ACKNOWLEDGEMENTS. Special thanks are due to my friends close by and abroad for balancing the life. Finally, I wish to thank my parents for everything.

Size: px
Start display at page:

Download "ACKNOWLEDGEMENTS. Special thanks are due to my friends close by and abroad for balancing the life. Finally, I wish to thank my parents for everything."

Transcription

1

2

3 iii ACKNOWLEDGEMENTS The research presented in this dissertation was carried out at the Mechanical Process and Recycling Laboratory, Department of Materials Science and Rock Engineering, Helsinki University of Technology (HUT). Concrete castings were performed at the concrete laboratory of Fortum Technologies in Vantaa and the statistical modelling was carried out at the Laboratory of Computational Engineering, Department of Electrical and Communications Engineering, Helsinki University of Technology. The work was financed by Lohja Rudus Oy Ab, and the initiation for the dissertation was introduced by the Scancem Scientific Counsel. I would profoundly like to thank Professor Kari Heiskanen, under whose supervision this study was carried out. The technical support was never compromised whenever that was needed. I also wish to thank Professor Göran Fagerlund and Dr Ernst M rtsell for reviewing the dissertation and for their comments and suggestions on the text. Additionally, I want to extend my gratitude to Professor Jouko Lampinen for reviewing the statistical approach of the dissertation. The support of my employer, Lohja Rudus Oy Ab is gratefully acknowledged and especially the encouragement received from M.Sc. Martti Kärkkäinen and M.Sc. Kauko Linna is highly appreciated. I want to express my sincere gratitude to M.Sc. Ville Toivanen for diligent and intelligent assisting work, Mr. Tuomo Rimpiläinen and his group at Fortum for accurately performed concrete castings and M.Sc. Aki Vehtari for the demanding statistical modelling. Also the work of Mr. Ilkka Kalliomäki for the statistical program is gratefully appreciated. It has been truly enjoyable and easy to work with all of you. Many thanks are due to my colleagues at Lohja Rudus Oy Ab and the Laboratory of Mechanical Process and Recycle for their support and interest as well as for interdisciplinary and witty discussions. Special thanks are due to my friends close by and abroad for balancing the life. Finally, I wish to thank my parents for everything. Hanna Järvenpää

4 iv CONTENTS Abstract Acknowledgements Contents Definitions and notations ii iii iv xi 1. Introduction 1 2. Effects of aggregate characteristics on concrete properties Workability Effect of paste and water content Effect of aggregate grading, surface area and size Effect of aggregate shape, angularity and surface texture Effect of aggregate mineralogy Effect of aggregate absorption Effect of superplasticizer and air-entraining agent Air percentage Air-void formation and stability Effect of water-cement ratio Effect of aggregate grading Effect of aggregate shape, angularity and surface texture Effect of aggregate mineralogy Effect of superplasticizer Bleeding Definition of stability, viscosity and cohesion Effect of cement and workability Effect of aggregate surface area, grading and size Effect of aggregate shape, angularity and surface texture Effect of superplasticizer and air-entraining agent 17

5 v 2.4 Compressive strength Effect of water-cement ratio and aggregate-paste interface Effect of compaction degree Effect of aggregate size Effect of aggregate strength Effect of aggregate shape, angularity and surface texture Effect of aggregate surface area Effect of aggregate mineralogy Effect of superplasticizer and air-entraining agent Drying shrinkage Mechanism of drying shrinkage Effect of water-cement ratio Effect of aggregate content Effect of elastic modulus of aggregate Effect of aggregate grading, shape, size, angularity and surface texture Effect of aggregate shrinkage properties Effect of superplasticizer and air-entraining agent Data analysis methods and Excel program used Inputs outputs Bayesian statistics and Gaussian processes for prediction of the fine aggregate-concrete interaction Bayesian methods Gaussian Process Relevance values of inputs Deviance Information Criterion (DIC) for model evaluation Data pre-processing Model selection Model errors (prediction errors) 35

6 vi 3.3 Excel program for prediction of fine aggregate concrete interaction General principles of the Excel program Predicting the correlation in input variables Sensitivity analysis and its reliability Error estimation Experimental programme Materials Aggregate products Cement Admixtures Test programme Aggregate (inputs) Concrete (outputs) Mix designs and concrete mixes, mixing procedure, test specimens Mix designs and concrete mixes Mixing procedure Test specimens Testing methods and potential input values, aggregates Mineralogical composition, fines and semi-coarse fractions Specific surface area, fines Grading, fines Particle density, fines and semi-coarse factions Particle porosity, fines and semi-coarse fractions Zeta potential, fines Resistance to fragmentation, semi-coarse fractions Elongation, flakiness, particle volume and quantity, semi-coarse fractions Angularity/roundness, semi-coarse fractions Surface texture, semi-coarse fractions 61

7 vii 4.5 Testing methods and concrete output values Workability Air % and density of fresh concrete Bleeding Compressive strength and density of the hardened concrete Testing methods for drying shrinkage, weight loss and air-parameters, hardened concrete Drying shrinkage and weight loss Thin section analysis; air %, specific surface area and spacing factor of air void 67 5 Aggregate test results and discussion Mineralogical composition Grading Specific surface area Particle density Particle porosity Zeta potential Resistance to fragmentation Elongation, flakiness, particle volume and quantity Angularity and surface texture Discussion of the test results for aggregates Concrete test results and discussion Workability Air % Air %, fresh concrete Air %, hardened concrete Bleeding Compressive strength Drying shrinkage and weight change Results Discussion 118

8 viii 7. Models for the fine aggregate concrete interaction Model for the flow value Sensitivity analysis flow value Reliability of the sensitivity analysis flow value Flow value SC-flakiness 3.15/4.0 mm, SC-angularity and SC-elongation 3.15/4.0 mm Flow value SC-pore area Å and SC- pore area >900Å Flow value F- mica %, F- C u, F- BET value and F-zeta potential Model for air %, fresh concrete Sensitivity analysis air % Reliability of the sensitivity analysis air % Air % - SC-pore area Å and SC-pore area Å Air % - SC- flakiness 3.15/4.0 mm and SC-angularity Air % - F- C u and F- BET value Model for the bleeding Sensitivity analysis bleeding Reliability of the sensitivity analysis bleeding Bleeding SC- total pore area and SC- average pore area Bleeding SC- elongation 0.8/1.0 mm, SC- flakiness 1.6/2.0 mm and SC- elongation 1.6/2.0 mm Bleeding F- BET value, F- zeta potential, F- density and C u Model for the compressive strength Sensitivity analysis compressive strength Reliability of the sensitivity analysis compressive strength Compressive strength SC- flakiness 3.15/4.0 mm and SC- quantity 1.6/2.0 mm Compressive strength SC- Los Angeles value Compressive strength SC- pore area Å Discussion of the models 147

9 ix 8. Predicting with the models Principles of the predictions Combined effect of two input characteristics Predictions with different solutions for concrete aggregate combination Predicting with the flow value model SC- pore area Å and SC- flakiness 3.15/4.0 mm vs. flow value F- C u and SC- flakiness 3.15/4.0 mm vs. flow value Effect of different aggregate combinations on the flow value Predicting with the air % model SC- pore area Å and SC- flakiness 3.15/4.0 mm vs. air % SC- pore area Å and F- C u vs. air % Effect of different aggregate combinations on the air % Predicting with the bleeding model SC- total pore area and F- BET value vs. bleeding SC- total pore area and SC- elongation 0.8/1.0 mm vs. bleeding Effect of different aggregate combinations on the bleeding Predicting with the compressive strength model SC- Los Angeles value and SC- flakiness 3.15/4.0 mm vs. compressive strength SC- Los Angeles value and SC- pore area Å vs. compressive strength Effect of different aggregate combinations on the compressive strength Discussion of the predictions made with the models Verification of the models with two new aggregtate products Procedure Identification of the new aggregate products Results of the modelled and measured values Evaluation of the verification of the modelled and measured values 183

10 x 10. Conclusion Need for future research 188 REFERENCES 189

11 DEFINITIONS AND NOTATIONS xi Aggregate Granular material used in construction. Aggregate may be natural, manufactured or recycled. Natural aggregate Aggregate from mineral sources which has been subjected to nothing more than mechanical processing. Natural fine aggregate Designation given to smaller aggregate sizes with upper nominal size less than or equal to 4 mm. Fine aggregate can be produced from natural disintegration of rock or gravel and/or by the crushing of rock or gravel. Fines Particle size fraction of an aggregate, which passes the mm sieve (aggregate testing) or the mm sieve (concrete castings). AE air-entrainment (concrete) ARD Automatic Relevance Determination DIC Deviance Information Criterion F fines (< mm or mm see above) FG future gravel GP Gaussian Process gsh good shape gst good strength (Los Angeles value) MCMC Markov Chain Monte Carlo N no admixture (concrete) PG past gravel psh poor shape pst poor strength (Los Angeles value) SC semi-coarse SCF semi-coarse fraction ( mm) WR superplasticizer (concrete)

12 1 1. Introduction For the aggregate producer, the concrete aggregates are end products, while, for the concrete manufacturer, the aggregates are raw materials to be used for mix designs and successful concrete production. With the aggregate production, the quality of the aggregate products can be influenced, but the raw material the gravel or rock - may have characteristics which cannot be modified by the production process. Similarly, with the concrete mix designs can be influenced how the aggregate affects the properties of concrete, but there is also a limit, whether technical and/or economic, in the mix design modification after which it is useful to select a more suitable aggregate product. Concrete aggregates have been studied relative largely in the past decades, though most of the research has been done to the coarse aggregate and only to one or few quality characteristics at a time. In order to optimise the aggregate-concrete chain, one has to know what are the aggregate quality characteristics that dominate different concrete properties, and how basic changes in the concrete mix design affect the influences. The need for knowledge is increasing as conventional concrete aggregate supplies are becoming depleted, and environmental aspects prevent the use of existing sources. The objectives of this work are: 1. To determine which the most important fine aggregate characteristics are that affect the concrete workability, compressive strength, air %, bleeding and drying shrinkage. 2. To determine how the aggregate characteristics affect the concrete properties separately and together. 3. To determine how basic changes in the concrete mix design, i.e. change in paste amount and admixtures (air-entraining agent, superplasticizer) affect the aggregate influences. 4. To become a program with which can be predicted how the fine aggregate concrete interaction affects the concrete workability, compressive strength, air % and bleeding.

13 2 When the objectives are fulfilled, the results are applicable for optimising both the aggregate and concrete production. The experimental studies were divided to aggregate tests for aggregate characteristics determination and to concrete tests where the tested aggregates were used for concrete production and observation of behaviour in concrete properties. To be able to relate the extent that the aggregate has on the concrete as compared against the effect of the changes in mix design, the test programme was built to monitor the aggregate behaviour in different mix designs including variations in admixtures and paste amount. The test programme was additionally constructed in such a way that the effect of the fines and the semi-coarse fraction could be interpreted separately. The fines are defined in this study as the fractions bellow 125 µm and the semi-coarse fractions as particle sizes between 125 µm and 4 mm. The aggregate characteristics and basic concrete mix design changes are regarded as input variables and the fresh and hardened concrete properties are the outputs to be modelled. Due to the restricted amount of data and lack of knowledge concerning which aggregate characteristics are relevant and their relationships to the concrete properties, it was decided to apply the Bayesian statistics and non-parametric non-linear Gaussian process models. The Bayesian statistics is based on learning processes where the prior information is combined with the evidence from the data. The results are treated as probability distributions expressing our beliefs regarding how likely the different predictions are. In a non-parametric model, the aggregate characteristics concrete properties relationship is determined from the data without reference to an explicitly parameterised model and thus, the possible different behaviour of aggregate inputs in the different mix design conditions can be concurred in the model. The adoption of a nonlinear model enables the possibility of non-linear conduct of the aggregate characteristics. The statistical modelling was performed by the laboratory of computational engineering at the Technical University of Helsinki.

14 3 The results of the models in this study are presented in the following formats: 1. ARD (Automatic Relevance Determination) listing gives the relevance values of the inputs, i.e. aggregate characteristics and basic mix design changes, in each model. The relevance value determines the distance of each input in the space of the model over which the input is expected to vary significantly. 2. Sensitivity analysis figures present how the output changes when an input variable is changed by a small amount. The predicted output is calculated and by comparing the change made in the input variable to the change perceived in the output, we see how the model reacts to changes in that particular input variable. The sensitivity analysis figures are presented for all inputs in each of the models except for SEM/N/AE/WR which are of nature on/off. 3. Sensitivity reliability figures present the difference between the modelled and measured values and thus, indicate how reliable the model is in general and on specific input-output-mix design combination. 4. Example predictions of combined effect of two aggregate input characteristics present how the models can be used for predictions for the combined effect on one output of two input characteristics over their total variation range. The result is shown by means of 3D surface charts and in the calculations the other characteristics of the model have fixed values. The charts are drawn for each mix design separately, as the mix design parameters (SEM, N/AE/WR) are additional influencing characteristics for the output. The example predictions are executed to one possible future solution for concrete aggregate combination. 5. Example predictions with different solutions for concrete aggregate combination present how the models can be used for predicting the effect of different aggregate products on the concrete properties. The results are shown in column charts and for each mix design separately, as the mix design parameters (SEM(N/AE/WR) are additional influencing characteristics for the outputs. In these predictions all the input characteristics are changed according to the aggregate combination. The example predictions are executed to three different solutions for concrete aggregate combinations; past and future gravel and combination of filler aggregate and crushed rock including predictions for the variation on shape and strength characteristics. 6. Verification of the models with two new aggregate products.

