Characterization of the Behavior of Granular Road Material Containing Glass Cullet

Transcription

1 Characterization of the Behavior of Granular Road Material Containing Glass Cullet S. Senadheera & P. Nash Department of Civil Engineering, Texas Tech University, Lubbock, Texas, USA A. Rana Frank X. Spencer Associates, El Paso, Texas, USA ABSTRACT: Granular base layers in highway pavements use large quantities of aggregate materials and are therefore good applications for possible use of waste materials such as glass. The engineering suitability of glass cullet as a granular embankment material is well documented. This paper presents results from resilient modulus tests on a blend of caliche, which is a conventional granular material used in pavement subbase layers, and glass cullet. Resilient modulus of granular material depends on factors such as aggregate mineralogy, particle characteristics, density, moisture content and gradation. However, when a material blend is used, the material response appears to also depend on relative strengths and compatibility of constituent materials in the blend. The general acceptance has been that by blending glass cullet with conventional materials, its engineering properties, particularly the strength, decreases. However, results from resilient modulus tests indicate that for relatively weaker granular base materials such as caliche, the introduction of glass cullet increases the strength of material blend. However, the strength gain appears to be accompanied by a likelihood of the material to fail by dilation at higher stress levels. As long as the caliche-glass cullet blend is not subjected to excessive loading, the presence of cullet in the blend appears to strengthen the resilient properties of the granular material. KEY WORDS: Glass cullet, roads, granular material, resilient modulus 1 INTRODUCTION Waste glass constitutes a significant portion of the solid waste that go to landfills every year. Over the years, a market for waste glass has been developed in glass container manufacture, and highway reflective markers and signs. Both these industries require sorting of glass by glass type and color, which is time consuming and often expensive. Therefore, a market for unsorted waste glass is needed to further reduce the burden on landfills. One such application is using glass cullet as a partial replacement for conventional pavement base and subbase materials. Glass cullet is produced by crushing waste glass and removing the debris such as paper and bottle caps to levels acceptable for a highway construction material. Highway applications such as pavement bases use large quantities of granular materials, and therefore would be an ideal market for glass cullet. Based on its engineering characteristics, glass cullet appears to be a good material to partially replace conventional granular materials.

2 The primary objective of this research was to determine the resilient (elastic) properties of different glass cullet-caliche blends. Caliche is weathered limestone that is often used as a pavement subbase material in many parts of Texas. Static load testing of granular materials does not simulate the repetitive vehicular loading that occurs in pavements. The engineering property that models the behavior of material under such repetitive loading is the resilient modulus. It is determined from laboratory testing of cylindrical granular material specimens and is calculated as the ratio of repetitive deviatoric stress and resilient (recoverable) strain. For this research, the AASHTO Test Method T was used with modifications (AASHTO 1994). Specimens prepared from four different caliche-glass cullet blends were tested for resilient modulus. In addition, particle size analysis and compaction tests were conducted on the material blends prior to resilient modulus testing. This paper presents the findings from resilient modulus tests and discusses the influence of glass cullet on the resilient behavior of caliche. The research into characterization of pavement granular materials started many decades ago. The first constitutive model for granular pavement materials was presented in 1962 when Biarez proposed his stress-dependent constitutive model. Since then, many researchers have investigated this topic. Hicks and Monismith (1971) out that stress level is the most significant factor that influences the resilient properties of granular pavement materials. They also found that factors such as material type, density, gradation, particle roughness, particle angularity, fines content and degree of saturation influence resilient response. Bulge (dilatancy) failure is typical in cohesionless granular aggregates when subjected to high stress levels. This type of failure was first observed in dense sand where the specimen volume increased during bulging failure. The term dilatancy was originally introduced by Reynolds (Reynolds 1885). He found that during plastic deformation under load, dense granular media such as sands or densely packed glass balls increase their volume and dilate. This is referred to as positive dilatancy. In contrast, loose granular media decrease in volume showing negative dilatancy (Sobolevsky 1995). Properties of a granular material that govern dilatancy may be classified into two groups; intrinsic factors such as surface texture, shape, size and elastic properties of grains and extrinsic factors such as grading and porosity or packing of aggregations (George and Shah 1974). George and Shah tentatively concluded that dilatancy increases with decreasing particle surface texture. The standardized testing protocols to determine resilient modulus of granular materials have evolved over the years. These protocols have been based on pre-conditioning the specimen to a point where it no longer shows significant permanent deformation during the load cycles when resilient properties are measured. Houston et al. (1993) indicated that one of the objectives of preconditioning is to induce any plastic strains that are prone to occur, so that mostly elastic strains remain during resilient modulus testing. During the sample conditioning stage, it is hoped that any voids in the specimens are supposedly removed and a good contact between the specimen and load platens is achieved (Nazarian and Feliberti 1993). However, the significance and validity of the preconditioning stage continue to be the subject of much debate. The resilient response of granular material to repeated loading is expressed using the parameter resilient modulus which is determined by repeated load triaxial testing. Using results from laboratory repeated load tests, numerous constitutive models have been proposed to predict resilient modulus. The most commonly used examples of such models for granular pavement base and subbase materials are the bulk stress model (Eq. 1) and a model proposed by Uzan that includes the bulk stress and deviatoric stress as parameters (Eq. 2). For the remainder of this paper, this model will be referred to as the three-parameter model. M = θ Eq. 1 R K 1 K 2

