moved to the other side). After a series of interchanges of the tamping tool between the two sides, the procedure was completed. Finally, additional ballast (except in Case 6 (see Table 2) which was prepared using solely fouling material) was added to the ground surface near the sleeper to produce a flat ground surface (the fouling material was added in Case 6). After tamping application, again one hundred loading cycles were applied (Fig. 7d). The tamping application implemented in this research is similar to the field tamping application reported in Selig and Waters (1994), except for the vibration action. As it was necessary to avoid the unnecessary segregation of materials (between ballast and fouling material), no vibration action was applied. It should be noted that maintenance in the field is carried out repeatedly during a life cycle, whereas it was applied only one time in the model tests. Therefore, less particle crushing will occur in the model tests than in field conditions. However, since particle crushing was not included in this study and the same experimental conditions were followed for all the tests, the results given by the model tests should rather be considered as the effects of tamping tools on the settlement characteristics.

Fig. 8. Photograph of the image capturing process.

3.6. Particle image velocimetry (PIV) Since the displacement transducers are an indication of the total deformation rather than possible local deformation induced by tamping tools, a PIV approach was implemented to study the strain distributions beneath the railway sleeper. In this method, a digital single-lens reflex (DSLR) camera was fixed on a tripod in front of the test apparatus as shown in Fig. 8. A number of images were captured at frequent time intervals during each loading cycle using a remote controller. The images can be used to track the particle movements in vertical and horizontal directions (i.e., along the longer direction of the sand box). First, the images are analysed in Flow-PIV which is a PIV based 2D software. The Flow-PIV produces velocity vectors on a user-defined area on the ballast layer in an image. The displacement vector is then generated in the software called “FEA Visualizer” by using the velocity vector produced. It is noted that the displacements in the software are measured at nodal points (see Fig. 9) and the strains between two consecutive nodal points are calculated by Eqs. (3a) and (3b). The strains for the elements (or mesh) are calculated using the strains between the nodal points as given by Eqs. (4a) and (4b). Finally, the maximum shear strain is evaluated using vertical, horizontal and shear strains as given in Eq. (5). More details on the PIV analysis can be found in Kumara (2013).
 δh;1  δh;2 ð3aÞ εh;1  2 ¼ x

658

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

Fig. 9. Schematic diagram showing nodes and elements for strain calculations.

where εh,1–2 is horizontal strain between nodes 1 and 2, δh,1 is horizontal displacement at node 1, δh,2 is horizontal displacement at node 2 and x is grid size in horizontal direction (see Fig. 9).
 δv;1  δv;4 εv;1  4 ¼ ð3bÞ y where εv,1–4 is vertical strain between nodes 1 and 4, δv,1 is vertical displacement at node 1, δv,4 is vertical displacement at node 4 and y is grid size in vertical direction (see Fig. 9).
 εh;12 þ εh;45 εh;1 ¼ ð4aÞ 2 where εh,1 is horizontal strain at element 1, εh,1–2 is horizontal strain between nodes 1 and 2, and εh,4–5 is horizontal strain between nodes 4 and 5 (see Fig. 9).
 εv;14 þ εv;25 εv;1 ¼ ð4bÞ 2 where εv,1 is vertical strain at element 1, εv,1–4 is vertical strain between nodes 1 and 4, and εv,2–5 is vertical strain between nodes 2 and 5 (see Fig. 9). pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5Þ γ max ¼ ðεv  εh Þþ γ vh where γmax is maximum shear strain, εv is vertical strain, εh is horizontal strain and γvh is shear strain. 3.7. Triaxial compression test A laboratory triaxial compression test was conducted to study the deformation characteristics of fresh and fouled ballasts with similar conditions as those used in the model tests (i.e. materials 1/5th the field scale and 80% relative density). The samples were prepared under dry conditions to simulate the materials in a typical railway track. The tests were conducted according to the standard by the Japanese Geotechnical Society (JGS 1998) with an 80 kPa of confining pressure, sc. Because the high angularity of ballast required the use of a relatively thick membrane, a slightly larger sc was selected to eliminate any error which may have arisen from a small sc, including the membrane force error. It should be acknowledged that the railway track is subjected to a slightly smaller sc (e.g., 40–50 kPa, Sevi and Ge 2012; Tennakoon et al., 2012). We also noticed that the studies by Anderson and Fair 2008 (i.e., 40–140 kPa), Indraratna et al., 2007 (i.e., 50–300 kPa) and Indraratna and Nimbalkar 2011 (i.e., 10–100 kPa) among others have used even larger confining pressures to study the