15 4 2 EFFECTS OF AGGREGTE CHARCTERISTICS ON CONCRETE POPERTIES 2.1 Workability Effect of paste and water content Any mix made with given materials and having a certain consistency will have a point at which the water ratio or the voids ratio is at a minimum. This point of maximum solids content also determines the optimum paste consistency. POWERS (1968) has defined the optimum paste consistency as follows: Optimum paste consistency is that consistency at which the solids content of the paste and the paste content of the mix are such that they produce the maximum solids content possible with the given materials.. Several commercial programs for particle packing prediction are now available, e.g. LPDM (Linear Packing Density Model) and Europack. For the concrete mix to be plastic, the volume of the cement paste must be sufficient to fill the interstitial space of the compacted aggregate, plus an increment that causes a certain dispersion of the aggregate particles (figure 1). (POWERS 1968, KRONLÖF 1997) Figure 1. Dry compacted state of the aggregate skeleton (A) Aggregate particles dispersed in cement paste (B) (POWERS 1968)

16 5 In plastic mixes, it is not paste alone that causes the dispersion of aggregate particles and plasticity; the volume of the paste is always augmented by a certain amount of air. According to the excess paste theory, the consistency of concrete depends on two factors: the volume of cement paste in excess of the amount required to fill the voids of the compacted aggregate; and the consistency of the paste itself. If the aggregate/cement ratio or the cement content of the mix for given materials is kept constant, the workability is then governed by the water amount. (POWERS 1968) MØRTSELL (1996) proposed functions for workability prediction of mortars and concrete. The inputs of the functions are the flow resistance ratio ( λ Q ) which characterises the matrix phase and the air voids space modulus ( H ) which characterises the aggregate phase. For a given matrix and aggregate phase, the volume relations between the phases determine the workability of the mortar or concrete. The matrix phase consists of water, cement, pozzolanes, admixtures and filler (<0.125 mm aggregate). m Effect of aggregate grading, surface area and size When the grading is changed for given materials, then the surface area of the aggregate combination is also affected. The greater the exposed surface area is the more water and cement paste will be required to wet that area and, therefore, the less water and paste will remain to lubricate the mix and thus the lower its workability will be. Special mix proportioning methods based on the specific surface of the aggregate combination have been proposed by, e.g. SINGH (1959). In his method the specific surface is determined by the water permeability method or an estimation is made on the basis of a shape factor. The specific surface of a combination of spheres of fine aggregate, S f, is given by: S p1 p2 p3 p4 p5 p6 S f = Equation where

17 6 S = the specific surface of the smallest size group, i.e. No. 100/ No. 200 sieves (0.075/0.150 mm) p...p 1 6 = percentage weight of six groups in the fine aggregate, from the smallest to the largest ( mm) LOUDON ( ) studied the shapes of different fine aggregates and he arrived at a determination of angularity factor, f, where the angularity is expressed as the ratio of the specific surface of a size group to the specific surface of spheres of a corresponding size group. He suggested that f = 1.1 for a rounded fine aggregate, f = 1.25 for a fine aggregate of medium angularity, and f = 1.4 for an angular fine aggregate. The specific surface is thus shape corrected by multiplying the specific surface of spheres by an appropriate angularity factor. MURDOCK (1960) suggested a modified method that also takes the maximum particle size into account. In addition to having a relatively smaller surface area requiring wetting an increased maximum aggregate size also presents the possibility of denser packing. With a given quantity of paste, decreasing the percentage of fine aggregate decreases the surface area, and hence the surface tension, thus tending to increase the mobility of the mix. In lean mixes (those with a low amount of cement) and gap graded mixes, the percentage of fine aggregate should, however, be high enough to ensure sufficient cohesion. Concrete mixes of which good mobility is required should also have an adequate surface area enabling good cohesion and shear resistance (see 2.2) Effect of aggregate shape, angularity and surface texture As was discussed earlier, the shape factors influence the specific surface; aggregates which are flaky, elongated and/or angular thus require more paste to wet the surfaces. The shape and texture of the aggregate also affect the bulk density of the aggregate skeleton and for rough, poor-shaped aggregate, the bulk density is therefore less than that of smooth, well-rounded particles of the same density owing to particle friction and

18 7 interference. Filling the voids and overcoming the friction call for a higher content of fine aggregate and water. A higher cement amount may also be necessary, to keep the strength constant. When a mix becomes richer, the angularity and grading of the aggregates become less important until, with high proportions of cement, the aggregate particles are little other than plums floating in cement paste. The test results obtained by MURDOCK (1960) show that, when the aggregate/cement ratio is reduced to 2, the effects of angularity and grading become negligible. KAPLAN (1958) studied 13 different coarse aggregates and came to the conclusion that increased angularity and/or flakiness lead to a reduction in the workability of concrete. Changes in the angularity, however, have a greater effect on the workability of concrete than changes in the flakiness. In Kaplan s studies flakiness caused 20% of the variation in the workability, whilst 59% was due to angularity. Although in his study there was a wide variation in the surface texture of the aggregates, Kaplan did not find any correlation between aggregate surface texture and the workability of concrete. In his research WILLIS (1967 studied nine different fine and coarse aggregates. He found that an equal change in shape characteristics caused fine aggregate to need two to three times more water than coarse aggregates. He also noted that the shape characteristics described the concrete behaviour best, whereas the mortar tests also included the effects of clay, mica and other deleterious materials. Owing to the method that was used to determine the shape (flow rate through an orifice), we can conclude that the findings reported by Willis are actually caused by a combination of shape, angularity and surface texture effects Effect of aggregate mineralogy Clay minerals are normally sheet-shaped i.e. they have more surface area than other minerals of the same grain size. The ratio of thickness to length for clay particles is

19 8 normally near 20. This makes the surface area of a clay particle nearly three times that of a cube of the same volume (non-expanding clays). Clays normally have a charged surface, and thus they attract charged ions and/or water molecules to adsorb on the surface. With some clays, the activity of the surface is increased by a sort of internal surface into which charged ions and water molecules can find their way (expanding clays). These absorbed ions and molecules expand the clay, and the surface area can be increased by a factor of 25 or more. (VELDE 1995) DANIELSEN AND RUESLÅTTEN (1984) studied micas in the size range of 0.15/0.30 mm. They found that micas have a negative effect on the workability properties of concrete. The effect is even greater for muscovite than for biotite, but only in the case of newly crushed, unweathered micas. For gravel-based, weathered micas, there is no difference in behaviour between muscovite and biotite. Particle degradation during mixing (flaky and elongated or otherwise mechanically weak particles) may cause an increased water requirement, slump loss, and a reduced air content of the air-entrained concrete. Additionally, if the aggregate particles have a coating which is soft or loosely adherent, the coating may be removed during the mixing and this would increase the fines amount of the grading Effect of aggregate absorption The mix design procedure now prevailing in Finland is based on the total water/cement ratio, i.e. the aggregate is considered to be in the bone-dry state. The mix design procedure most commonly applied in other countries is based on the effective water/cement ratio, which excludes the water absorbed by the aggregates (figure 2). This is also the case with EN 206-1, Concrete Part 1: Specification, performance, production and conformity.

20 9 Figure 2. Different aggregate moisture conditions (NEVILLE 1995) The effective water/cement ratio and the free water content are difficult to determine. For both coarse and fine aggregates, the absorption of bone-dry aggregate to the state of saturated and surface dry (SSD) is determined with standard tests that are hence accurate, though both methods have their own reproducibility and repeatability errors. Normally, it is assumed that, at the time of the setting of concrete, the aggregate is in an SSD condition. When aggregates are dry, e.g. in spring and summer the particles may quickly get coated with cement paste which prevents the further ingress of water necessary for saturation; in consequence, the effective water/cement ratio is higher than assumed. On the other hand, when the water/cement ratio is calculated on the bone-dry basis, the effective water content is always less than calculated recipe water. In this case, too, the effective water content varies according to the prevailing moisture content of the aggregate products and mix designs, as in richer mixes the coating effect of the cement tends to be quicker than in leaner mixes. (SINGH 1958, NEVILLE 1995) Most of the water absorption occurs by the outer layer of the aggregate particles. Some aggregates, especially gravel products, can have a weathered patina outer layer. The minerals of the outer surface can be altered and/or some minerals may have been leached away, causing enhanced porosity. The weathered gravel with a patina outer layer absorbs more than crushed product produced from the same raw material. This is due to the fresh, less porous unweathered surfaces, which appear during crushing. (KAPLAN 1958)

21 Effect of superplasticizer and air-entraining agent When an air-entraining admixture is used the water content and/or the share of fine aggregate can be reduced. An 1 per cent increase in air is equivalent to a 1 per cent increase in fine aggregate or a 3 per cent increase in the unit water content (ACI COMMITTEE 309, 1981). The reason for the improved workability brought about by the entrained air is that the air bubbles, kept spherical by surface tension, act as a fine aggregate having a very low surface friction and considerable elasticity. Figure 3 shows the indicative reduction of the water content as a function of the percentage of added air and the cement content. Figure 3. Reduction in the mixing water due to entrained air (NRCA 1993) For very lean mixes with an aggregate/cement ratio of 8 or more, and particularly when an angular aggregate is used, the improvement in workability caused by air entrainment is such that the resultant decrease in the water/cement ratio compensates fully for the loss of strength resulting from the presence of the voids. (POWERS 1968) Superplasticizers adsorb onto the surface of cement and aggregate particles and alter the electrical charge of the surface and/or cause physical interference (steric repulsion) between particles. The deflocculation and dispersion of cement and aggregate fines is thus enhanced and the workability is increased.

22 Air percentage Air-void formation and stability Entrapped air voids are unintentional voids. They are characteristically 1000 microns or more in diameter and, because the periphery of the voids follows the contour of the surrounding aggregate particles, they are usually irregular in shape. Entrained air voids, in the contrast, are spherical or nearly so, owing to the hydrostatic pressure to which they are subjected by the surrounding paste of water, cement and aggregate fines. These voids are typically between 10 and 1000 microns in diameter. (MIELENZ ET AL. 1958) Air-entraining agents adsorb at air-water interfaces, and thus the air voids that are formed during mixing become stabilised as they are covered by a sheath of air-entraining molecules that repel one another. Repellence prevents the coalescence of voids and ensures uniform dispersion of the entrained air. The soluble air-entraining agents will also precipitate on the surface of cement and aggregate particles, and will reduce the hydrophilic quality of the surface and render it hydrophobic. Air voids tend to cling to the hydrophobic surface of the particles. It is thus anticipated that void-particle adhesion is most significant for certain ranges of particle size. Studies of ore flotation indicate that particles between about 10 and 50 microns in sizes are most susceptible to void adhesion. (NEVILLE 1995, MIELENZ ET AL. 1958) Often there is a discrepancy between the air content measured in fresh concrete and air content determined in hardened concrete. Three mechanisms have been proposed for airvoid instability in fresh concrete (FAGERLUND 1990): 1. Loss of coarse air voids due to handling and compaction as the large bubbles move upwards by buoyancy 2. Dissolution of small bubbles in water as the bubbles collapse due to pressure caused by surface tension 3. Transfer of air from small bubbles to coarse bubbles as small bubbles coalescence with larger bubbles

23 Effect of water-cement ratio The amount of entrained air is smaller with lower water/cement ratios, i.e. with higher cement concentrations. WHITING (1985) has reported that dosages as much as ten times greater are needed for 6.5 ±1.0% entrainment for concrete with a w/c of (SSD) and a maximum aggregate size of 25 mm as compared against dosages used in conventional concrete mixes. The same phenomenon can be seen in figure 4, the air entrainment in cement paste where is presented for different w/c ratios and airentrainment agent dosages. (POWERS 1968) Figure 4. Air entrainment in cement paste as influenced by the w/c ratio and the dosage of air-entraining agent (POWERS 1968) Effect of aggregate grading The air content of concrete increases if the proportion of intermediate size ( µm) of fine aggregate is increased. The maximum size of the space subtended by intermediate particles varies from about 30 to 130 microns. The size range is suitable for enmeshing air-entrained bubbles that are big enough to withstand rapid dissolution in the mixing water. An increase in the finer sizes of aggregate or cement beyond the optimum

24 13 intermediate size decreases the air content because the available volume among the particles is decreased and hence the air bubbles become smaller. The smaller bubbles are subjected to greater pressure than bigger bubbles, thus increasing the dissolution of the bubbles. Further, an increase in coarse aggregate size decreases the available interstitial space of optimum dimensions and thus decreases the air content in concrete. (MIELENZ ET AL. 1958, SINGH 1959, PIGEON AND PLEAU 1995) Effect of aggregate shape, angularity and surface texture When the shape and/or angularity of the aggregate particles deviates from sphericity, the interstitial space between the particles decreases if the most compact arrangement of the particles is achieved. If good workability is required, the paste content of the mix design is increased, which enlarges the interstitial space and leads to successful air entrainment (NICHOLS JR. 1982, MIELENZ 1958). On the basis of his tests, SINGH (1959) concluded that angular particles derive great benefit for purposeful air entrainment as they resist compaction and thus increase the interstitial space between particles. BACKSTROM ET AL. (1958) studied eleven aggregates, including aggregates with smooth and rough surfaces. They found that surface texture had a to be rather striking effect on the values of the specific surface and spacing factor in air-entrained castings. The average value of the surface area was 742 in -1 for the smooth aggregates and 1037 in -1 for the rough aggregates. The average values of the spacing factor were and in., respectively. They found a fairly good correlation between the spacing factor and the freezing and thawing resistance with seven out of the eleven aggregates they tested. Four concrete castings expanded more than would have been expected according to the spacing factor. These aggregates all had smooth surfaces, and in petrographic examination they were found to contain appreciable amounts of weathered and physically unsound materials.