3 In Eq. 1, θ is the bulk stress (first stress invariant of the normal stress tensor) and K 1, K 2 are regression constants determined by statistical analysis of laboratory data. The bulk stress is equal to the sum of three principal stresses. Previous research has shown that higher quality granular materials exhibit larger K 1 values and smaller K 2 values (Rada and Witczak 1981). The major limitation of the bulk stress model is that it does not account for shear stresses and shear strains during loading (Uzan 1981). Also, the bulk stress model does not appropriately handle volumetric strain or dilative behavior of granular soil under triaxial loading conditions (Brown and Pappin 1985). However, bulk stress model is still widely used and the AASHTO T test procedure recommends the use of bulk stress model for all unbound granular base and subbase materials. M R = k θ 2σ 3 Eq. 2 1 k k d For the three-parameter model shown in Eq. 2, θ is the bulk stress and σ d is the deviatoric stress. The three parameters k 1, k 2 and k 3 are regression constants. Previous research has shown that three-parameter constitutive model predicts laboratory measured resilient moduli more effectively (Uzan 1981). Clean Washington Center (CWS), a division of the Washington State Department of Trade and Economic Development, sponsored a study that conducted extensive laboratory tests, including resilient modulus tests, on glass cullet and cullet-conventional material blends (CWS 1993). Two glass cullet gradations, with ¼ inch and ¾ inch maximum size, were used in this study. Resilient modulus values for conventional base material (crushed rock) and crushed rock-glass cullet blend were determined during the pre-conditioning stage of the test with a 4 psi confining stress and a 8 psi deviator stress. The pre-conditioning data showed that addition of cullet reduced the resilient modulus of the blend. The conventional base material (crushed rock) showed a resilient modulus of 38.3 ksi. When 15% glass cullet (¼ inch size) was used in the blend, the resilient modulus reduced by approximately 12%. For 50% glass cullet, the reductions were between 15 and 23 percent. Within the 50% cullet blend, the ¾ inch cullet size showed a higher resilient modulus compared to the ¼ inch size. This report also investigated whether the presence of glass cullet causes breakdown of crushed rock particles during sample preparation and/or the repeated loading phase. However, no appreciable change in the overall gradation was observed. The CWS study also showed that, at the same axial strain, specimens with ¼ inch cullet showed higher volumetric strains (higher dilatancy) than specimens with ¾ inch cullet. This behavior is in agreement with previous studies that showed a higher tendency to dilate for dense specimens with a higher degree of packing (Shockley and Ahlvin 1960). 2 TEST PROGRAM 2.1 Materials The two materials used in the test program were caliche and glass cullet. Caliche, which is weathered limestone, is used as a subbase material in many parts of Texas. Caliche was obtained from the Butler Pit in Lubbock, Texas. Glass cullet used in this study was obtained from the City of Abilene Recycling Center. It was produced by crushing municipal waste glass to a maximum size retaining on #4 sieve. Once crushed, the glass was cleaned of paper labels and cork using a vacuum system and metal caps were removed using magnets.