Fig. 10. Settlement versus number of loading cycles in (a) pre- and (b) posttamping application.

behaviour of railway ballast. Therefore, the confining pressure may be as high as that applied in some applications. The effects of confining pressure on strength properties have been thoroughly studied by Lackenby et al. (2007) and Thakur et al. (2013) among others. Specimens with 100 mm diameter and 200 mm height were used for the tests. 4. Results and discussion The discussion here is divided into three main issues, namely the effects of particle size distribution (PSD), the effect of tamping application on the settlement characteristics and the deterioration of deformation characteristics of ballasts by intrusion of fouling materials. 4.1. Effects of particle size distribution (PSDs) on settlement characteristics Fig. 10a and b shows the results of settlement with number of loading cycles for the tested materials in pre- and posttamping applications respectively. As expected, settlement was reduced in the post-tamping application, which is due to the densification of the model trackbed by the number of loading cycles in the pre-maintenance application. However, it should be noted that larger settlement may occur in the post-tamping

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

659

affect settlement in both the pre- and post-tamping applications. Moreover, the relationship of S100 vs. FIp is quite similar to the relationships of emax and emin vs. FIp (see Fig. 11). The settlements in both pre- and post-tamping applications decrease with the addition of fouling material until 30% of FIp. Fig. 12 also indicates that the settlements in both pre- and post-tamping applications remain approximately the same between 30 and 50% FIp. Then, the settlements in both preand post-tamping applications start to increase with the addition of fouling material in a manner similar to that observed for the void ratios versus FIp. Therefore, it can be concluded that there is a clear relationship between the voids in crushed stone-fouling material mixtures and their settlement characteristics. Fig. 11. Maximum and minimum void ratios versus fouling index.

4.2. Effects of tamping application on settlement characteristics A number of empirical relationships of strain and loading cycles (Chrismer and Selig, 1993; Shenton, 1978) and settlement and loading cycles (Indraratna et al., 2002b; Sekine et al., 2005) have been reported in the literature. To study the effects of tamping application on settlement characteristics, the empirical equation proposed by Sekine et al. (2005) (i.e., Eq. (6)) was selected in this paper because it is capable of distinguishing between the initial and residual settlement characteristics.
 S ¼ c 1  e  αN þ βN ð6Þ Fig. 12. Settlement after one hundred loading cycles versus fouling index.

application if the effects of particle crushing (due to a large number of loading cycles and the vibration action of the tamping tools) are simulated in a manner similar to the field conditions. Fig. 10a and b clearly shows that an intrusion of fouling materials into the ballast alters the settlement characteristics from fresh ballast significantly, in both the pre- and post-tamping applications. As shown in Fig. 10a and b, the smallest settlement was observed when the specimens had values of 30 and 50% on the fouling index, FIp. This small observed settlement can be attributed to the variation of the void ratio with FIp as well (see Fig. 11). In Fig. 11, it can be seen that the smallest values of the void ratios were observed on the specimens with 30–50% of FIp. This is attributed to the filling of the voids between larger-sized ballast particles by smaller-sized fouling materials, which can indeed be justified by the results of Indraratna et al. (2011a), Kumara et al. (2012) and Lade et al. (1998) as well. The results also indicate that fresh ballast produces the highest value of the void ratios, and that these reduce with the addition of fouling material up to the densest packing. Then, the void ratios increase with the addition of fouling material until to the void ratios of the fouling material. Fig. 12 shows the results of final settlement after one hundred loading cycles, S100 vs. FIp for all the model trackbeds. It clearly indicates that particle size distributions

where S is settlement after N number of loading cycles, c is initial settlement, α is initial settlement rate and β is residual settlement rate. Fig. 13a–c shows the relations of c, α and β with FIp respectively. Fig. 13a–c clearly explain how these parameters are influenced by different magnitudes of fouling. As shown in Fig. 13a, initial settlement, c, reduced with FIp to 30–50% and then increased. This tendency also occurred on the railway tracks in the pre- and post-tamping applications. As shown in Fig. 13b and c, while the initial settlement rate, α, reduced with FIp, the residual settlement rate, β, increased with FIp in both the pre-and post-maintenance applications. However, the reduction of α with FIp is larger in the post-tamping application than in the pre-tamping application. That is, the initial settlement rate decreased significantly with FIp in the posttamping application (see Fig. 13b). The results also indicated that the residual settlement rate became higher in the posttamping application than in the pre-tamping application on the specimens characterised by more than 30% of FIp (see Fig. 13c). That is, the settlement in post-maintenance application reaches the allowable settlement within a smaller number of train passes compared to the pre-maintenance application when ballast is fouled by 30% or more of FIp. That is, in practical use, tamping application is effective for fouled ballasts with a maximum of 30% of FIp. As discussed by Selig and Waters (1994), the settlement rate should be faster on fouled ballasts and our findings satisfy this over 30% of FIp. Therefore, it is

660

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

Fig. 13. (a) Initial settlement, (b) initial settlement rate and (c) residual settlement rate versus fouling index.