25 Effect of aggregate mineralogy The effect of aggregates on the air-entrainment agent function varies according to the chemical and mineralogic compositions and degrees of alteration (see 2.5.4). Aggregates composed with alkali earth and metallic ions, e.g. limestone, dolomite, blast furnace slags and glassy basalts, are expected to have the most considerable effect on the performance of the air-entraining agent. (MIELENZ 1958) Effect of superplasticizer If superplasticizers are used to increase the workability of concrete, the air content of airentrained concrete generally increases if the other mix design parameters are constant. In some cases, however, the air is not stable, i.e. the air-void system created during the concrete manufacturing changes before the concrete is hardened. This has been explained by two phenomena (PLANTE ET AL. 1989): 1. superplasticizers can entrain large bubbles, which are thus easily lost during handling and compaction 2. superplasticizers increase the paste fluidity, thereby promoting the coalescence of air-voids. 2.3 Bleeding Definition of stability, viscosity and cohesion Stability is defined as the flow of fresh concrete without applied force and is measured by bleeding and segregation characteristics. Bleeding occurs when the mortar is unstable and releases free water. Normal bleeding, which occurs in the form of uniform seepage, is not necessarily undesirable. It is, e.g. good preventive curing against plastic shrinkage cracking. Segregation is defined as the instability of a mix, caused by a weak matrix that

26 15 cannot retain individual aggregate particles in a homogeneous dispersion. Segregation is possible in the case of both wet and dry consistencies. Viscosity is defined as the quotient of shear stress divided by the rate of shear in a steady flow. The viscosity of the matrix can also be said to contribute to the ease with which the aggregate particles can move and rearrange themselves within the mix. Cohesion is defined as the force of adhesion between the matrix and the aggregate particles. It provides the tensile strength of fresh concrete that resists segregation. Internal friction occurs when a mix is displaced and the aggregate particles translate and rotate. (ACI COMMITTEE 309, 1981) Effect of cement and workability The fineness and the amount of cement greatly affect the bleeding tendency of concrete. Finer cement decreases this tendency owing to its larger surface area, earlier hydration and lower sedimentation rate. In addition, less bleeding occurs when cement has a high alkali and C 3 A content. (NEVILLE, 1995) A water content above that needed to achieve a workable mix produces greater fluidity and decreased friction. Additionally, the water-cement ratio increases; this reduces the cohesion within the mix and hence increases the potential for segregation and excessive bleeding. An overly dry mix may also result in loss of cohesion and dry segregation. (ACI COMMITTEE 309, 1981) Effect of aggregate surface area, grading and size The amount and surface area of the fine aggregate, especially that smaller than 150 µm, influences the bleeding of the concrete. The increased bleeding caused by the angularity of the fine aggregate can be controlled by the surface area. The bleeding tendency is

27 16 reduced by using a finer fine aggregate or by adding separate fines to the mix. The fines automatically contained in the crushed fine aggregate as a result of the crushing phenomenon is also suitable, though care should be taken that the amount is not too much. Mixes with gap grading normally require less water to achieve good workability than continuous grading with an otherwise similar recipe. Gap grading reduces the sizes of coarse fine aggregate and small coarse aggregate, and the tendency for bleeding and even segregation is enhanced if the concrete has a high workability without enough cohesion (cement, fines or air %). Additionally, if the fine aggregate fraction becomes coarse, the cohesion is reduced thus making the mix harsh and the tendency for bleeding increases. In contrast, as the fine aggregate becomes finer, the water requirement increases and the concrete mix becomes increasingly sticky. If the coarse aggregate has a large maximum size and if, in addition, the particles are flaky an excessively workable concrete should be avoided because pockets of bleed water may collect on the undersize of the coarse aggregate particles Effect of aggregate shape, angularity and surface texture Flakiness, elongation, angularity and surface texture of the aggregate, especially with the fine aggregate, all reduce the workability of concrete. The viscosity of the paste increases if only water is added, and if the surface area of the paste (cement, additives, fines and air) is too low, the extra water can overcome the cohesion and vigorous bleeding or even segregation can occur.

28 Effect of superplasticizer and air-entraining agent Superplasticizers generally reduce bleeding except if there is a very high slump when the concrete can become unstable and heavy bleeding or even segregation can occur. (NEVILLE 1995) Air entrainment also reduces bleeding. The reduction is caused by the displacement of the paste, the buoyancy of the bubbles and their surface area. (POWERS 1968) 2.4 Compressive strength Effect of the water-cement ratio and aggregate-paste interface Concrete is a heterogeneous material. Its properties depend on the properties of its component phases and the interactions between them. If concrete is fully compacted, the compressive strength for a given set of materials at a given age is inversely proportional to the water/cement ratio. It has been observed, however, if the water/cementitious material ratio and the fine aggregate/cement ratio of the concrete and mortar are constant, the cement paste has the highest compressive strength and ductility compared to mortar and concrete. Additionally, mortar has a somewhat higher compressive strength and a little more ductility than concrete, but otherwise possesses a similar stress-strain curve, figure 5.(MARTIN ET AL. 1991) According to DARWIN (1999) the lower strength of mortar and concrete results from stress concentrations induced in the cement paste by the aggregate particles. The stress peaks are due to differences in the elastic properties of aggregate and paste. Failure of the paste-aggregate interface also plays a role here, but generally to a lesser degree.

29 18 Figure 5. Stress-strain curves for concrete, mortar and cement paste with a water/cementitious material ratio of 0.5 (MARTIN ET AL. 1991) The weakness of the aggregate-matrix interface may be explained by the following phenomena: a) development of a higher porosity than the bulk matrix (higher w/c ratio) b) formation of larger crystal particles of the hydration products c) deposition of calcium hydroxide crystals with a preferential orientation on the interface MONTEIRO, MASO AND OLLIVIER (1985) found that the thickness of the transition phase is determined by the intensity of the surface effects produced by the aggregate. The thickness is larger for larger aggregates, and it is also a function of the size and shape of the fine aggregate particles. The surface effects originated by the fine aggregate particles interfere with those caused by the large aggregate, and the intensity of this interference determines the final thickness of the transition zone. PING et al. (1991) discovered, however, that for very fine limestone particles (radius mm) the transition zone was denser than bulk paste. They concluded this to be due to chemical reactions between limestone particles and portland cement.

30 Effect of compaction degree If the compaction of concrete is insufficient, the compressive strength is reduced. KAPLAN (1960) observed, for example, that the compressive strength of concrete with a voids content of 15% was reduced by approximately 72% when compared against the strength of fully consolidated concrete This result was irrespective of the mix proportions or the age at which the test was done. However, the reduction in strength due to a rise in voids up to a content of 15 % was much greater than that, owing to an increase from 15% up to 30%. The reduction percentage in concrete having a voids content of 30% was found to be 92%. WALKER AND BLOEM (1959) concluded that, at a given water/cement ratio the compressive strength of concrete containing up to about 10% entrained air, is reduced by approximately 5% for every 1% of air added. Their conclusion agrees fairly well with the results of KAPLAN (1960). WRIGHT (1953) concluded that the effect on compressive strength is materially the same, irrespective of whether the air is entrained intentionally in the form of numerous minute bubbles or occurs unintentionally in the from of large irregular voids Effect of aggregate size WALKER AND BLOEM (1960) have shown that, at a fixed water/cement ratio, strength decreases as the maximum size of aggregate increases, particularly for sizes larger than 38 mm (1½ in.). The optimum size tends to decrease with increasing strength. This phenomenon is caused by many parameters related to the heterogeneity of concrete, e.g. the interface zone, lower bond stresses between the aggregate particles and the matrix, maximum paste thickness and different dimensional changes of the paste and aggregate at both early and later age (ALEXANDER ET AL. 1961, LALLARD AND BELLOC 1997). However, reduction in the maximum aggregate size increases the specific surface of the aggregates and thus the incidentally entrapped air tends to be higher or if the workability is kept constant, the w/c ratio becomes higher. Both cases have a decreasing effect on the compressive strength, unless the workability is controlled with superplasticizer.

31 Effect of aggregate strength Most normal-weight aggregates have strengths much greater than the strength of the cement paste. Thus, up to concrete strengths of about 35 to 40 MPa, the effects of different good-quality aggregates are usually small. However, the aggregate strength required is considerably higher than the normal range of concrete strengths, because the actual stresses at the interface of individual particles within the concrete may be far in excess of the nominal compressive stress that is applied. In higher strength classes, the aggregate strength properties - which are also a function of particle shape - as well as the bond between the paste and aggregate begin to play a more important role. As the concrete is a heterogeneous material, the best compressive strength results are achieved with aggregate, which has a high strength (e.g. a good Los Angeles value) and low modulus of elasticity, i.e. a modulus of elasticity that is not very different from hydrated cement paste. When the elasticity values are closer to each other, the bond stresses are lower; thus less microcracking is induced and higher compressive strength values can be achieved. For flexural strength, the compatibility of the modulus of elasticity is even greater. (NEVILLE 1995) Effect of aggregate shape, angularity and surface texture WILLIS (1967) found that the shape of the fine aggregate had a markedly greater effect on compressive strength than the coarse aggregate. Fine aggregate influenced the compressive strength primarily through its effect on the need for mixing water, whereas with coarse aggregate, other factors in addition to the water requirement affected the compressive strength, e.g. elasticity, bond and mineralogy. ALEXANDER (1959) concluded that if even slightly angular projections or depressions are present on the surface of an otherwise smooth aggregate pebble, the mechanism of tensile failure can change from a preferential rupture of the bond to a preferential rupture through the paste in the region of the surface irregularity. (Figure 6. )

32 21 A B B C Paste Aggregate A B C Figure 6. The rupture mechanism depends on the relationship between bond and paste strengths as well as on the degree of irregularity on the surface of the crushed aggregate particle. ALEXANDER (1959) KAPLAN (1959) studied 13 different coarse aggregates and found, that the most important factor in coarse aggregate affecting the compressive strength was the surface texture. A rougher surface results in a greater adhesive force between the cement matrix and the aggregate. In this study, the surface texture was determined by comparing the traced line length from a magnification of 125 times against the length of an unevenness line drawn as a series of chords. One explanation for surface texture is the porosity of the particle surface. A porous, dry surface absorbs water and thus positively influences the bond between the aggregate particle and the paste. Additionally, if the aggregates are drier than SSD, the water/cement ratio will be reduced by the absorption of the aggregates and, consequently, the strength will increase. (STOCK ET AL. 1979, NEVILLE 1995) When it comes to the effect of the shape, angularity and surface texture, it is somewhat difficult to compare the results obtained by researchers, because nearly all the studies have been conducted using different testing methods to determine the same characteristics. Also, the terminology is overlapping to some extent, e.g. the line used to distinguish between the surface texture and angularity is vague.