4 2.2 Preliminary Laboratory Tests The particle size analysis and compaction tests were conducted for the caliche-cullet blends before the resilient modulus tests were conducted. Four materials were tested; the control specimen prepared with 100% caliche, and three other material blends were prepared with 20, 30 and 50 percent cullet. Sieve Analysis was performed on both caliche and glass cullet separately in order to separate materials to different sizes. Then, the two materials were blended to achieve the Texas Department of Transportation (TxDOT) gradation for Grade I flexible base materials (TxDOT 1993). The median point was selected within each gradation band. The TxDOT gradation limit requirements and the gradation of material blend are shown in Table 1 Table 1: Gradation of caliche-glass cullet blends Sieve Size Gradation of Tested Material (Cumulative % Retained) TxDOT Specification for Grade 1 Flexible Base Material (13) 1-3/ / / No No The modified Proctor compaction test was conducted on the four material blends to obtain moisture-density relationship for each blend. The test was conducted according to AASHTO Specification T as recommended in AASHTO T procedure (AASHTO 1995). Table 2 shows the moisture-density relationship test results for each blend. The optimum moisture content decreased with increasing glass cullet content. This is expected because glass cullet, being an impervious material, reduces the water demand of the blend. The dry density of 100% glass cullet sample did not show any sensitivity to the moisture content. Table 2: Moisture-density relationship results Caliche-Glass Blend Optimum Moisture Content (%) Max. Dry Density (lb/ft3) 100: : : : :100 Insensitive to moisture Resilient Modulus Test The resilient modulus tests were conducted in accordance with AASHTO T with some modifications. This test method recommends vibratory compaction of the specimen using a vibratory head, but no exact specifications are provided for the number of soil layers per specimen, the amount of vibration energy and the duration of vibration. In this research, a vibratory table was used to support the compaction mold and the specimen, and no vibratory head was used. Instead, a static head was used to help compact the specimen from the top. The vibrating table was supported on four legs and the table was supported on springs attached to the top of each leg. A pneumatic vibrator attached to the table. The compaction method used in this research was similar to that developed by Zaman et al. (1994). The specimens were six inches in diameter and 12 inches high and were prepared in a