Fig. 14. Displacement vector for fresh ballast in (a) pre- and (b) post-tamping application.

important to understand how tamping application affects the subsequent settlement. As reported earlier, a particle image velocimetry (PIV) analysis was conducted to understand the possible local deformations induced by the inserted tamping tools by tracking particle movements. Fig. 14a and b show the displacement vectors (of fresh ballast) obtained from the FEA Visualizer in the pre- and post-tamping applications respectively. As discussed in the section on particle image velocimetry, the displacement vectors shown in Fig. 14a and b were used to determine the strains on the elements by using Eqs. (3) and (4). The comparisons of the maximum shear strain distributions (Eq. (5)) in the pre- and post-tamping application for fresh and fouled ballasts with 30% of FIp are shown in Fig. 15a–d. As shown in Fig. 15a, fresh ballast showed a densification process beneath the sleeper in the pre-tamping application. Fig. 15b showed that the tamping application loosened ballast mostly at the sides of the sleeper, thereby indicating strain localisation in the ballast at the sides of the sleeper in the post-tamping application. As shown in Fig. 15c and d, fouled ballasts also showed similar behaviour, particularly in the post-tamping application, where strain localisation occurred in the ballast at

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

661

Fig. 15. Maximum shear strain distributions in (a) pre-, (b) post-tamping application of fresh ballast, (c) pre- and (d) post-tamping application of fouled ballast of 30% of FIp.

Fig. 16. Deviator stress versus axial strain of fresh and fouled ballasts with different values of FIp.

the sides of the sleeper. The results also revealed that fouled ballasts showed strain localisation beneath the sleeper more than fresh ballast in the post-tamping application, which could be interpreted as fouled ballasts being affected more than fresh ballast by the tamping tools. The increase in the residual settlement rate of fouled ballasts with over 30% of FIp in the post-tamping application can be attributed to the damage caused by inserting tamping tools during maintenance.

4.3. Effects of ballast fouling on strength properties Fig. 16 shows the stress–strain relationships for fresh and fouled ballast with different FIp (see Eq. 1) given in Table 2. Because the failure stress was not clear, and was particularly unclear on the strain hardening specimens, the failure frictional angle, ϕf, was evaluated using the peak stress values, A single test is sufficient to determine ϕf due to non-cohesion of

Fig. 17. Variation of failure friction angle with dry density.

particles in both fresh and fouled ballast. As shown in Fig. 16, fresh ballast showed the highest stress–strain values while fouling material showed the smallest stress–strain values. The results also revealed that specimens with 30% or more of FIp showed strain softening behaviour, whereas the other specimens (i.e., with 15% or less of FIp) showed strain hardening behaviour. That is, the results on the stress–strain relationships revealed that ballast loses its strain hardening behaviour when it is fouled by 30% or more of fouling materials. Fig. 17 shows the variation of failure friction angle against dry density for the specimens tested. As shown in Fig. 17, the largest dry density was observed on the specimens with 30–50% FIp. It also showed the fresh ballast specimen has the strongest strength properties (evaluated by ϕf here), and that these gradually decreased with increasing quantities of fouling material. Similar results on the strength properties of coal-fouled ballasts have been reported in Tennakoon et al. (2012). It should be noted that the magnitude of deviator stress increases with increasing confining pressure, sc (Anderson and Fair, 2008;