33 Effect of aggregate surface area STOCK ET AL. (1975) conducted tests to study the effect of the aggregate concentration on the compressive strength of concrete. The results show that the strength of cement paste in tension and in compression is reduced by the addition of 20% by volume of graded aggregate, and it fell to a minimum value at a volume fraction of 30% to 35% and then increased with a further addition of aggregate. When the specific surface of aggregate is increased for a constant mix proportion, the amount of cement relative to the surface of the aggregate decreases. LALLARD AND BELLOC (1997) state that as the maximum paste thickness (MPT) between aggregate particles decreases the compressive strength increases. In dry packing of particles, it has been observed that the highest stresses exist at the contact points of aggregate particles. Thus, when paste is introduced into the packing and it is placed between two close aggregates, the paste will be highly stressed, yielding a greater matrix strength. The results of GOBLE AND COHEN (1999) also showed that the mortar strength increased and the strain-stress behaviour became more ductile as the quantity of the transition zone material was increased, i.e. as the aggregate surface area was increased. They comment, however, that increasing surface area causes stiffer mixtures, which is probably why in the test series performed by SINGH (1958) it was noted that the increase in the aggregate surface area caused more voids around the surface of the aggregate particles and thus a decrease in compressive strength Effect of aggregate mineralogy The mineral size, texture and mineralogical composition as well as the shape, angularity and surface texture affect the strength properties of the aggregate products. Additionally, the electrostatic conditions as well as the behaviour together with admixtures, additives and cement depend on the mineralogy. Some chemical bond may exist between the

34 23 aggregate and cement paste in the case of limestone and dolomite aggregates and possibly also siliceous aggregates. (NEVILLE 1995) In their studies, DANIELSEN AND RUESLÅTTEN (1984) found that altered feldspars (An rich plagioclases) have an almost continuous transition zone from the mineral phase to the cement paste phase. For unaltered feldspars, the contact zone was completely discontinuous. They concluded that the altered feldspars, with their cation deficiency in the crystal structure, make the diffusion of Ca from the cement paste into the Si-Al framework possible. A similar phenomenon also occurs with mica minerals during weathering. While unweathered mica (0.15/0.30 mm) caused a loss of strength in mortar, mortar made with weathered mica didn t deviate form the strength of the reference mortar. Potassium leached during the weathering process helps the hydrated calcium ions to find adsorption sites on the mica surfaces. When mica is present in the coarse aggregate, the most important factor is not the total amount of mica but its distribution. If the mica is in bundles, then even smaller amounts of mica can be detrimental, though its effect can be also seen from the strength determinations Effect of superplasticizer and air-entraining agent Superplasticizers are used to increase the workability, to reduce the w/c ratio and/or to save cement. Changes in the w/c ratio and cement amount have clear effects on compressive strength. By reducing the water without compromising the workability, the 24-hour early strength can be increased by 50% to 75%. Owing to the better dispersion of the cement particles, a greater amount of reactive surface area of cement is exposed, which can also lead to increased compressive strength. (NEVILLE 1995) The effect of entrained air has been discussed in chapter

35 Drying shrinkage Mechanism of drying shrinkage Concrete holds water in various states with different bonding energies. These are capillary water, which is free from the influence of surface forces, adsorbed water; which is bound to a solid surface; and interlayer water, which penetrates between a pair of solid surfaces. Drying shrinkage is observed as a result of the forces of contraction arising as the water is removed by drying. There are many models for determining the drying shrinkage of concrete. There is, however, widespread agreement that the dominant factors are the modulus of elasticity of the aggregate and cement paste (or their ratio), the aggregate content and, aggregate and paste shrinkage. (PICKETT, 1956, HANSEN AND NIELSEN 1965, HANSEN AND ALMUDAIHEEM 1987) Effect of water-cement ratio Shrinkage is greater the higher the water-cement ratio is, because the w/c ratio determines the amount of evaporable water in the cement paste and, additionally, the rate of evaporation. BROOKS (1989) concluded that the shrinkage depends on the water/cement ratio up to a w/c ratio of approximately 0.6, after which the additional water in the cement paste takes the form of free water. Unlike the physically (adsorbed) and chemically (interlayer) bound water, the free water does not contribute to shrinkage. Hence, the change in the volume of drying concrete is not equal to the volume of water removed.

36 Effect of aggregate content The aggregate in concrete restrains the drying shrinkage; this explains the higher aggregate content the smaller shrinkage with a constant w/c ratio. According to the model of HANSEN AND ALMUDAIHEEM (1987), the shrinkage decreases by about 18% when the aggregate content is changed from 65% to 70%. This change is independent of the w/c ratio, though the restraining effect of the aggregate is more pronounced with an increasing w/c ratio. The effect of the aggregate content on concrete shrinkage has also been reported by, e.g. PICKETT (1956) Effect of elastic modulus of aggregate The total restraining effect of aggregate depends not only on the volume concentration of the particles but also on the elastic properties of the particles and paste. The modulus ratio is defined as the ratio of the elastic modulus of the dispersed particles to the hydration products. For normal-weight concrete, the modulus ratio is typically in the range of 4 to 7. According to the model presented by HANSEN AND ALMUDAIHEEM (1987), the difference in dying shrinkage of concrete having a volume of aggregate in the range of 60% to 80% is about 30% when the modulus ratio increases from 4 to 7. When the effect of same change in the modulus ratio is predicted with the model by presented HANSEN AND NIELSEN (1965), the decrease in drying shrinkage is, however, only 8%. The reason for this difference between the two models lies in the calculation of Young s modulus of elasticity, especially, how the aggregate effect is taken into account Effect of aggregate grading, shape, size, angularity and surface texture The effect of aggregate grading, shape and size on concrete shrinkage is indirect and depends on how these influence the amount of water amount in the concrete. On the other hand, aggregate properties that enhance the bond between the paste and aggregate,

37 26 e.g. surface texture, angularity and porosity (see 2.3.6) decrease the drying shrinkage. (ACI COMMITTEE 221, 1997) Effect of aggregate shrinkage properties Some aggregates are known to shrink on drying. In most cases, these aggregates also have a high water absorption. Generally, aggregates containing quartz or feldspar and granite, limestone, dolomite as well as some basalts can be classified as low-shrinkage producing aggregates. Aggregates containing sandstone, shale, slate, graywacke, or some types of basalt have been associated with high-shrinkage concrete. However, the properties of a given aggregate type, such as granite, limestone or sandstone, can vary considerably within different sources. This can result in significant variation in the shrinkage of concrete made with a given type of aggregate. (ACI COMMITTEE 221, 1997) In their studies, HANSEN AND NIELSEN (1965) concluded, that if any appreciable shrinkage occurs in the aggregate material, the restraining effect of the particles is reduced and that it is not usually possible to bring the concrete shrinkage within reasonable limits by adjusting the composition of the concrete mix. Similar results were reported previously by CARLSON (1939), as can be seen from table 1. Table 1. Drying shrinkage of concrete with different aggregates (CARLSON 1939) Aggregate Particle density [ Mg/m 3 ] Absorption [ % ] 1-year drying shrinkage, RH 50% [ o/oo] Sandstone Slate Granite Limestone Quartz

38 27 The presence of clay on the aggregate lowers its restraining effect on shrinkage. Moreover, because the clay itself is subject to shrinkage, clay coatings can increase the shrinkage by up to 70%. (POWERS 1959) Effect of superplasticizer and air-entraining agent If superplasticizer is used for water reduction then two opposite phenomena affect the drying shrinkage. A lowered w/c ratio reduces the shrinkage, whereas the enhanced dispersion of cement increases the effective surface area of the paste and thus increases the shrinkage. BROOKS (1989) studied five different plasticizers and superplasticizers in water reduced and cement reduced concrete mixes and found that the admixtures increase the deformation (shrinkage and creep) by 3% to 132% compared to plain concrete. His suggestion was that for admixture flowing concrete (high workability), the deformation expectation should be increased by 20%. Entrainment of air has been found to have no effect on shrinkage. (KEENE 1960)

39 28 3. DATA ANALYSIS METHODS AND EXCEL PROGRAM USED 3.1 Inputs outputs We studied how the fine aggregate characteristics affect the concrete properties. To be able to relate the extent of the effect that the aggregate has on the concrete compared to the effect of the mix design changes, the testing program was build to contain six different mix designs in which 21 fine aggregate products were studied altogether in 215 castings. See section 4. The fine aggregate characteristics and mix design parameters are input variables, and the fresh and hardened concrete properties are the outputs to be modelled. (Figure 7) Mix design parameters Fine aggregate characteristics Air %, fresh concrete Flow value Bleeding Compressive strength INPUTS OUTPUTS Figure 7. Input output scheme These outputs have been modelled with the methods described in chapters 3.2 and 3.3. The models can be used with the Excel program described in chapter 3.4. Additionally, concrete drying shrinkage and the air % in hardened concrete were studied, but these were not modelled.

40 Bayesian statistics and Gaussian processes for prediction of the fine aggregate-concrete interaction Bayesian methods Bayesian statistical methods use probability to quantify uncertainty in inferences. The result of Bayesian learning is a probability distribution expressing our beliefs regarding how likely the different predictions are. The prior information from the problem is combined with the evidence from the data, giving the posterior probability of the solutions. Predictions are made by integrating over this posterior distribution. The effect of the prior information diminishes with increased evidence from the data and in the case of insufficient data, the prior dominates in the solution. The article of GELMAN ET AL gives a good introduction to Bayesian methods Gaussian Process As it is not known what the parameterised form of the input-output relationship should be, we use non-parametric non-linear Gaussian process (GP) models (RASMUSSEN 1996, ABRAHAMSEN 1997, MACKAY 1998, NEAL 1997, NEAL 1999). In a nonparametric model, the input-output relationship is determined from the data without reference to an explicitly parameterised physical model. Gaussian processes are a natural way of specifying prior distributions over possible relationships between the inputs and the output. In material science, Gaussian Processes have been applied, e.g. to the problem of predicting the microstructures of forged materials (BAILER-JONES ET AL. 1998) and the austenite formation in steel (BAILER-JONES ET AL. 1999). Based on the training data () 1 () 1 ( n) ( ) {( x, y ),...,( x y n )} D =, (having n data points), our primary purpose is to predict the new output, y ( n+1), for a new case where we have

41 30 observed only the new input vector, x ( n+1). With Gaussian processes predictive ( n+1) distribution of y is Gaussian, with the mean and variance given by ( + ) [ y ] 1 D k C y ( n+ 1) 1 [ y D] V k C k E n 1 Var = Equation 2 = Equation 3 where, C is the n by n covariance matrix of the observed targets () 1 ( n) { y y } y =,..., is the vector of known values for these targets k is the vector of covariances between V is the prior variance of ( n+1) y (i.e. ( n+1) y and the n known outputs ( ) ( 1) [ y ] n+ 1, y n+ Cov ). There are many possibilities for the covariance function, some of which are discussed in (RASMUSSEN 1996, ABRAHAMSEN 1997, MACKAY 1998, NEAL 1999). For example, a regression model based on a class of smooth functions can be obtained using a covariance function of the form () i ( j ) 2 2 ( x x ) d p 2 2 C ij = s exp ru u u + ij σ Equation 4 u = 1 The first term of this covariance function expresses that the cases with nearby inputs should have highly correlated outputs. The s parameter gives the overall scale of the local correlations. The r u parameters are multiplied by the co-ordinate wise distances in input space and thus allow for different distance measures for each input dimension. For irrelevant inputs, the corresponding r u should be small in order for the model to ignore these inputs. The second term is the noise model, where d = 1 when i=j. For the noise model, we tested normal and t 4 distributions. The t 4 distribution is Student's t distribution with 4 degrees of freedom, which is a quite safe and robust choice when the true noise distribution is unknown. ij It should be noted that this noise model is only for the outputs, and we assume here that the inputs are noise-free. This assumption is wrong (we know there are measurement

42 31 errors in input variables), but we assume that this simplification still gives the model acceptable accuracy. A noise model for the inputs would improve estimate of predictive distribution and would allow reconstruction of the regression over the true noiseless input but such a noise model would be more complex to implement and to use. (CARROLL ET AL. 1995, CORNFORD ET AL. 1998, WRIGHT 1999). Our prior knowledge is usually insufficient to fix the appropriate values for the hyperparameters in the covariance function (σ, s, and the r u for the model above). Therefore the hyperparameters are given prior distributions and predictions are made by integrating (averaging) over the posterior distribution for hyperparameters. This integration can be done using Markov Chain Monte Carlo (MCMC) methods (GILKS ET AL. 1996, GAMERMAN 1997, ROBERT & CASELLA 1999). In Monte Carlo methods expectations of integrals are approximated by using a sample of values drawn from the posterior distribution of parameters. In MCMC, samples are generated using a Markov chain that has the desired posterior distribution as its equilibrium distribution. We have used Flexible Bayesian Modeling (FBM) software (NEAL), which implements the methods described in (NEAL 1996, NEAL 1997, NEAL 1999). The Gaussian process specification used was gp-spec log nin / 0.05: :0.5:1 The noise model specification used was model-spec log real 0.05:0.5:4 The initial values for the model parameters were set as gp-gen log fix The MCMC sampling parameters were set as mc-spec log repeat 10 sample-variances heatbath 0.9 hybrid negate The length and the number of the chains and the burn-in length were decided using visual inspection of trends and the potential scale reduction method (GELMAN AND RUBIN 1992A, GELMAN AND RUBIN 1992B).