5 split mold mounted on a base plate which was attached to the vibrating table. The specimen was prepared in 12 layers with compacted height of each layer being one inch. Each layer was vibrated for 60 seconds, followed by dropping a 5.5lb ram over a 12 inch distance on top of each compacted layer. The number of drops varied between 10 and 15 to ensure uniform compaction over the height of specimen. More drops were used in the top layers. Optimum moisture content determined from the moisture-density relationship test was used in specimen preparation. The compacted specimen was transported while in the split mold and transferred to the triaxial cell prior to testing. The AASHTO T procedure recommends a haversine loading pattern with a loading duration of 0.1 seconds and an unloading duration of 0.9 seconds. The loading cycle used in this research had a step-loading and unloading pattern with a 0.5-second loading duration and a 0.5 second unloading duration. The shape of the loading pulse and the loading-unloading durations were carefully chosen to match the actual loading pulse to the desired loading pulse and to ensure appropriate recovery of the specimen during the unloading phase. The load pulse shape and the duration of loading and unloading were identical for all specimens. The resilient modulus test loading sequence used in this research was identical to the procedure outlined in AASHTO T ANALYSIS OF TEST RESULTS The results were analyzed for both pre-conditioning and post-conditioning. As per the AASHTO test procedure, specimens were subjected to 1000 pre-conditioning cycles at a confining pressure of 15 psi and a deviatoric stress of 15 psi. Resilient modulus values were calculated at the end of each 100 pre-conditioning cycles and plotted in Figure 1. The data show that blends with higher glass contents consistently gave higher resilient modulus values during the pre-conditioning stage. This is contrary to the observations made by Shin and Sonntag (1994). However, Shin and Sonntag conducted their pre-conditioning at lower stress levels, i.e. confining stress and deviatoric stresses of 4 psi and 8 psi respectively. The post-conditioning resilient modulus data was first fitted to the bulk stress model (Eq. 1) and is shown in Figure 2. The bulk stress model appears to fit the data reasonably well and it does show the stress dependency of resilient modulus in the material tested. Another interesting observation from Figure 2 is that resilient modulus of the caliche-glass cullet blends increase with increasing glass cullet content. This is also contrary to the observations made by Shin and Sonntag (1994). Table 3 shows K 1 and K 2 values calculated for the two-parameter bulk stress model. The trend in this data is consistent with results of resilient modulus tests published by Rada and Witczak who noted that stronger granular material showed higher K 1 values and lower K 2 values (Rada and Witczak 1981). It should be noted that the 50:50 caliche-glass cullet blend did not produce any K 1 & K 2 values because it failed by bulging during the pre-conditioning stage. Table 3 also provides k 1, k 2 and k 3 values from the three-parameter model in Eq. 2. Theses values were used to predict resilient moduli which were compared with measured values for 100% caliche in Figure 3. It shows that predicted moduli at lower confining stresses (3, 5, 10 and 15 psi) agreed reasonably well with measured values. However, predicted moduli deviated from measured values at higher confining stress levels (20 psi). The coefficient of determination (R 2 ) values show a better statistical fit with the threeparameter model compared to the two-parameter bulk stress model (Table 3). In addition, the three-parameter model also better predicts the trends in laboratory data better. Axial strains, both resilient and permanent, were calculated for each specimen tested. During the pre-conditioning stage, cumulative axial permanent strain increased with increasing glass cullet content. The specimens containing glass cullet showed six to seven

6 times more permanent axial strains during the pre-conditioning stage. This is also evident from Figure 4 where the accumulation of axial strain at the end of 1000 pre-conditioning cycles and the 15 loading sequences is shown. This is a very significant difference which may be attributed to degradation of caliche particles due to sharp edges of the glass cullet particles. This observation is contrary to that observed by Shin and Sonntag (1994) who monitored the difference in gradation of crushed stone-cullet blends and observed no significant differences before and after testing Resilient Modulus (psi) AASHTO T Conditioning Cycles 0% Cullet 20% Cullet 30% Cullet Figure 1: Resilient modulus during pre-conditioning stage y = 3489x R 2 = Resilient Modulus (psi) y = x R 2 = y = x R 2 = Bulk Stress (psi) '30% Cullet' '20% Cullet' '0% Cullet' Power ('20% Cullet') Power ('30% Cullet') Power ('0% Cullet') Figure 2: Laboratory measured resilient moduli and best-fit curves using bulk stress model. This stark difference in performance between this research and research conducted by Shin and Sonntag may be due to the quality of crushed stone used. The crushed stone (caliche) used in this research was of significantly lower quality than the crushed stone typically used in Texas pavement base courses. For a bulk stress of 25 psi, the resilient modulus values for

7 caliche used in this research and crushed rock used by Shin and Sonntag were calculated to be 15 ksi and 40 ksi respectively. This indicates that caliche used in this research was much less stronger compared to the crushed rockused by Shin and Sonntag (1994). Table 3: Constitutive model parameters for tested material blends Caliche- Bulk Stress Model Uzan Model Glass Blend Model Parameters Model Parameters R 2 K 1 K 2 R 2 k 1 k 2 K 3 100: : : : Resilient Modulus (psi) Deviatoric Stress (psi) 3 psi (Pred) 5 psi (Pred) 10 psi (Pred) 15 psi (Pred) 20 psi (Pred) 3 psi (Meas) 5 psi (Meas) 10 psi (Meas) 15 psi (Meas) 20 psi (Meas) Figure 3: Measured vs. predicted resilient moduli for the 100% caliche specimen using the three- parameter model Figure 4 also shows the accumulation of permanent strain during the 15 testing sequences used to determine the resilient modulus. It does not conclusively show any significant difference in accumulated permanent strain during the 15 loading sequences between specimens with different glass cullet contents. However, it shows that even after preconditioning, significant accumulation of permanent strain does occur. These results also show the effect of stress history on specimen response. It is evident that permanent deformation jumps whenever the specimen is subjected to a stress level (deviatoric stress) that is higher than it was subjected to in the past, including the pre-conditioning stage. Another interesting observation that was made from the post-conditioning testing is the warning of imminent dilatant failure of the specimen with 30% glass cullet. Even though the specimen did not fail at the end of loading sequence 15, it did show an unusual jump in the axial strain by reversing its trend of being the strongest specimen until that point (Figure 5). It would be reasonable to postulate that had the stress sequences continued to higher stress levels, even the specimen with 20% cullet and possibly the specimen with 0% cullet would have shown similar bulging tendencies. As discussed previously, the resilient modulus