662

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663

Tennakoon et al., 2012). However, as reported by Anderson and Fair (2008) and Tennakoon et al. (2012), the stress–strain behaviour (e.g., strain hardening behaviour) remains more or less the same regardless of the value of sc. Therefore, although we conducted triaxial compression tests under a slightly higher sc, the results do not differ from the field conditions except in the case of a slight difference in the magnitude of deviator stress. As our main purpose was to compare the stress–strain behaviour of fresh and fouled ballast, the results, obtained using a slightly higher sc, simulated the field conditions appropriately. The triaxial compression test results suggested that the strength properties of fresh ballast deteriorate due to the intrusion of fouling material into fresh ballast. 5. Conclusions The effects of altered gradation curves by fouling material intrusion into fresh ballast and the tamping application on the settlement characteristics on ballasted railway tracks were investigated using a series of cyclic loading model tests conducted on a scaled-down ballasted railway track. The strength properties of fresh and fouled ballasts were also studied from laboratory triaxial compression test. The following conclusions were drawn from this research. The uniform nature of the gradation curve of fresh ballast is altered into a moderately uniform one by the intruded fouling material. The settlement characteristics of the ballasted railway tracks are altered subsequently. The relationship between the settlement curves of fouled ballast with the fouling index, FIp, and that of the void ratio is likely due to the filling of voids between larger-sized ballast by small-sized fouling material. The model test results indicate that tamping significantly altered the settlement characteristics of fouled ballast, and had a particularly clear effect on residual settlement. The extent of the decrease in the magnitude and the rate of the initial settlement in the post-maintenance application was larger than in the pre-maintenance application. In contrast, the increase in the residual settlement rate, which is responsible for long-term settlement, in the post-tamping application was larger than in the pre-tamping application when the fouled ballast was characterised by over 30% of FIp. The particle image velocimetry (PIV) results on the strain distributions verifies the damage caused by the tamping tools on fouled ballast where it is found that the local deformations are induced by the inserted tamping tools on the sides of the sleepers. The results suggest, for practical purposes, that tamping should be applied to fouled ballast before FIp exceeds 30%. The results from the triaxial compression test indicate that the strength properties also deteriorate due to the intruded fouling material. The triaxial compression test results also revealed that ballast loses its strain hardening behaviour when it is fouled by 30% or more of fouling material. Acknowledgements The Japanese Government is acknowledged for providing financial assistance through a Monbukagakusho scholarship for

the first author to study in his PhD at Yokohama National University (YNU), Japan. Ms. Kaori Ishii, a former graduate student of the YNU is also appreciated for her support during the laboratory experiments. The first author also acknowledges the President and Prof. Yoshiaki Kikuchi of Tokyo University of Science (TUS), Japan for the President Postdoctoral Fellowship offered to the first author where the manuscript was prepared. The authors are also thankful to the two anonymous referees whose high quality reviews and suggestions improved the quality of the paper to the current standard. Last but not least, the first author is also thankful to his wife Dr. Chamila Niroshinie of the TUS for her continuous assistance throughout. References Anderson, W.F., Fair, P., 2008. Behavior of railroad ballast under monotonic and cyclic loading. J. Geotech. Geoenviron. Eng. 134 (3), 316–327. Anderson, W.F., Fair, P., Key, A.J., McMichael, P., 2001. The deformation behavior of two layer railway ballast beds. In: Proceedings of the 15th International Conference on Soil Mechanics and Geotechnical Engineering, vol. 3, Istanbul, Turkey, pp. 2041–2044. Barksdale, R.D., 1972. Laboratory evaluation of rutting in base course materials. In: Proceedings of the 3rd International Conference on the Structural Design of Asphalt Pavements, London, pp. 161–174. Brown, S.F., Hyde, A.F.L., 1975. Significance of cyclic confining stress in repeated-load triaxial testing of granular material. Transp. Res. Rec. 537, 49–58. Brown, S.F., Kwan, J., Thom, N.H., 2007. Identifying the key parameters that influence geogrid reinforcement of railway ballast. Geotext. Geomembranes 25 (6), 326–335. Cambio, D., Ge, L., 2007. Effects of parallel gradation on strength properties of ballast materials. Adv. Meas. Model. Soil. Behav. 173, 1–7. Chrismer, G.P., Selig, E.T., 1993. Computer model for ballast maintenance planning. In: Proceedings of the 5th International Heavy Haul Railway Conference, Beijing, China, pp. 223–227. Feldman, F., Nissen, D., 2002. Alternative testing method for the measurement of ballast fouling: percentage void contamination. In: Proceedings of the Cost Efficient Railways through Engineering (CORE 2002), Wollongong, Australia, pp. 101–111. Hayano, K., Ishii, K., Muramoto, K., 2013. Effects of ballast thickness and tietamper repair on settlement characteristics of railway ballasted tracks. In: Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris, pp. 1275–1278. Holtz, W.G., Gibbs, H.J., 1956. Triaxial shear tests on pervious gravelly soils. J. Soil. Mech. Found. Div. 82 (1), 1–22. Hossain, Z., Indraratna, B., Darve, F., Thakur, P.K., 2007. DEM analysis of angular ballast breakage under cyclic loading. Geomech. Geoengin. 2 (3), 175–181. Huang, H., Tutumluer, E., 2011. Discrete element modeling for fouled railroad ballast. Constr. Build. Mater. 25 (8), 3306–3312. Indraratna, B., Ionescu, D., Christie, H.D., 1998. Shear behaviour of railway ballast based on large-scale triaxial tests. J. Geotech. Geoenviron. Eng. 124 (5), 439–449. Indraratna, B., Khabbaz, H., Lackenby, J., Salim, W., 2002a. Engineering Behaviour of Railway Ballast – A Critical Review. University of Wollongong, Wollongong, Australia. Indraratna, B., Khabbaz, H., Salim, W., Christie, D., 2006. Geotechnical properties of ballast and the role of geosynthetics in rail track stabilization. Ground Improv. 10 (3), 91–101. Indraratna, B., Lackenby, J., Christie, D., 2005. Effect of confining pressure on the degradation of ballast under cyclic loading. Geotechnique 55 (4), 325–328. Indraratna, B., Ngo, N.T., Rujikiatkamjorn, C., 2011a. Behavior of geogridreinforced ballast under various levels of fouling. Geotext. Geomembranes 29 (3), 313–322.