43 Relevance values of inputs In the GP model using a covariance function of the equation 4 the r u parameters are sometimes called Automatic Relevance Determination (ARD) (GIBBS 1997, NEAL 1997, NEAL 1999). The ARD parameter determines the distance to the particular direction in the n-dimensional space (n = number of inputs) over which the data point is expected to vary significantly, i.e. the ARD listing can be referred to as a listing of the relevance values of the inputs. We computed the relevance value for each input for each posterior sample of relevance parameters ( r u ). This yields a sample from the posterior distribution of the relevance values, which may be summarised to provide an estimate of the mean (asterisk) and median (diamond) values for each input, plus 25%-75% (box) and 10%-90% (line) quantiles. The quantiles describe the uncertainty of each input in relevance value. Figure 8 shows the relevance value listing of the inputs for the compressive strength 91 d model. (See chapter 7.4.) Higher value describes a higher relevance for the specific input. WR Flkn 3.15/4.0 mm AE Los Angeles QNTY 1.6/2.0 mm SEM Pore area Å Figure 8. Example of a relevance value listing of inputs, compressive strength 91 d Asterisk mean value; diamond median value; box % quantiles; line % quantiles

44 Deviance Information Criterion (DIC) for model evaluation The purpose of interpolation problems is not usually to obtain the closest fit to the data but to find a balance between fitting the data and making sensible predictions about new events. Hugely complex models are often over-parameterised and, while fitting the data precisely, they interpolate and extrapolate poorly. Within the classical modelling framework, model comparison takes place by defining a measure of fit, typically the deviance statistic, and complexity, the number of free parameters in the model. (GIBBS 1997, SPIEGELHALTER ET AL. 1998) Deviance Information Criterion (DIC) was recently proposed by SPIEGELHALTER ET AL. (1998) for comparison of arbitrarily complex Bayesian models. DIC is based on comparison of the posterior distribution of the deviance ( ) = 2 log p( yθ ) + 2log f ( y) D θ, where y is the observed data and θ are the lowest-level parameters directly influencing the fit. The standardising term f ( y) is a function of the data alone and hence does not affect model comparison. The fit of a model is summarised by the posterior expectation of the deviance [ D] D = E. θ y The model complexity is measured by the effective number of parameters p D = E θ y ( ) [ D] D E [ θ ] = D D( θ ) θ y p D, defined as

45 34 The fit and the complexity are then added to form a Deviance Information Criterion DIC = D + p D ( ) + D = D θ 2 p The DIC and quantiles for it can be easily obtained from the MCMC analysis Data pre-processing For computational reasons, input and output variables were normalised to have zero mean and unit variance (BISHOP 1995 P.298, NEAL 1999). Some of the outputs (air %, bleeding 60min) had values close to zero, but it is known that the values for these outputs are always greater than zero. In order to assure that the predictions and predictive quantiles for these outputs would always be greater than zero, log transformation was used Model selection First we made models with different noise models for each output with full set of potential inputs (see chapter 4.4.). To compare different noise models, we calculated the mean square error (MSE) and 90% quantiles of absolute error of the test data and Bayesian Deviance Information Criterion (DIC). The t 4 noise model was clearly better than the Normal noise model for all outputs. Then, using relevance values of the inputs, smaller sets of inputs were selected and new models were made (some inputs were favoured over others, based on expert knowledge e.g. BET vs. pore area (fines); see chapter 4.4. We continued this approach for each output until the DIC increased. The best model according to the DIC was selected and then, using backward selection, the input set was still reduced. The model with the lowest DIC was selected. If several models had statistically similar DIC values, the model with the least inputs was selected. Models

46 35 having the least inputs had similar errors compared to the errors of the model having all inputs. Depending on which output was modelled, seven to twelve input variables were needed. (see section 7.) Model errors (prediction errors) To estimate prediction errors we used a ten-fold cross-validation (10-CV) error estimate, i.e. nine tenths of the data was used for training and the one tenth was left out for error evaluation, and this scheme was repeated ten times (STONE 1974, GEISSER 1975, GELFAND AND DEY 1994). All the castings were used for inferences, but error estimates were computed only for castings with A and B aggregate products (no REF was used). Quantiles of estimated prediction errors were obtained by re-sampling. Cross-validation was used to produce cross-validation predictive densities (GELFAND 1996). Expectations and quantiles were then easy to estimate by re-sampling MCMC samples and data points. 3.3 Excel program for prediction of fine aggregate concrete interaction General principles of the Excel program When the desired input combination (fine aggregate characteristics and mix design parameters) are entered into the Excel program, it will calculate the expectation value for the output and 10% and 90% quantiles for the prediction ( 3.3.2) suggest adjustments to other input variables when one variable is changed ( 3.3.2) show how marginal changes of one input affect the specific output, i.e. it demonstrates, the output sensitivity to an input variation ( 3.3.3) show the reliability of the sensitivity analysis ( 3.3.3).

47 Predicting the correlation in input variables When a single input variable is changed by a large amount, we would like to take into account the correlation between the inputs. All other inputs should be adjusted in such a way that the new input vector is similar to those found in the training data set. This can be done by calculating the covariance matrix of the data and adjusting the other inputs according to the relative magnitude of the elements in the covariance matrix, x j, changed where σ = x j, unchanged + dxi. Equation 5 σ 2 ij 2 ii i is the index of the manually changed input dx i is the change made to that input j is the index of the input to be adjusted 2 σ ij is one element in the covariance matrix Σ For practical purposes, it is better to use a regularised estimate for the covariance matrix. This is done by Principal Component Analysis (PCA) (BISHOP 1995). First, the maximum likelihood estimate for the covariance matrix is computed with 1 Σ = N N i= 1 ( x ( i) µ )( x ( i) µ ) T. Then, the eigenvectors v i and eigenvalues λ i of the matrix Σ are computed, choosing M largest ones. The regularised estimate for the T covariance matrix is then Σ ~ = VΛV, where the matrix Λ has M largest eigenvalues λi on the main diagonal and the matrix V contains corresponding eigenvectors v i as columns. Using a regularised estimate has the advantage of making more conservative adjustments to the inputs because less significant and noisy correlation effects are ignored.

48 Sensitivity analysis and its reliability Sensitivity analysis answers the question: How does the output change when an input changes? At each data point, a single input variable is changed by a small amount (for example, ±2%) and the predicted output is calculated. By comparing the change made in the input variable to the change perceived in the output, we see how the model reacts to changes in that particular input variable. A useful graph can be made by plotting the input variable on the horizontal axis and connecting the predicted outputs of the original and changed inputs with a line (Figure 9). Sensitivity analysis Comp. Stength [MPa] N30 WR30 AE30 N35 WR35 AE SC-Flkn 3.15/4.0 mm Figure 9. Example of an input-output sensitivity analysis The slope of this line then represents the sensitivity of the model in one data point. If the lines are horizontal, the change in the input variable has no effect on the output. Upward and downward slopes suggest positive and negative effects in the output, respectively. Having two data points with different slopes close to one another does not necessarily mean that the model is incorrect; the change in the slope could be due to a large change

49 38 in other input variables, not illustrated in any way in the graph. The mutual correlation of the input variables is ignored in this analysis, as the changes made in the inputs are small. We would also like to estimate the reliability of our sensitivity analysis. By plotting the input variable of interest on the horizontal axis and connecting the predicted and measured output variables on the vertical axis with a line, we can identify ranges in the input variable where prediction errors are large, and thus where the model is not to be trusted (Figure 10). Sensitivity analysis - difference between modelled and measured values Comp. Stength [MPa] N30 WR30 AE30 N35 WR35 AE SC-Flkn 3.15/4.0 mm Figure 10. Example of the reliability of a sensitivity analysis In these areas, the results of the sensitivity analysis are likely to be incorrect as well. In contrast, where the prediction errors are small, the results of the sensitivity analysis are deemed to be more plausible.

50 Error estimation For determining the error in output measurements, we repeated 15% of the castings. From these repetitions we determined the repeatability error of measurement; it is the median of the 90% quantile of the absolute error. It includes the error made by the laboratory personnel as well as errors arising from differences in the conditions during the repetitions. In addition, it includes the deviations within the raw materials between the repetition castings. The skewness of the error prediction distribution can be observed from the 10/50/90% quantiles of the 90% quantile of the absolute error. In the text is used notation, e.g. 40 mm (30 60 mm), in which the 40 mm is the median (50%) of the 90% quantile of the absolute error, and the values in parentheses give the 10% and 90% quantiles for the 90% quantile of the absolute error. The model error includes the repeatability error of the measurement and the errors due to the selected inputs and model. As discussed earlier, the model selection is normally done by choosing the model where the fit and complexity are optimised. The input selection is, of course, crucially important to the model performance. If the inputs aren t capable of describing the output phenomenon, then the lack of the fit of the model will be obvious and hence the model error will be large. Also, the model error describes the median of the 90% quantile of the absolute error and the notation in the text is similar to the repeatability error notation. The group average is the median of the 90% quantile for the mix design group results, i.e. castings without admixture (N), with superplasticizer (WR) and with air-entraining agent (AE). The largest group is that, where all the castings are included (all). When we compare the group average to the repeatability error, we can evaluate the variation caused by other factors, presumably that caused by the fine aggregate characteristics.

51 40 4 EXPERIMENTAL PROGRAMME 4.1 Materials Aggregate products Fine Aggregate products A total twenty-one (21) different aggregate products were used in this study. Thirteen (13) of these were gravel materials and eight (8) crushed rocks. Table 2 is presents the aggregate products, their material source, rock type, used size fractions and geographical origin. Table 2. List of the aggregate products Aggregate Material source Rock type Used size fractions fines / semi-coarse Geographical origin A2 Rock Granite F & SC Uusimaa A3 Rock Granite F & SC Uusimaa A6 Rock Mica gneiss F Pohjois-Savo A7 Rock Tonalite F & SC Uusimaa A8 Rock Garnet bearing granite F & SC Pohjanmaa A10 Rock Tonalite SC Pohjanmaa A15 Rock Mafic metavolcanite F Pohjois-Savo A16 Rock Gabbro F & SC Kanta-Häme B1 Gravel Granitic gravel F & SC Uusimaa B2 Gravel Granitic gravel F & SC Uusimaa B3 Gravel Granitic gravel F Kymenlaakso B6 Gravel Granitic gravel F Varsinais-Suomi B7 Gravel Granitic gravel F & SC Pohjois-Savo B8 Gravel Sandstone, granitic gravel F & SC Satakunta B9 Gravel Sandstone F Satakunta B10 Gravel Granitic gravel F Satakunta B11 Gravel Granitic gravel F & SC Uusimaa B12 Gravel Granitic gravel F & SC Päijät-Häme B13 Gravel Granitic gravel F & SC Päijät-Häme B14 Gravel Granitic gravel F & SC Uusimaa REF Gravel Granitic gravel SC Uusimaa

52 41 Aggregate REF is the normal concrete laboratory aggregate used in Finland and was chosen as a reference aggregate for this study. More accurate mineralogical compositions of the aggregate products are presented in tables in chapter 5. In order to control the mix design grading curve, the aggregate products were sieved to six (6) nominal size fractions; 0/0.125, 0.125/0.25, 0.25/0.5, 0.5/1.0, 1.0/2.0 and 2.0/4.0 mm. Down to 0.25 mm the sieving was performed with a Mogen sieving apparatus, and the two smallest size fractions were separated using a Sweco sieving machine. To ensure accurate sieving results the size fractions 1.0/2.0 and 2.0/4.0 mm were sieved twice and the size fractions 0.25/0.5 and 0.5/1.0 mm and 0/0.125 and 0.125/0.25 mm three times. The REF aggregate was already sieved to narrow nominal size fractions; 0.1/0.6, 0.5/1.2, 1/2, 2/3 and 2/5 mm. Only the finest size fraction, 0.1/0.6 mm, was sieved using the Sweco to obtain the fraction 0.125/0.6 mm. Appendix 1 presents the original gradings of the aggregate products. Coarse aggregate The coarse aggregate was a combination of two coarse aggregate products. The nominal sizes of the aggregate products were 5/10 and 8/14 mm. Both aggregate products were uncrushed granitic gravels from Uusimaa Cement The cement was Finnish rapid hardening cement, CEM IIA 42.5R. Table 3 presents its chemical composition and physical characteristics. As it was already known from the beginning that the total time for the castings would be long, extra care was taken to prevent the cement from ageing. To verify the quality of the cement along the castings, four (4) strength determinations were made. Table 4 shows the initial compressive strength results and the three determinations done during the castings.

53 42 Table 3. Chemical composition and physical properties of cement Chemical composition [ % ] Physical properties CaO 61.3 Property Value SiO Particle density 3.14 Mg/m 3 Al 2 O Specific surface (Blaine) 473 m 2 /kg MgO 3.0 Standard consistency 31.0 % Fe 2 O Setting time, beginning 90 min SO Setting time, end 150 min Na 2 O eq 1.3 TiO Table 4. Compressive strength determinations Age [ d ] Compressive strength [ MPa] Admixtures Two admixtures were used: a superplasticizer and an air-entrainment agent. The superplasticizer was a sulphonated naphthalene formaldehyde condensate, Mighty 150. A fatty acid soap-based product, Ilma-Parmix, was used as an air entrainment admixture. The air entrainment agent was diluted to solution of 1:19. Both substances are commercially available and commonly used admixtures in the concrete industry. The dry material content for the water reducer is 42% and for the diluted air-entraining agent 5%.