8 increased with increasing cullet content within the stress levels in the testing protocol. However, the bulging failure of the specimen with 50% cullet during pre-conditioning stage may suggest that with increasing cullet content, the stress level at which dilatant (bulging) failure occurs would decrease Cumulative Permanent Axial Strain AASHTO T Sequence No. 0% Cullet 20% Cullet 30% Cullet Figure 4: Accumulation of permanent axial strain during pre- and post-conditioning testing Based on the constituent materials and their combinations tested in this research, the material blends with 0% and 20% cullet content would be able to withstand a stress state of a 20 psi confining stress and a 40 psi deviatoric stress. This translates to a principal stress ratio of 3.0. The material blend with 30% cullet showed signs of imminent dilatant failure at this stress state (Figure 5), and therefore, it would be able to withstand a stress state represented by the las loading sequence. As it was mentioned previously, the material with 50% cullet failed during pre-conditioning stage. Although the moisture-density test results showed increasing maximum dry density with decreasing glass cullet content, the resilient modulus showed a reverse (increasing) trend with increasing glass cullet content. The reason for this reversal may be attributed to differences in compaction for the two tests indicated above. It is possible that vibratory compaction combined with impact compaction, which was used for resilient modulus specimen preparation, significantly densified the specimen. Hicks and Monismith (1971) indicated that generally, with increasing density of granular materials, K 1 value increases and K 2 value remains relatively constant or decreases. If this theory is applied to data in Table 3, it shows that the density of specimens used for resilient modulus increased with increased glass cullet content. This may have attributed to increased resilient moduli with increasing cullet content. A volume increase (positive dilatancy) is generally associated with the breakdown of interlocking, both microscopic (surface friction) and macroscopic (mechanical locking). For the 50:50 caliche-cullet blend, it is possible that increased amount of sharp glass cullet edges had a more abrading effect on the caliche than other samples with lower amounts of glass cullet. Therefore, the sample possibly underwent a higher degree of densification during specimen preparation. In addition, replacement of caliche by larger amounts of glass cullet increased the proportion of particles with smoother surface texture. Abrasion of particles and

9 the resulting densification of material may have eased the sliding friction resistance that existed between glass and caliche particles. This is supported by the findings of George and Shah (1974). Therefore, during the conditioning stage, this highly densified caliche glass cullet blend expanded under stress and resulted in bulge failure (positive dilatancy). Axial Strain in Last Cycle Dilatant Behavior of 30% Cullet Blend AASHTO T Sequence No. 0% Cullet 20% Cullet 30% Cullet Figure 5: Total axial strain in last cycle in loading sequence #15. 4 CONCLUSION It is important to develop applications for productive re-use of waste materials. This will reduce the burden on landfills and help preserve scarce natural resources. In order to do this, careful investigation of the engineering characteristics of materials such as glass cullet is essential. In this research, the resilient characteristics of caliche-glass cullet blends were investigated. It was found that up to a threshold level of 30% glass cullet, the resilient modulus of the caliche-glass cullet blend increased with increasing cullet content. It was also found that significant differences exist between compaction methods adopted for the moisture density relationship test and the resilient modulus test. After specimen preparation, the specimens were subjected to further densification during both pre-conditioning and post-conditioning testing. The degree of densification increased with increasing glass cullet content, and therefore, higher densification contributed to higher stiffness values. Following densification, and particularly at higher stress levels, the caliche-glass cullet blends display dilatant tendencies, thereby jeopardizing the stability of the material. Specifically, the material blend with 30% glass cullet showed signs of dilatant failure at stress sequence 15 with 20 psi confining stress and 40 psi deviatoric stress. The material blends with 0% and 20% glass cullet contents did not show any signs of dilatant behavior at this stress level. The material blend with 50% glass cullet possibly underwent the most densification of the four blends during sample preparation, and resulted in dilatant failure during pre-conditioning phase. The presence of higher amount of smooth glass surfaces may also have contributed to dilation. It was also found from this research that glass cullet has an abrading effect on caliche material. This finding contradicted with the observations by Shin and Sonntag (1994), but the