J.J. Kumara, K. Hayano / Soils and Foundations 56 (2016) 652–663 Indraratna, B., Nimbalkar, S., 2011. Implications of ballast breakage on ballasted railway track based on numerical modeling. In: Proceedings of the 13th International Conference of the International Association for Computer Methods and Advances in Geomechanics, Sydney, Australia, pp. 1085–1092. Indraratna, B., Nimbalkar, S., Coop, M., Sloan, S., 2014. A constitutive model for coal-fouled ballast capturing the effects of particle degradation. Comput. Geotech. 61 (9), 96–107. Indraratna, B., Nimbalkar, S., Rujikiatkamjorn, C., 2011b. Stabilisation of ballast and subgrade with geosynthetic grids and drains for rail infrastructure. In: Proceedings of the International Conference on Advances in Geotechnical Engineering, Perth, Australia, pp. 99–112. Indraratna, B., Nimbalkar, S., Tennakoon, N., 2010. The behaviour of ballasted track foundations: track drainage and geosynthetic reinforcement. In: Proceedings of the GeoFlorida 2010 Conference on Advances in Analysis, Modeling, & Design, Florida, USA, pp. 2378–2387. Indraratna, B., Salim, W., 2005. Mechanics of Ballasted Rail Tracks – A Geotechnical Perspective. Taylor and Francis, London. Indraratna, B., Salim, W., Christie, D., 2002b, Improvement of recycled ballast using geosynthetics, In: Proceedings of the 7th International Conference on Geosynthetics, Nice, France, pp. 1177–1182. Indraratna, B., Salim, W., Ionescu, D., Christie, D., 2001. Stress-strain and degradation behaviour of railway ballast under static and dynamic loading. In: Proceedings of the 15th International Conference on Soil Mechanics and Geotechnical Engineering, vol. 3, Istanbul, Turkey, pp. 2093–2096. Indraratna, B., Shahin, M.A., Salim, W., 2007. Stabilisation of granular media and formation soil using geosynthetics with special reference to railway engineering. Ground Improv. 11 (1), 27–43. Indraratna, B., Su, L., Rujikiatkamjorn, C., 2011c. A new parameter for classification and evaluation of railway ballast fouling. Can. Geotech. J. 48 (2), 322–326. Indraratna, B., Tennakoon, N., Nimbalkar, S., Rujikiatkamjorn, C., 2013. Behaviour of clay fouled ballast under drained triaxial testing. Geotechnique 63 (5), 410–419. Ionescu, D., 2004. Evaluation of the Engineering Behaviour of Railway Ballast. University of Wollongong, Wollongong, Australia. Ionescu, D., Indraratna, B., Christie, H.D., 1998. Behaviour of railway ballast under dynamic loads. In: Proceedings of the 13th Southeast Asian Geotechnical Conference, Taipei, Taiwan, pp. 69–74. Janardhanam, R., Desai, C.S., 1983. Three-dimensional testing and modeling of ballast. J. Geotech. Eng. 109 (6), 783–796. JGS 0527, 1998. Method for Triaxial Compression Test on Unsaturated Soils. Japanese Geotechnical Society, Tokyo, Japan. JIS A 1224, 2009. Test Method for Minimum and Maximum Densities of Sands. Japanese Industrial Standards, Tokyo, Japan. Kumara, G.H.A.J.J., 2013. Development of Prediction Methods for Deformation Characteristics of Fouled Ballasts Based on Laboratory Experiments and Discrete Element Method. Yokohama National University, Yokohama, Japan. Kumara, G.H.A.J.J., Hayano, K., Ogiwara, K., Takeuchi, M., 2012. Fundamental study on the simple evaluation methods for particle size distribution and maximum/minimum void ratio of sand-gravel mixtures. In: Miura, S.,