54 Test programme Aggregate (inputs) The test programme was divided into tests for the fines (F) and semi-coarse (SC) size fractions separately partly because this is natural due to the magnitude difference between them, but also in order to ascertain the effects of the characteristics in isolation from each other. This is especially important with the processed fine aggregate products, e.g. mixture of uncrushed gravel and rock products. For the purpose of better revealing the variations in the characteristics, the fines were tested as the size fraction 0/0.063 mm. Table 5 presents a list of the basic characteristics of the fine aggregate studied and the test methods used. It also shows to which size fractions the tests have been applied. Table 5. Basic characteristics of the fine aggregate studied and the test methods used Basic characteristic of the fine aggregate Test method Tested size fraction(s) Mineralogical composition Röntgen diffraction (X ray) 0/0.063 mm 2.0/4.0 mm Specific surface area 1. BET-analysis 2. Laser diffraction 0/0.063 mm Grading Laser diffraction 0/0.063 mm Particle density Helium pycnometer 0/0.063 mm 2.0/4.0 mm Particle porosity Mercury intrusion porosimetry 0/0.063 mm 0.5/1.0 mm Zeta potential Zeta potential 0/0.063 mm Resistance to fragmentation Los Angeles value, modified 4.0/5.6 mm

55 44 Elongation Image analysis 0.8/1.0 mm 1.6/2.0 mm 3.15/4.0 mm Flakiness Image analysis together with average particle volume measurement 0.8/1.0 mm 1.6/2.0 mm 3.15/4.0 mm Angularity/roundness Image analysis 0.8/1.0 mm 1.6/2.0 mm 3.15/4.0 mm Surface texture Image analysis 0.8/1.0 mm 1.6/2.0 mm 3.15/4.0 mm Concrete (outputs) The total quantity of castings was 215, which were divided into six different mix designs. Two different cement amounts, 300 kg/m 3 and 350 kg/m 3, were studied without admixtures and with two admixtures, superplasticizer and air-entraining agent. The quantity of the castings using different mix designs are presented in table 6. Table 6. The quantity of castings using different mix designs Mix Cement amount design [ kg/m 3 ] Admixture Number of castings N no admixture 34 N no admixture 41 WR superplasticizer 38 WR superplasticizer 30 AE air-entraining agent 29 AE air-entraining agent 43

56 45 The drying shrinkage and weight loss measurements were mainly performed on mix designs N35, WR30 and AE35 and there are therefore a greater number of castings with these mix designs. In order to determine deviations between castings, 15 % of them were repeated (see appendix 2). In addition, two mixes were repeated several times (5 and 6 times) along the castings to ascertain that no major changes had occurred during the one year that it took to conclude all the castings. As far as deviations within castings were concerned, the focus was on compressive strength results. Each compressive strength result is an average of three (3) cubes; this also applies to the results for density, because the same cubes were used for the density measurements. The concrete test programme consists of tests for both fresh and hardened concrete. The tests were selected so as to cover the majority of the building code requirements and the practical concrete tests performed during concrete production. They were also designed to reveal potential difficulties arising from variations in the aggregate inputs. The lists of tests are presented in tables 7 and 8 for fresh and hardened concrete respectively. Table 7. Measured characteristics of fresh concrete and test methods Concrete characteristics Test method Testing age (after mixing) Remarks Workability Slump Flow value 5 min 7 min All castings All castings Density of fresh concrete Unit mass of the concrete in 8 l container 10 min All castings Air %, fresh concrete Pressure method 12 min All castings Bleeding Bleeding test 10, 30, 60 min All castings

57 46 Table 8. Measured characteristics of hardened concrete and test methods Concrete characteristics Test method Testing age Remarks Compressive strength Compressive strength measurement 24 h 28, 91 d All castings Density of hardened concrete Particle density 24 h 28, 91 d All castings Drying shrinkage Measurement of the length changes Up to 231 d 64 of 215 castings Weight loss Measurement of the weight changes Up to 231 d 64 of 215 castings Air %, hardened concrete Determination from the thin section 56 ± 2 d (AE mix designs) 72 of 215 castings Specific surface area of air void Determination from the thin section 56 ± 2 d (AE mix designs) 72 of 215 castings Spacing factor of air void Determination from the thin section 56 ± 2 d (AE mix designs) 72 of 215 castings 4.3 Mix designs and concrete mixes, mixing procedure, test specimens Mix designs and concrete mixes Mix designs Six (6) different mix designs were applied to all 215 castings, as shown in table 9. The designs consisted of two different cement amounts, corresponding to low and high paste volumes, together with three admixture classes; no admixture, superplasticizer or airentraining agent.

58 47 Table 9. Mix designs The mix designs were based on the following rules: 1. Two cement amounts, 300 and 350 kg/m 3 2. Same water-cement ratio for both cement amounts 3. Both cement amounts without admixture (N), with superplasticizer (WR) and with air-entraining (AE) agent 4. Same starting workability with one fines/semi-coarse aggregate combination in low paste mix designs (B3/B14, slump 110±10 mm) 5. Same superplasticizer dosage % from the cement amount in the WR mix designs 6. Same air-entraining agent dosage % from the cement amount in the AE mix designs 7. The water amount includes the water in the admixtures 8. Same fines % for the cement amount in all mix designs 9. Same combined grading curve for all mix designs, though fines amount (passing- % mm) varies according to rule # 8 Mix design Cement amount [ kg ] Water [ l ] Aggregates [ l ] Fines [ kg ] 20%C Superplasticizer [ kg ] 1.2%C Air-entr. 1:19 [ kg ] 0.425%C Air [ % ] W/C ratio Paste amount [ % ] Paste amount [ % ] w/ air N N WR WR AE AE The mix designs were made with the assumption that the air % for the N and WR mix designs would be 1% and for the AE mix designs 5%. However, it was known that there would be great deviations between the actual measured values and the theoretical values. Actually, this is one of the points of interest in this study, and it has to be taken into account in the interpretation of the other results, e.g. compressive strength and slump/spread values.

59 48 Combined grading curve In all mix designs, the ratio of the fine and coarse aggregates was kept constant. The percentage of the fine aggregate in the total aggregate was always 42.8 % and thus that of the coarse aggregate was 57.2 % (5/10 mm 20.2 %-unit and 8/14 mm 37.0 %-unit). The target combined grading curve for all the mix designs is presented in table 10. The passing-% of the mm sieve is dependent on the cement content and is further slightly affected by the total aggregate amount of the mix designs. The target value varies from 3.1% to 4.1%. Table 10. Target combined grading curves for all mix designs Mix design mm mm mm mm mm mm mm mm mm N N WR WR AE AE Fine aggregate combinations The fine aggregate products had been sieved into narrow size fractions: i.e. fines, 0/0.125 mm, and five semi-coarse fractions, 0.125/0.25, 0.25/0.5, 0.5/1.0, 1.0/2.0 and 2.0/4.0 mm.. Figure 11 presents the ways in which the fine aggregate fractions were combined. The combination of the fines and semi-coarse fractions could range from: the same aggregate or two different aggregate products or a maximum of 4 different aggregate products both for the fines and semi-coarse fractions

60 49 A 100 % A 100 % C-D-E-F max 4 pc. <0.125 mm A 100 % B 100 % G-H-I-J max 4 pc mm - 4 mm Figure 11. Combinations of the fine aggregate combining; fines and semi-coarse size fractions Concrete castings The list of all concrete mixes for the 215 castings with mix design type and fine aggregate combinations is contained in appendix 2. The list also indicates repetitions of castings Mixing procedure The castings were mixed using a 50 litre Zyklos pan mixer, and the size of the batches varied between 27 and 32 litres. A larger batch size was needed when shrinkage and weight loss prisms were also cast. The filling order of the mixer was: coarse aggregate products, cement and fine aggregate products. The following figure 12, illustrates the mixing cycle. The total mixing time was five minutes.

61 50 TOTAL MIXING TIME FIVE MINUTES 1 min 2 min 5 min 1/3 of the water + possible admixture 2/3 of the water Coarse aggregate, cement, fine aggregate Dry mixing time, total 1 min Wet mixing time, total 4 min Figure 12. Mixing cycle, total mixing time five minutes Test specimens For the purpose of compressive strength determinations, nine 100 mm cubes were cast from each batch. The strength determinations were conducted on three ages: 24 h, 28 d and 91 d, each representing an average of three cubes. The same cubes were used for density measurements. One additional cube was cast for the preparation of thin-section samples. An impregnated pre-sample was made from each cube at the age of 56 ± 2 days, and only from the AE mix designs were made final tin-section samples. Other pre-samples were stored as a reserve information source in case some phenomenon might need extra clarification. For the drying shrinkage and weight loss determinations, two 100 x 100 x 500 mm prisms were needed. The determinations were made for 64 out of 215 castings. The measurements consisted of drying shrinkage and weight loss determinations at ages of up to 231 days.

62 Testing methods and potential input values, aggregates Mineralogical composition, fines and semi-coarse fractions The mineralogical composition of the fine aggregate products was determined, using the X-ray diffraction method, at the Geological Survey of Finland. For all the aggregates the determination was made for two size fractions, <0.063 mm and 2.0/4.0 mm. Both size fractions were ground to fine powder before sample preparation. The <0.063 mm size fraction was additionally tested using two oriented mounts for enhanced determination of clay minerals. From these concentrated and oriented samples the X ray diffraction spectrum was determined with the 2θ-angle region After analysis, one sample was treated with ethylene glycol for 24 hours and the other was heated for 1 hour at a temperature of 550 C. Following this, the X ray diffraction spectrum was recorded for the treated sample. By comparing the prior and post-treatment X ray spectra with the information obtained from infrared spectrometric analysis of untreated material, it was possible to recognise and identify even minor quantities of clay minerals. A semiquantitative estimation of individual minerals was done by using experimentally obtained absorption coefficients. The potential input values gained from the mineralogical composition: 1. F and SC clay, (clay amount, fines and semi-coarse), [ % ] 2. F and SC - mica, (mica amount, fines and semi-coarse), [ % ] 3. F and SC amphibole, (amphibole amount, fines and semi-coarse), [ % ] 4. F and SC quartz, (quartz amount, fines and semi-coarse) [ % ] Specific surface area, fines The specific surface area of the fines, <0.063 mm, was determined using a NOVA 1000 Gas Sorption Analyser by the Quantachrome Corporation (BET method) and Coulter s LS Particle Size Analyser, which is based on the laser diffraction (LD) principle.

63 52 BET method The basis of adsorption method is the fact that the amount of gas adsorbed on a gas-solid interface under specific conditions is proportional to the interfacial area presented. The most commonly used adsorption method is the adsorption isotherm derived by Brunauer, Emmet and Teller (BET) using simplifying asusmptions; the isotherm gives the amount n of gas adsorbed relative to the monolayer amount n m as a function of (1) the gas pressure p divided by the saturation vapour pressure p s and (2) a constant C that depends on the adsorption energy. [( 1 p / p )( 1 p / p + ( Cp / p ))] 1 n / nm = Cp / ps s s s Equation 6 The BET equation has generally been found to be very useful for physical adsorption for non-porous solids, e.g. aggregates, in the pressure range 0.05 < p/p s < 0.35, and it is usually satisfactory to determine just a single point near the upper limit of validity of the BET isotherm. Since the constant C for nitrogen gas is generally much greater than unity (generally ), 1-p/p s, can be neglected compared with C p/p s, and the BET adsorption isotherm is reduced to the relation: ( 1 p ) n = n / Equation 7 m p s The monolayer capacity n m can be calculated using equation Q* and the ideal gas equation and thus, the equation becomes where, pvm n m = ( 1 p / p s ) Equation 8 RT p = ambient pressure (atm) T = ambient temperature ( K) R = gas constant (82.1 cc atm/ K mole) V and p/p s are measured values

64 53 The total surface area S t of the sample can be expressed as S = n NA M Equation 9 t m ca / where, N = Avogadro s number (6.023 x molecules/mol) A ca = 16.2 Å 2, cross-sectional area for the hexagonal close-packed nitrogen at 77 K (temperature of liquid nitrogen) M = molecular weight of the adsorbate (nitrogen (N 2 ) 28 g/mol) By combining the equations Q* and Q* the total surface area equation becomes a form: S t ( 1 p / p ) pvnaca s = Equation 10 RT The degree of physical adsorption increases when temperature decreases. Thus, the nitrogen gas is allowed to adsorb on the sample surface at the temperature of liquid nitrogen (77 K). When the sample is transferred to ambient temperature, the adsorbed nitrogen gas starts to evaporate from the surface of the sample and this volume of desorbed gas is measured (RUMPF1990). The specific surface area is obtained when the total surface area is divided by the mass of the sample. Prior testing the samples were dried at 80 º C. LD method Laser diffraction (LD) is a method where the particle size distribution is determined from the light scattering information of different size particles. A sample, which has been dispersed either in gas or liquid, is lead through a coherent light (laser beam). When the coherent light meets the particle surfaces, the light scatters and thus, a diffraction pattern is formed. Both the scattered and unscattered light is then focused to a detector plane through a transform lens, figure 13.

