An increase in demand for aggregates and environmental and economic concerns over the disposal of waste has led to an interest in the use of recycled materials for construction purposes. Many recycled materials have the potential for use in construction but little research has been carried out into their performance. This paper identifies some of the properties of waste glass which suggest possibilities for its use in geotechnical situations. Three samples of crushed waste glass are tested for permeability, shear strength, crushing resistance and small-strain stiffness and the results show that the glass samples have the potential to perform as well as gravel.
In recent years there has been an increasing worldwide demand for the construction of buildings, roads and airfields which has led to a local depletion of aggregates. In some urban areas, the enormous quantities of aggregate that have already been used means that local materials are no longer available and the deficit has to be made up by importing materials from other locations.
Most countries have areas of land covered by spoil heaps which are unsightly and prevent large areas of land being used for anything else. Moreover in the UK, the introduction of the landfill tax makes tipping uneconomic in many cases. If the large amount of waste materials generated were used instead of natural materials in the construction industry there would be three benefits:
conserving natural resources, disposing of waste materials (which are often unsightly) and freeing up valuable land for other uses.
The generation of waste glass constituted 6.7Mt in 1960 and continued to grow over the following two decades. In the 1980s, glass containers were widely replaced by plastics and the fraction of waste glass dropped from 15Mt in 1980 to 12.5Mt in 1988 where it has approximately stabilised (Ahmed & Lovell, 1992). In Britain in 1993, over 500,000t of bottled glass was recycled through bottle bank schemes, rising to over 800,000t by 1995 (British Glass, 1994). Increasing public awareness of recycling schemes, coupled with new landfill taxes make the search for alternative uses for waste glass desirable.
Ideally, most waste glass would be recycled into new glass.However, since only colour sorted and contamination-free waste glass is feasible for reuse in the glass industry, there is a surplus of waste glass which cannot be reused by glass manufacturers. The increased amount of green and mixed bottles being collected, which are surplus to current requirements, will also need alternative uses.
Therefore significant quantities of glass may be available for other applications.
Growing stockpiles of glass in the US in the 1970s led to work being conducted on possible pavement uses. This work found that glass could be used within hot rolled asphalt, producing what has become known as Glassphalt. The American problem with glass disposal is at its most extreme in New York, where 27,000t of mixed colour waste glass was collected in 1993, and was expected to rise to 110,000t by 1997 (Concrete International, 1995). New York has been recycling waste glass into Glassphalt for some time, however the city now produces more debris than can be used in this application and as a result, alternative uses have been investigated.
In 1995, the New York State Energy Research & Development Authority awarded a two year research contract to Columbia University to investigate the feasibility of replacing aggregates in concrete with waste glass. This aimed to establish that the use of waste glass in concrete could save the city tipping fees of $65/t at that time and would help relieve its shortage of landfill sites. In addition it was found that the possible savings to concrete block manufacturers could be significant.
There are two major sources of glass supply: l Waste from glass factories l Glass waste in household refuse Conventional crushing equipment is used to produce glass aggregates which have been found to consist of a high percentage of flat or longish elements.There are no technical problems standing in the way of reusing glass in the construction industry.However, glass waste is widely scattered and while there may be many sources, the cost of transportation from source to site may prove prohibitive in some instances.
The glass used in this work was obtained from Mac-Glass Recycling in Dalkeith, Edinburgh.
Three samples were available, enabling the effect of particle size to be considered as part of the testing programme.The 'fine'and 'coarse'crushed materials were produced by crushing the glass and then refining the finished product by grinding the sharp edges, producing fairly smooth granular particles.The 'raw'crushed glass was obtained by crushing the raw material without any further refinement, producing very angular finished particles.
Measurement of some relevant geotechnical properties
Particle size distribution charts for the samples are given in Figure 1 and primary characteristics have been identified using standard laboratory techniques (Table 1).
Differences in the particle densities measured are probably caused by differences in the glass composition; the fine and coarse glass comprise clear and green glass whereas the raw glass is sourced from brown glass.
The geotechnical requirements of the three samples were defined as high shear strength and smallstrain stiffness, high permeability and high crushing resistance. The testing programme was designed to measure these properties of the glass and of a sample of gravel to give a basis for comparison. The gravel used was quartz dolerite with a nominal particle size of 6mm. A particle size distribution chart for the gravel is shown in Figure 1 and its basic properties are given in Ta b l e 2 .