10 differences may be attributed to the difference in quality of the two conventional materials tested in the two research efforts. It is generally perceived that use of waste materials in combination with conventional materials will weaken the material blend. However, this may not hold true for every scenario. For example, as it was shown in this research, relatively weaker conventional materials may be strengthened by the addition of glass cullet. Therefore, when a recycled material is considered as a partial replacement for a conventional material, characteristics of each constituent material along with that of the blend must be thoroughly investigated. REFERENCES American Association of Highway and Transportation Officials (AASHTO), Standard Specifications for Transportation Materials and Methods of Sampling and Testing, 17 th Edition, Part II: Tests, Washington, D. C., Brown, S.F., and Pappin, J. W., Analysis of Pavements with Granular Bases. In Transportation Research Record 1022, TRB, Washington, D.C. 1985, pp Clean Washington Center (CWS), Glass Feedstock Evaluation Project: Engineering Suitability Evaluation. Report No. 5, Dames and Moore, Inc., Seattle, Washington, George, K.P. and N.S.Shah, N. S., Dilatancy of Granular Media in Triaxial Shear. In Transportation Research Record 487, TRB, Washington, D.C. 1974, pp Hicks, R.G., and Monismith, C. L., Factors Influencing the Resilient Response of Granular Materials. Highway Soils Engineering, Highway Research Record 345, Washington D.C., 1971, pp Houston, W.N., Houston, S. L. and Anderson, T. W., Stress State Considerations for Resilient Modulus Testing of Pavement Subgrade. In Transportation Research Record 1406, Washington, D.C. 1993, pp Nazarian,S., and Feliberti, M., Methodology for Resilient Modulus Testing of Cohesionless Subgrades. In Transportation Research Record 1406, Washington, D.C. 1993, pp Rada, G., and Witczak, M. W., Comprehensive Evaluation of Laboratory Resilient Moduli Results for Granular Material. In Transportation Research Record 810, Washington, D.C. 1981, pp Reynolds, O., On the Dilatancy of Media Composed of Rigid Particles in Contact with Experimental Illustrations. In Philosophical Magazine and Journal of Science, 5 th Series, Vol. 20, No. 127, 1885, pp Shin, C.J. and Sonntag, V., Using Recovered Glass as Construction Aggregate Feedstock. In Transportation Research Record 1437, TRB, Washington, D.C. 1994, pp Shockley, W.G. and R.G.Ahlvin, R. G., Nonuniform Conditions in Triaxial Test Specimens. Research Conference on Shear Strength of Cohesive Soils, ASCE (1960), pp Sobolevsky, D., Strength of Dilating Soil and Load-Holding Capacity of Deep Foundations. A.A.Balkema, Rotterdam, Netherlands, Texas Department of Transportation (TxDOT), Standard Specifications for Construction of Highways, Streets and Bridges. Austin, Texas, Uzan, J., Characterization of Granular Material. In Transportation Research Record 810, TRB, Washington, D.C. 1981, pp Zaman, M., Chen, D. and Laguros, J., Resilient Moduli of Granular Materials. Journal of Transportation Engineering, Vol. 120, No. 6, November/December, 1994, pp