663

Ishikawa, N., Yoshida, N., Hisari, Y., Abe, N. (Eds.), Advances in Transportation Geotechnics II. CRC Press, pp. 572–577. Lackenby, J., Indraratna, B., McDowell, G., Christie, D., 2007. Effect of confining pressure on ballast degradation and deformation under cyclic triaxial loading. Geotechnique 57 (6), 527–536. Lade, P.V., Liggio Jr., C.D., Yamamuro, J.A., 1998. Effects of non-plastic fines on minimum and maximum void ratios of sand. Geotech. Test. J. 21 (4), 336–347. Leps, T.M., 1970. Review of shearing strength of rockfill. J. Soil. Mech. Found. Div. 96 (4), 1159–1170. Marschi, N.D., Chan, C.K., Seed, H.B., 1972. Evaluation of properties of rockfill materials. J. Soil. Mech. Found. Div. 98 (1), 95–114. Pen, L.M.L., Powrie, W., Zervos, A., Ahmed, S., Aingaran, S., 2013. Dependence of shape on particle size for a crushed rock railway ballast. Granul. Matter 15 (6), 849–961. Raymond, G.P., Bathurst, R.J., 1994. Repeated-load response of aggregates in relation to track quality index. Can. Geotech. J. 31 (4), 547–554. Raymond, G.P., Gaskin, P.N., Svec, O., 1978. Selection and performance of railroad ballast. In: Kerr, A.D. (Ed.), Railroad Track Mechanics and Technology. Pergamon Press, Oxford, pp. 369–385. Sekine, E., Ishikawa, T., Kouno, A., 2005. Effect of ballast thickness on cyclic plastic deformation of ballasted track. Railw. Tech. Res. Inst. (RTRI) 19 (2), 17–22 in Japanese. Selig, E.T., 1985. Ballast for heavy duty track. In: Track Technology, Proceedings of a Conference organised by the Institute of Engineers (ICE), Nottingham, 245–252. Selig, E.T., Waters, J.M., 1994. Track Geotechnology and Substructure Management. Thomas Telford, London. Sevi, A., Ge, L., 2012. Cyclic behaviors of railroad ballast with the parallel gradation scaling framework. J. Mater. Civil. Eng. 24 (7), 797–804. Shenton, M.J., 1975. Deformation of railway ballast under repeated loading conditions. In: Kerr, A.D. (Eds.), Railroad Track Mechanics and Technology, pp. 387–404. Shenton, M.J., 1978. Deformation of railway ballast under repeated loading conditions. In: Kerr, A.D. (Ed.), Railroad Track Mechanics and Technology. Pergamon Press, Oxford, pp. 369–385. Stewart, H.E., 1986. Permanent strain from cyclic variable-amplitude loadings. J. Geotech. Eng. 112 (6), 646–660. Tennakoon, N., Indraratna, B., Nimbalkar, S., Rujikiatkamjorn, C., 2012. Deformation and degradation of clay fouled ballast subjected to monotonic loading. In: Indraratna, B., Rujikiatkamjorn, C., Vinod, J. (Eds.), Ground Improvement and Ground Control: Transport Infrastructure Development and Natural Hazards Mitigation. Research Publishing, Singapore, pp. 1521–1528. Thakur, P.K., Vinod, J.S., Indraratna, B., 2013. Effect of confining pressure and frequency on the deformation of ballast. Geotechnique 63 (9), 786–790. Vallerga, B.A., Seed, H.B., Monismith, C.L., Cooper, R.S., 1957. Effect of shape, size and surface roughness of aggregate particles on the strength of granular materials. ASTM Spec. Tech. Publ. 212, 63–74. Wright, S.E., 1983. Damage caused to ballast by mechanical maintenance techniques. Br. Rail Res. Tech. Memo. 15.