The following standard test procedures were adopted:
l Constant head permeability testing l Aggregate crushing value (ACV) testing, obtained by comparing the coefficient of uniformity of the samples before and after crushing l Direct shear testing using a 300mm square large shear box and 100mm diameter triaxial cell From the results of these tests the geotechnical performance of the waste glass can be assessed.
Constant head permeability testing
A standard 75mm diameter permeameter was used to determine the permeability of the glass and gravel samples.The samples were placed in the apparatus in three layers and tamped to the required relative density of 70%. The high relative density was selected in an attempt to simulate field compaction conditions. The permeameter was connected to a constant head apparatus and the samples were flushed through with de-aired water to remove any air voids.
The resulting permeability for each material is given in Ta b l e 4 . The raw glass proved to be nearly as permeable as the 6mm gravel but, not surprisingly, the permeability decreased dramatically with particle size with the coarse glass only approximately one tenth as permeable as the gravel and the fine glass only about one two hundredth of the permeability of the gravel and much more like the permeability of a coarse sand in accordance with its grading.
Aggregate crushing value
The standard test on the crushing properties for bulk aggregates is the Aggregate Crushing Value (ACV) test as set out in BS812: Pt110 (1990). Although there is no specific relationship between crushing value and compressive strength, the crushing value is a useful guide when dealing with aggregates of unknown performance.
For this programme, some modifications to the standard procedure were introduced for convenience. First, instead of sieving out a 1kg sample between the specified ranges, this testing programme used a 1kg sample of each material's complete particle distribution in a similar manner to that proposed by Hardin (1985). In addition, it was decided to use the coefficient of uniformity before and after crushing as a comparison rather than sieving out one size. This allows the behaviour of all particle sizes to be examined and gives a fuller picture of each of the material's behaviour.
The particle size distribution charts for the materials both before and after crushing were plotted and values for coefficients of uniformity obtained. The ratio of the crushed to uncrushed coefficients was found and given as the modified aggregate crushing value for this programme.
Table 3 shows the change in grading of the samples due to the crushing process.
Results show that while the d60
particle size is relatively unaffected by the crushing process, the d10particle size is, on average, halved. This gives the coefficient of uniformity for the crushed material to be roughly twice that of the uncrushed and results in average ACVs of 2.
The fine glass proved to have the strongest resistance to crushing which can probably be attributed to the small initial particle size, the material having already been crushed down to its most stable particle size during manufacture.
It was noted that the change in particle size of all the glass samples was principally due to grinding between particles as they rearranged themselves within the sample rather than splitting of individual particles due to lines of weakness within them.
Therefore the main change in particle size distribution was simply due to the generation of dust from abrasion rather than cracking of individual elements. This resulted in a generally good performance from the glass and one which was comparable with gravels.
Direct shear testing
Direct shear testing was chosen as the most straightforward method to determine the shear strength of the glass samples as there are no problems of membrane puncture which may occur with triaxial testing. To cross check the results obtained, one envelope of drained and undrained triaxial tests were carried out on the samples of raw glass and the gravel.
The direct shear strength tests were performed on a 300mm square direct shear machine, or large shear box, in the manner described by Woodward & Blewett (1998). Dry samples were placed in the box and compacted in three layers to obtain a final relative density of 70% which was the largest value that could be achieved without crushing the glass and gravel particles. The normal stress was applied to the sample and the initial settlement noted.
The shear box provides a constant rate of displacement until failure and during this displacement both hor izontal force and ver t ical settlement are recorded. The normal stresses chosen to produce each test envelope were 100kPa, 200kPa and 300kPa and the rate of displacement was 0.5mm/min.
Figure 2 shows the build up of shear strength in both the glass and the gravel samples under a normal overburden stress of 200kPa. These results are presented as typical plots of all the tests carried out. The gradual increase in shear stress with displacement is a characteristic of all the samples tested, although the glass samples exhibit a less stiff build up to ultimate strength than the gravel. There is little difference between the behaviour of each of the glass samples and their ultimate shear strengths (taken at 30mm displacement) are less than 10kPa smaller than that of the gravel. Steady state conditions do not appear to have been reached in the tests due to the large displacement which would be required. The friction angles obtained from the test envelopes are given in Table 4 but it should be noted that these angles may actually be greater if steady state conditions had been reached.