65 54 Figure 13. Optical arrangement of laser diffraction method to obtain size distribution of particles (FELTON 1990) The unscattered light is focused to a point on the optical axis and the scattered light forms a pattern of rings around the central spot. The diffraction pattern is the net diffraction of all the separate particles. Movement of the particles does not cause movement of the diffraction pattern, because light scattered at an angle θ, will always give the same radial displacement in the detector, irrespective of the particles position in the illuminating beam. This diffraction pattern produced is known as the Fraunhofer diffraction pattern. From the diffraction pattern can thus, be determined the particle size distribution. (FELTON 1990, STANLEY-WOODS AND LINES 1992) In the Fraunhofer diffraction calculations, it is assumed that all the particles are spheres. As the spheres are the only particles having an equal diameter when measured either from projected area or volume, it is obvious that the shape of the particles affect the achieved result. For irregular particles the grading result depends on the orientation of the particles. The particles are detected as a set of spheres having the average diameter between the smallest and largest projected area diameter thus, generally causing wider span in the grading curve and possibly transfer of the average particle diameter. (WEICHTER 1986) The samples were tested in water dispersion. To ensure proper dispersion the samples were first wetted in a beaker by a small amount of water after which the beaker was placed in an ultrasonic bath for one minute.

66 55 The main difference between the two methods is that the BET method determines the actual surface area, including the surface area of accessible pores, by measuring the amount of adsorbed gas on the gas-solid interface, while in the LD method the area is calculated from the grading curve, with the assumption that all the particles are spheres and non-porous. The potential input values are (used denotation in bold): 1. F - BET value (specific surface area by BET, fines), [m 2 /g] 2. F - LD value (specific surface area by LD, fines), [m 2 /g] Grading, fines The grading of the fines, <0.063mm, was determined by means of laser diffraction analysis. The equipment used was a Coulter LS Particle Size Analyser. The principle of the method is described in chapter The potential input values are (used denotation in bold): 1. F - H f, [%], is the sum of the passing-% for sieve sizes of 2, 4, 8, 11.2, 16, 22.4, 31.5, 45, 63, 80, 125 µm, fines 2. F mm, [% ], is the passing-% for the sieve size of 8 µm, fines 3. F - C u, [ - ], is the ratio of the sieve sizes for which the passing-% is 60% and 10%, fines Particle density, fines and semi-coarse fractions The particle density measurements were made for size fractions <0.063 mm and 2.0/4.0 mm using a Quantachrome helium pycnometer, AccuPyc The principle of the Helium pycnometer method is that the volume of a sample is determined by measuring the pressure change of helium in a calibrated volume. The

67 56 density can be calculated when the mass of the sample is given. Prior testing the samples were dried at temperature of 80 º C. The potential input values are (used denotation in bold): 1. F density (particle density, fines), [ Mg/m 3 ] 2. SC density (particle density, semi-coarse fractions), [ Mg/m 3 ] Particle porosity, fines and semi-coarse fractions The particle porosity values are based on mercury intrusion porosimetry measurements and were conducted using a Micrometrics Poresize The determinations were made for the size fractions <0.063 mm and 0.5/1.0 mm. The principle of the method is based on the fact that mercury behaves as a non-wetting substance towards most materials. Consequently, it does not penetrate into the pores of the material and one must apply pressure to make it do so. The most commonly applied intrusion equation is the Washburn equation, which states that the pore size is inversely proportional to the applied pressure: where, ( 2σ cosθ ) P r = LV / Equation 11 P = pressure applied to the mercury σ LV = surface tension of the mercury surface θ = r = contact angle of the mercury radius of the capillary If the mercury intrusion method is applied to loose powder material, the result obtained includes both the accessible pores in the particles (intra-particle voids) as well as the voids between the particles (inter-particle voids). However, if the loose material is of single size, granular material the mercury fills the inter-particle voids space predominantly without pressure (VAN BRAKEL ET AL. 1981, KLOUBEK 1994) and thus, the result obtained represents intra-particle voids, i.e. particle surface porosity. The former

68 57 case applies to the tested aggregate fines (<0.063 mm) and the latter case to the tested aggregate semi-coarse fraction 0.5/1.0 mm. A contact angle of and a surface tension of 485 mn/m were used in the calculations. Prior testing the samples were dried at 105 C and then evacuated in a penetrometer before mercury filling to minimum 3 Pa. The measured range of pores was from 300 µm (3 Pa) down to 6 nm (200 MPa). The potential input values are (used denotation in bold) : 1. F avg. pore size (average pore size, fines), [µm] 2. F tot. pore area (total pore area, fines), [ m 2 /g ] 3. SC avg. pore size (average pore size, semi-coarse), [µm] 4. SC tot. pore area (total pore area, semi-coarse), [ m 2 /g ] SC pore area Å/ Å/>900Å (incremental pore area Å, Å and >900Å, semi-coarse), [ m 2 /g ] Zeta potential, fines The Zeta potential measurements were performed for the fines, <0.063 mm, using a Coluter Delsa 440. Because the zeta potential measures the average electric charge of the aggregate particle surfaces, the tests were carried out for samples without admixtures (N), with superplasticizer (WR) and with air entrainment agent (AE). The dosages of the admixtures were according to the mix designs, i.e. 0.2% and 0.07% from the amount of aggregate for the WR and AE measurements respectively. In order to obtain results that can be related to the environment of the concrete, the electrolyte was made of 4 litres of ion-exchanged water and 1.00 kg of cement. After mixing, the electrolyte was filtered twice, first with a coarse filter and then with a 0.22 µm membrane filter. The ph value of the electrolyte was The zeta potential

69 58 measurements were conducted by means of the following routine: 100 mg of fines was added to a beaker containing 100 ml of electrolyte. The beaker was placed in an ultrasonic bath for one minute to ensure proper dispersion, after which the beaker was placed on a magnetic stirrer. The sample was allowed to balance itself for 10 minutes before measurements were made. Before any admixture addition, the N (no-admixture) value was measured to ensure that the sample was clean of any impurities. After the admixture had been added, the sample was again allowed to balance itself for 10 minutes before measurements were conducted. The potential input value is (used denotation in bold): 1. F - Zeta pot. (Zeta potential value, fines), [ mv ] Resistance to fragmentation, semi-coarse fractions The European standard EN : Methods for the determination of resistance to fragmentation was applied. A slight modification was made to the tested size fraction, i.e. the size fraction was 4.0/5.6 though the smallest size fraction given in the EN standard is 4.0/8.0 mm. The shape properties of the aggregate products were not altered, e.g. by bar sieving, and thus the results resemble the product characteristics and not only the raw material characteristics. The potential input value is (used denotation in bold): 1. SC - LA value (mod.) (Los Angeles value (mod.), semi-coarse), [ % ] Elongation, flakiness, particle volume and quantity, semi-coarse fractions The determination of elongation, flakiness, particle quantity, surface area, angularity and surface texture area were all based on scanned images. The scanner was a normal office scanner (AGFA SnapScan 600). For image processing, an image analysis tool was developed.

70 59 The pre-processing of the scanned images The images were scanned in colour, and thus they are first converted to greyscale. The aggregate particles are recognised from the background using simple thresholding. The grey level distribution of the image has two peaks, one corresponding to the background and one to the particles. The minimum between the two peaks in the histogram marks the decision boundary: pixels which are darker than this boundary belong to the background and pixels which are lighter belong to particles. In order to reduce noise, the binary image produced by means of thresholding is then filtered using a 3-by-3 median filter. This operation marks each pixel black or white, depending on which was more common in the pixel s 3-by-3 neighbourhood. Next, a morphological opening operation is performed. This tends to even out the particle boundaries and removes small holes, caused by speckles in the image, from the particles. Finally, connected areas in the image are sought and holes possibly still left inside these areas are filled. The determinations of elongation and flakiness were conducted for three narrow size fractions: 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm. The same procedure was applied in the case of each size fraction: a. The theoretical geometric mean for the sphere diameter of the narrow size fraction was calculated (e.g. 3.15/4.0 mm 3.55 mm) b. Particles of the narrow size fraction were spread on a scanner screen c. The image of the particles was scanned with a resolution of 1000 pixels per inch d. The quantity and the areas of the particles were determined by an image analysis program. (The quantity was cross-checked by manual counting) e. The average area of one particle from the narrow aggregate size fraction was calculated and the pixels were transformed to mm value f. From the average area was calculated the average equivalent 2D diameter (=circle) g. ELONGATION = [ avg. equivalent 2D diameter / theoretical sphere diameter ] h. As the quantity of the scanned particles and the particle density are known the average particle volume the average equivalent 3D diameter (=sphere)

71 60 i. FLAKINESS = [ avg. equivalent 2D diameter / avg. equivalent 3D diameter ] The particle quantity per each size fraction was calculated from the mix information by means of the equation: ( Total amount of aggregates [ l ] * Percentage of size fraction [ % ] ) Average particle volume of size fraction [ dm 3 ] The percentage of the size fraction was calculated from the combined grading curve and was the same for all mix designs i.e. 14.0% for 3.15/4.0 mm, 9.5% for 1.6/2.0 mm and 8.0% for 0.8/1.0 mm. The total amount of aggregate can be seen from table 9. The elongation, flakiness and quantities was calculated for each mix according to the aggregate combination and mix design. The potential input values are (denotation in bold): 1. SC Elng 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm (Elongation of size fractions, 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm), [ - ] 2. SC Flkn 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm (Flakiness of size fractions, 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm), [ - ] 3. SC Qnty 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm (Particle quantity of size fractions, 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm), [ - ] 4. SC- Surface area 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm (Surface area of size fractions, 3.15/4.0, 1.6/2.0 and 0.8/1.0 mm), [ - ] 5. SC tot. surface (total surface, semi-coarse), [ - ] Angularity/roundness, semi-coarse fractions Angularity is determined from the morphological spectrum of the particle and is approximated by performing successive morphological eroding operations using diskshaped masks. The eroding operation rounds the edges of the particles and removes pixels from the particle, depending on the size of the mask and the ruggedness of the

72 61 border. The features are computed using masks with radii of 3, 5, 7, 10 and 12 pixels. The number of pixels eroded from the stone is then divided by the square of the mask radius. As a result a sequence of five numbers is acquired, representing the shape distribution of the stone border. An average of the five-number sequence was calculated, and for the purpose of input parameter, the angularity was additionally averaged according to the particle quantity of the three size fractions. (GONZALES AND WOODS 1993) Angularity is computed with regard to each aggregate particle independently. The results of a given image are then the expectation values of the independent particles. The potential input value is: 1. SC Angularity (angularity, semi-coarse), [ - ] Surface texture, semi-coarse fractions When the surface texture is determined, a square is fitted inside the aggregate particle and texture inside the square is considered. Because the size of the texture samples varies from particle to particle, the texture sample is copied periodically so that samples from different particles become comparable. A two-dimensional Fourier transform is computed from the texture sample. In the two-dimensional case the Fourier transform coefficients are too numerous to be useful alone as features. Instead, radial sums of the absolute values of the Fourier transform coefficients from [pi/2..pi/4] band is used. The input parameter was calculated as the weighed average of the particle quantity of the three particle sizes. (JAIN 1989) The surface texture is computed with regard to each aggregate particle independently. The results of a given image are then the expectation values of the independent particles. The potential input value is: 1. SC - Surface texture, (surface texture, semi-coarse) [ - ]

73 Testing methods and concrete output values All concrete castings were made and tests conducted at the Fortum concrete laboratory in Vantaa, except in the case of the thin-section analyses, which were carried out at the Finnish Research Centre in Otaniemi Workability For workability determination, the slump and flow value was applied according to testing methods ISO 4109 and SFS 5286 respectively (figure 14). The slump was measured 5 and 15 minutes and flow values 7 and 17 minutes after the mixing. Figure 14. The measuring of the slump and flow value The outputs are: 1. Slump 5 min, [ mm ] 2. Flow value 7 min, [ mm ]

74 Air % and density of fresh concrete The testing of the percentage of air in the concrete was performed according to the standard ISO 4848, as shown in figure 15. The testing was done 12 minutes after the mixing. Prior to the air % measurement, the 8 litre vessel filled with concrete using standard compaction routines was weighed, and in this way the unit density of the fresh concrete was determined. Figure 15. Measurement of the air % of fresh concrete The testing method for the air % measures all the air in the concrete: entrapped and entrained air. Therefore no judgement of the quality of the air can be made on the basis of this measurement. Consequently, thin-sections were made from the AE castings for the air parameter determinations (see 4.5.6). The outputs are: 1. Air %, fresh concrete, [ % ] 2. Density of the fresh concrete, [ kg/m 3 ] 3. Excess density of the fresh concrete, [ % ] (deviation from the theoretical density)

75 Bleeding The measurement of bleeding was done after intervals of 10 min, 30 min and 60 min. Approximately 2 litres of concrete was put into a plastic cylinder container and vibrated slightly. The container was covered with a lid and placed in a quiet place to rest. After a specified time the bleeding water was sucked out and weighed. The bleeding was calculated by means of the equation: Bleeding = V w / V, Equation 12 where V w = the amount of bleeding water, [ g ] V = the volume of the concrete With some concrete mixes, it was difficult to measure the bleeding water accurately due to the large overall amount of bleeding and the segregation tendency of the concrete. The bleeding liquid consisted of both water and very low viscosity cement paste, and hence the choice of the liquid was some times challenging. The outputs are: 1. Bleeding 10 min, [ g/cm 3 ] 4. Bleeding 30 min, [ g/cm 3 ] 3. Bleeding 60 min, [ g/cm 3 ] Compressive strength and density of the hardened concrete The compressive strength and density were determined at the ages of 24h, 28d and 91d. All the values are average for three 100 mm cubes. The density determinations were made as mass per volume calculations based on weighing the specimen in air and in water (figure 16).