Figure 3 shows the vertical - horizontal displacement relationships corresponding to the results in Figure 2 and are once again typical of those obtained throughout the testing programme. The vertical settlement of the glass and gravel samples are similarly contractive and are likely to be strongly linked to the crushing resistance of the materials which have been shown to be similar.
Many authors have examined the effect of aggregate crushing on laboratory test results (Hardin, 1985, Lade et al, 1996, Fumagalli, 1969). Particle breakage can occur even at low pressures, depending on the characteristics of the soil grains. It is generally recognised that the factors which affect the amount of particle crushing, both in the laboratory and in the field, are:
l Particle size distribution lParticleshape lTheeffectivestressconditions l The applied stress path l The void ratio of the deposit l Internal bedding or planes of weakness lParticlehardness l Time under loading l Presence of water (decreases particles apparent hardness) The potential for breakage increases with particle size as individual particles will contain more fractures or lines of weakness and there will be fewer contact points on the particle surface, each of which will carry an increased load. Watt (1991) carried out some approximate calculations to estimate the contact forces in various materials due to a normal stress of 98kPa.He found that in a medium sand the contact force would be 9.8N, in a gravel it would be 9.8kN and in a rockfill material it would be 9.8MN. This clearly demonstrates the relationship between particle size and breakage potential.
Hardin (1985) proposed that the lower limit on particle size with breakage potential is 0.074mm. Determination of the crushing strength or breakage potential of a deposit is important both in laboratory and field conditions. In triaxial compression and direct shear, crushing occurs at sliding contacts and significantly reduces the rate of dilation of the sample. Triaxial volumetric strain data and shear box vertical settlement data from these tests must therefore be treated with care when applying the results to field conditions. Fumagalli (1969) carried out triaxial testing on materials for rockfill dams and found that particle crushing caused reduction in friction angles of up to 10degrees.
Most of the samples in this study exhibit initial contraction with a small amount of dilation taking place at about 14mm displacement.The raw glass sample is purely contractive right up to the 30mm displacement limit. Since this material is very angular it is probable that breakage of the particles, rather than rolling, occurs leading to further settlement rather than dilation. The glass samples with smaller particles are in a more stable state which allows more rolling without breakage, thus the sample with the smallest particle sizes, the fine glass, exhibits the most dilation.
As the raw glass has been shown to have the best direct shear and permeability properties of the glass samples, a series of drained and undrained triaxial tests were carried out on this and the gravel samples. The triaxial testing system incorporates a hydraulic triaxial cell, manufactured by GDS, and of the type described by Bishop and Wesley (1975) which can accommodate 100mm diameter samples. In addition piezoelectric bender elements have been inserted in the cell in the manner described by Dyvik and Madshus (1985) to facilitate the measurement of shear wave velocity and hence the small-strain stiffness parameters, Gmax, of the materials tested.
The same sample preparation technique was used for all samples. The split form mould was assembled on the triaxial pedestal in the usual way. Two membranes were used to guard against puncture in all cases and the strain readings were subsequently corrected for this and membrane penetration (Nicholson et al, 1993). The membranes were filled with de-aired water. The material was saturated and placed in the mould in approximately 10 layers and tamped between each layer to achieve the target relative density of 70%.Once the top cap was in place, a negative pressure of 20kPa was applied to ensure that the sample structure held together on removal of the mould.The cell lid was placed and filled with water. The degree of saturation was measured and the samples flushed through with water until a B value of greater than 0.97 was achieved.
Pressures were applied to the sample by means of three GDS hydraulic digital pressure controllers, one for the cell pressure, one for the back pressure and one for the lower chamber which drove the piston and hence provided the axial force.Pore pressure was measured at the opposite end of the sample to the drain by means of a GDS digital pressure interface.
Only the raw glass and gravel were tested to verify the shear box results, as problems did occur with puncture of the membrane. Both drained and undrained shear tests were completed at effective confining pressures of 100kPa, 200kPa and 300kPa. Values of Gmax were calculated from bender element measurements at the start of each test using a 60Hz square wave input and measuring the time of flight to the maximum of the received signal.The authors acknowledge that there maybe some error associated with this measurement technique (Viggiani and Atkinson, 1995, Blewett et al, 1999) but for simplicity, and to give a comparison between the two materials, it was deemed acceptable in this instance.
The stress-strain behaviour in drained shear obtained from the raw glass and gravel test envelopes are shown in Figure 4.The build up of shear strength in both the gravel and glass samples is gradual over the strain range as seen previously in the shear box.The glass samples have a slightly smaller ultimate shearing resistance than the gravel, although only by approximately 20kPa. The friction angles obtained from these tests are shown in Table 4. The friction angles in triaxial shear are smaller than those obtained in the shear box and the results presented here therefore show that the peak friction angle of these granular materials is a function of the stress path and density.This type of behaviour was also shown by Lade and Duncan (1975) and Lade (1997) who developed the Lade and Duncan failure surface for granular soils.In this criteria the equivalent peak friction angle is greater than the Mohr-Coulomb for all stress paths other than triaxial. This study finds the differences between the two testing techniques to be 8degrees in the glass and 6degrees in the gravel.
Generally, the triaxial friction angles appear to be low suggesting that steady state conditions were not reached during the tests due to the large shear strains required.The stress-strain plots and the volumetric response of the samples indicate that this may be the case.
Figure 5 shows typical volumetric strain-axial strain relationships for the materials.Once again the predominantly contractive behaviour is apparent which can be attributed to the crushing behaviour of the samples. As expected more contraction occurs at higher confining pressures. In addition some of this contraction will be due to particle rearrangement.This is particularly likely in the case of the raw glass where the particles resemble flat 'plates' and could potentially cause a considerable reduction in sample size if they become arranged with their flat faces together rather than their edges during shear.Obviously, if the particles become arranged in this way, their shearing resistance is greatly reduced as their smooth surfaces allow them to slide more easily over each other. In the small scale of triaxial samples this effect could become significant, although on a larger scale, the greater number of particles should ensure adequate randomisation of their orientation preventing large losses in strength. It is therefore important to construct laboratory samples carefully and carry out many tests to ensure that the particle alignment is significantly randomised.
For this testing programme, the whole envelope of tests was carried out three times and the shear strengths obtained were similar enough to ensure that this particle arrangement effect was minimised.
To examine the crushing behaviour of the materials under triaxial conditions, a quantity of each of the samples was sieved before and after testing. Figures 6a and 6b show the particle size distributions resulting from drained shearing of the samples under a confining pressure of 200kPa.
Crushing occurs in both samples. The gravel exhibits crushing over the whole range of particle sizes, although the degree of crushing increases with decreasing particle size. The glass, however, shows no crushing of the larger particle sizes but a great deal of crushing within the smaller particles, probably as a result of the type of abrasion previously noted during the ACV test.
Figures 7 and 8 show the stress-strain relationships and the build up in pore pressures obtained from a series of undrained tests carried out on the samples. Once again, the build up in shear strength is continuous with axial strain and the raw glass has slightly less overall shearing resistance than the gravel.The undrained friction angles obtained are shown in Ta b l e 4 .The build up in excess pore pressure is also similar between the glass and the gravel. Under 100kPa and 200kPa confining pressures the glass samples exhibit a slightly greater initial build up in pore pressures and greater subsequent reduction in pore pressure due to dilation.
If it is assumed that one of the main factors effecting volumetric strain during drained shear, and excess pore pressures during undrained shear, is crushing then it is likely that the main crushing deformation of the glass takes place during the initial stages of shearing. However, once the particles have reached a more stable state, crushing reduces and particles begin to roll rather than break causing some dilation in the later stages of testing.
Figure 9 shows the relationship between shear wave velocity (swv ), and effective confining pressure (s¢ c), which was obtained from a series of bender element measurements carried out under the isotropic conditions at the start of each of the triaxial tests above.The shear wave velocity values are plotted on a semi-logarithmic scale and the resulting gradient can be used to obtain shear-wave velocity, at atmospheric pressure (98kPa) which is a useful basis for comparison of the materials.
Values of shear-wave velocity can be used to obtain the shear modulus Gmax, sometimes called the small-strain or initial modulus. This modulus is typically associated with shear strain levels of about 10-3% and below and hence is an important parameter for a wide range of geotechnical design applications. Gmax is a key parameter in small strain dynamic analyses such as those used to predict soil behaviour during earthquakes, explosions or machine and traffic vibrations. It can be equally important for small strain cyclic situations such as those caused by wind or wave loading.In addition it is now well established that small-strain behaviour plays an important role in the soil response to static loading. It has been shown (Burland, 1989) that the strain levels around engineering structures lie in the range of very small to small strains (up to 0.2%) The value of the small-strain stiffness Gmax, can be obtained from measurements of shear-wave velocity using:
Gmax= rV s2where: ris the density of the sample Vs is the shear wave velocity, in this case at atmospheric pressure.
A shear modulus value of 67MPa was obtained for the gravel samples and 70MPa for the glass.
Thus at small strains, the glass would appear to have a stiffness comparable to that of gravel.
Summary of results
Table 4 shows a summary of the properties of each of the materials tested.
The larger glass particles gave an improvement over the smaller ones due to the increase in the angularity of the particles, giving a greater degree of interlock for shear strength and increased drainage paths for permeability.This results in a material which is nearly as effective as gravel when used in situations requiring high shear strength or permeability and has similar small-strain stiffness as the gravel. It would therefore appear that this material has the potential to be used as a replacement to the gravel in certain situations.
A series of basic tests have been carried out to assess the geotechnical performance of waste crushed glass. These tests have shown that the material has some potential for use as a fill or drainage material. The performance of the material depends on the grading of the sample and the angularity of the individual particles.For example, if the particles are small, then the shear strength and strain (either vertical or volumetric) behaviour of the sample is less influenced by crushing, but the permeability of the sample will be small.However, if the particles are large then the sample will be more permeable but its shear strength and strain will be more influenced by crushing. In addition, if the particles are angular rather than rounded then the shear strength of the material will be improved but the influence of crushing on the strain characteristics will be greater.
As the properties of the glass depend on the type of material produced in the recycling process, a more detailed study is required to match the gradings and particle shapes available with specific properties.This is essential if detailed specifications for the use of crushed glass are to be developed in the future.
J Blewett gratefully acknowledges the support of the Mott MacDonald Charitable Trust and the UK Engineering and Physical Sciences Research Council.
The authors would also like to thank B MacBride of MacGlass Recycling, Dalkeith, Edinburgh, UK for the glass samples used in this study.
Ahmed I and Lovell CW (1992). Use of waste materials in highway construction: State of the art practice and evaluation of the selected waste products.Transportation research record,1345, National Research Council, Washington DC,1-9.
Bishop AW and Wesley LD (1975). A hydraulic triaxial apparatus for controlled stress path testing.Geotechnique 25, No 4, 657-670.
Blewett J, Blewett IJ and Woodward PK (1999). Measurement of shear-wave velocity using phase-sensitive detection techniques.Canadian Geotechnical Journal,36, No 5, October,934-939.
British Glass (1994).Waste glass recycling.Final report on the Etis project.
Burland JB (1989). Ninth Lauritis Bjerrum memorial lecture: 'Small is beautiful - the stiffness of soils at small strains.'
Canadian Geotechnical Journal,26,499-516.
Concrete international (1995).Waste glass research under way.15 April 1995.
Dyvik R and Madshus C (1985). Laboratory measurements of Gmax using bender elements. Proc. ASCE - Conference on advances in the art of testing soils under cyclic conditions, Detroit,186-196.
Fumagalli E (1969). Tests on cohesionless materials for rockfill dams. Journal of Soil mechanics and foundations division.
Hardin BO (1985).Crushing of soil particles. Journal of Geotechnical Engineering,111, No 10, Oct,1177-1192.
Lade PV (1997). Elasto-plastic stress strain theory for cohesionless soil with curved yield surfaces. International Journal of Solids & Structures,13,1019-1035.
Lade PV and Duncan JM (1975). Elasto-plastic stress-strain theory for cohesionless soils. Journal of Geotechnical Engineering division, ASCE,101, GT10,1037-1053.
Lade PV, Yamamuro JA and Bopp PA (1996).Significance of particle crushing in granular materials. Journal of Geotechnical Engineering,122, No 4, April,309-316.
Nicholson PG, Seed RB and Anwar HA (1993). Elimination of membrane compliance in undrained triaxial testing. Part 1.
Canadian Geotechnical Journal,30,727-746.
Viggiani G and Atkinson JH (1995).Interpretation of bender element tests.Geotechnique, 45, No 1,149-154.
Watt IAG (1991). Investigation into the causes and effects of particle breakage of a weathered sandstone material. MSc thesis, University of Newcastle upon Tyne (unpublished).
Woodward PK and Blewett J (1998).Shear strength of a sand/rubber composite at high normal stresses.Ground Engineering,31, No 9, September.