76 65 Figure 16. Density determination of the hardened concrete The 28d and 91d specimens were stored in a climate room with RH >95% and temperature 20 ± 2 C. At the testing age the specimens were taken out of the climate room 3 hours before testing. The outputs are: 1 3. Compressive strength 24h, 28d and 91d, [ MPa ] 4 6. Standard deviation of the compressive strength 24h, 28d and 91d, [ MPa ] 7 9. Density 24h, 28d and 91d, [ kg/m 3 ] Standard deviation of the density 24h, 28d and 91d, [ MPa ] Excess density 24h, 28d and 91d, [ % ] (deviation from the theoretical density)

77 Testing methods for drying shrinkage, weight loss and air parameters, hardened concrete Drying shrinkage and weight loss For drying shrinkage and weight loss determinations two 100 x 100 x 500 mm prisms with measuring bolts in the ends were cast. After demoulding, the prisms were placed in water for curing for 6 days before weight and length measurements were conducted; these were set as the zero values for the drying shrinkage and weight loss determinations. The prisms were stored in a climate room with RH 40 ± 3 % and temperature 20 ± 2 C. Weight and length determinations were done at the ages of 14, 21, 35, 49, 63, 91, 119, 147, 175, 203 and 231 days. A general view of the climate room and of the set-up for the length measurement can be seen in figures 17 and 18. Figures 17. Figure 18. General view of the drying shrinkage climate room (left) Set-up for the drying shrinkage measurement (right)

78 Thin section analysis; air %, specific surface area and spacing factor of air void From all the castings one thin-section pre-sample was made by cutting a 30x50x10 mm prism vertically from the middle of a 100 mm cube. The pre-samples were impregnated with clear resin at normal atmosphere pressure, and thus the reactions in the concrete were interrupted. Final petrographic thin sections were prepared from all the AE castings. The pre-samples were vacuum impregnated with fluorescent coloured resin and glued on slides. Finally the samples were diamond cut and grind down to sizes 30x50 mm 2 x 25 µm. A detailed description of the preparation method is contained in standard NT Build 381. The analyses were performed using a Leica DM LP polarisation and fluorescence microscope, applying the modified point-count method described in standard ASTM C457 (NT Build 381). Air pores which had diameter < 0.8 mm were counted as air voids and pores with greater diameter as entrapped air pores. The minimum analysed air pore size was with diameter mm. The analysed parameters were: air void percentage, entrapped air percentage, specific surface of the air voids (mm 2 /mm 3 ) and spacing factor of the air voids (mm).

79 68 5. AGGREGATE TEST RESULTS AND DISCUSSION The test programme was constructed in such a way that the fines, < mm, and the semi-coarse fraction, 0.125/4 mm, could be interpreted separately. This is vital. The reason is that, although the testing methods for these two groups are different, the test results can correlate quite strongly, and thus one phenomenon can hide the other. As two examples one may give the results of the F-LD specific surface area for rock products vs. the SC- Los Angeles value and the F- BET specific surface area for gravel products vs. the SC- total pore area (figures 19 and 20). Los Angeles value [%] LD SSA fines vs Los Angeles value (mod.) A -aggregate R = LD specific surface area [m 2 /g] Total pore area [m 2 /g] BET SSA fines vs Total pore area SC B - aggregate R = BET specific surface area [m2/g] Figures 19 and 20. Correlations between the fines and the semi-coarse fractions; Examples: F- LD vs. SC- Los Angeles and F- BET vs. SC-total pore area It is inevitable that there will be fines and semi-coarse inputs such that the separate fines inputs will correlate with each other; the same applies to the semi-coarse fraction inputs. The correlations between the inputs are taken into account in the Excel program for the prediction of fine aggregate concrete interaction by Principal Component Analysis (see chapter 3). Appendix 3 provides lists of the correlations within the fines and semi-coarse inputs.

80 Mineralogical composition The X ray diffraction determinations are only semi-quantitative and thus the actual quantities are difficult to determine with high precision. The results from the determinations are presented in table with an accuracy of within 1% unit for the lower concentrations (<10 %) and 5 %-unit for the higher concentrations ( 10%). Table 12. Mineralogical composition for the rock products; X-ray determination A2 [%] A3 [%] A6 [%] A7 [%] A8 [%] A10 [%] A15 [%] A16 [%] F SC F SC F SC F SC F SC F SC F SC F SC Quartz Potassium feldspar Plagioclase Hornblende Biotite Chlorite Carbonate 15 Pyroxene 15 Garnet 2 2 Table 13. Mineralogical composition for the B1-8 gravel products; X-ray determination B1 [%] B2 [%] B3 [%] B6 [%] B7 [%] B8 [%] F SC F SC F SC F SC F SC F SC Quartz Potassium feldspar Plagioclase Hornblende Biotite Muscovite Chlorite Smectite Kaolinite 15 2 Vermiculite 4 1 Talc 1 2

81 70 Table 14. Mineralogical composition for the B9-14 gravel products; X-ray determination B9 [%] B10 [%] B11 [%] B12 [%] B13 [%] B14 [%] F SC F SC F SC F SC F SC F SC Quartz Potassium feldspar Plagioclase Hornblende Biotite Muscovite Chlorite Smectite Kaolinite 7 Vermiculite 7 Talc 2 The mineralogical composition of the SCF and the fines can vary substantially both with the rock and the gravel products (tables 12-14). For the rock products, the reasons for the deviation lie in the heterogeneous of the rock product itself, enriching of minerals during the crushing process and representativeness of the samples. The gravel products, on the other hand, have been through weathering processes and hence the mineralogical composition, especially of the fines, has changed during a long period. The weathering of the gravel products can indirectly be measured, e.g. using the BET method, and as general, such products which have larger amounts of clay minerals show also higher values in the BET determination. 5.2 Grading The measurements were performed on three test specimens and each specimen being determined twice. The average coefficient of variation of the surface area (calculated from the grading information) was 4.6 ± 1.9%.

82 71 The potential input values i.e. fineness, shape of the grading curve and one passing-% value (0.008 µm), were chosen so that the grading curve could be described by them and its orientation would be unambiguous. Figure 21 illustrates the three potential input values from the grading curve for each of the fines. The scatter in the values was in general greater for gravel products than for rock products. Thus, the lowest and highest fineness values for the gravel and crushed rock products were 338 (B14) 644 (B11) and 425 (A8) 636 (A15) respectively. The 8 µm passing-% values varied with the gravel products within the range of 5.9% (B3) and 34.6% (B11), whereas the rock products had values within the range of 13.0% (A8) and 27.9% (A15). Furthermore, the C u values for the gravel products varied from 3.4% (B3) to 16.5% (B10) and for the rock products from 10.1% (A8) to 16.2% (A3). The mix values for the grading inputs were calculated proportional to the percentage of the different fines. Fineness, 8 µm passing-%, Cu Fineness Fineness 8 µm p-% Cu B1 B2 B3 B6 B7 B8 B9 B10 B11 B12 B13 B14 A2 A3 A6 A7 A8 A15 A16 Fines µm p-% and Cu Figure 21. Values for fineness, 8µm passing-% and C u for each of the fines

83 72 Figures 22 and 23 show the grading curves of the finest and coarsest gravel and rock fines. Figure 22 presents the cumulative grading curve and the figure 23 shows the same information in differential format. Cumulative grading curve Volume [%] A08 70 A15 60 B B Particle diameter [µm] Figure 22. Cumulative grading curves of A8, A15, B3 and B11 products, fines 10 Differential grading curve Volume [%] A08 A15 B03 B Particle diameter [µm] Figure 23. Differential grading curves of A8, A15, B3 and B11 products, fines

84 73 As can be seen from figure 22, the shape of the cumulative grading curves for the rock products are very similar to each other. In contrast, the shape of the cumulative grading curves for the gravel products vary considerably. This could of course be due to the choice of test material, but more likely, it results from the classification actions caused by the glacial periods and proceeded weathering. The rock products have been subjected only to compressive stressing processes, and thus the fines generated fall within narrower grading limits, (RUMPF 1990). The differential grading curve of the B11 reveals that the product has a fairly large amount of clay-size particles and thus indicates weathered material. This can additionally be seen from the SEM pictures, figures 24, of the fines as well as from their surface area values (see 5.3). B11 B3 A15 A8 Figure 24. SEM pictures of the fines B3, B11, A8 and A15. Magnification x 500.

85 74 The SEM pictures also accord well with the fineness and C u values of the fines. What can also be seen from the pictures is the shape properties of the fines. The gravel particles mainly have a cubical shape, whereas the rock particles are often flaky and/or elongated. The structure, mineral size and mineralogical composition of the rock has a vast influence on the shape of the particles. The exact amount of the crushed fines was known but was not used as a potential input, as in the everyday production it would not be possible to determine it. 5.3 Specific surface area The LD measurements were made 3*2 times (see 5.2) and the BET measurements three times. The average coefficient of variation was for the LD 4.6 ± 1.9% and 11.0 ± 9.2% for the BET. The specific surface area of the fines was determined by means of two methods: gas adsorption (BET method) and laser diffraction (LD). Of these two, the gas adsorption method can be considered a direct method and laser diffraction as an indirect method (see 4.4.2). The BET values of the fines varied between 1.77 m 2 /g (B3) and m 2 /g (B10) for the gravel products and between 1.25 m 2 /g (A8) and 2.73 m 2 /g (A15) for the rock products. The corresponding LD values of the fines varied from m 2 /g (B3) to m 2 /g (B11) for the gravel products and from m 2 /g (A8) to m 2 /g (A15) for the rock products. Figure 25 shows the results obtained from both determinations.

86 75 Specific surface area BET LD BET [m 2 /g] LD [m 2 /g] B1 B2 B3 B6 B7 B8 B9 B10 B11 B12 B13 B14 A2 A3 A6 A7 A8 A15 A Fines Figure 25. Specific surface area values determined using the BET method and LD The LD value describes the grading/fineness of the fines as one value. Because the method calculates the surface area, i.e. constitutes indirect determination, and assumes the particles to be spheres, it can be concluded that the value does not contain any information on the particle shape or the weathering properties of the fines. As BET is a direct method, it contains both types of information, although weathering is more prominent than the shape properties. In practice, this means that the BET value should be considered when the water-cement ratio of the mixes is determined with the bone-dry state of the aggregate products. Consequently, the LD surface area value should be considered when the w/c ratio is based on the saturated-surface dry (SSD) state of the aggregate products. Figure 26 shows pictures of the B10 and A7 fines that have close LD values, m 2 /g and m 2 /g, but very different BET values, m 2 /g and 2.44 m 2 /g.

87 76 B10 A7 Figure 26. SEM pictures of the fines B10 and A7. Magnification x 1500 The correlation between LD and BET is non-significant (R = 0.22), but if the gravel and rock products are interpreted as separate groups the correlation for rock products becomes significant (R= 0.76). The reason becomes clear if we consider the fact that the rock products are virtually non-weathered and that the main difference between the LD and BET results is thus due to the shape properties of the particles. 5.4 Particle density The average coefficient of measurement variation was 0.12 ± 0.03% and was determined from 12 aggregate fractions, 9 SCF and 3 fines. The determinations were carried out with three test specimens and each specimen was measured three times (3 x 3 times). The remaining 23 aggregate fractions were measured 1x 3 times. Fines The particle density values for the fines were quite constant for the gravel products. The values varied between Mg/m3 (B3) and Mg/m3 (B7). Greater variation, from Mg/m3 (A2) to Mg/m3 (A15), was observed in the fines of the rock products Figure 27 presents the particle density values for each fines.

88 Particle density B1 B2 B3 B6 B7 B8 B9 B10 B11 B12 B13 B14 A2 A3 A6 A7 Density [Mg/m3] A8 A15 A16 Fines Figure 27. Particle density values for the fines Semi-coarse fractions The gravel products showed a small scatter in the particle densities of the semi-coarse fractions. The lowest value was Mg/m 3 (B8) and highest Mg/m 3 (B7). For rock products, the scatter was greater: between Mg/m 3 (A3) and Mg/m 3 (A16). The results are presented in figure 28. Rock quarries are normally quite heterogeneous when it comes to rock types, and hence it is possible for particle density to vary significantly from one rock type to another. For example, a difference of 0.15 Mg/m 3 in particle density (2.80 Mg/m 3 instead of 2.65 Mg/m 3 ) in normal concrete with 1750 kg of aggregate affects the volume of 1 m 3 by 3.6 percentage, i.e. -36 litres/m 3.

89 Particle density 3.0 Density [mg/m 3 ] B1 B2 B7 B8 B10 B11 B12 B13 B14 REF A2 A3 A7 A8 A10 A16 Semi-coarse Figure 28. Particle density values for the semi-coarse fractions 5.5 Particle porosity The measurements were performed 2-6 times for each aggregate product (fines and SCF) and the average coefficients of measurement variation thus obtained are shown in table 15. Table 15. Average coefficient of particle porosity measurement variation Average coefficient of measurement variation Fines SCF Average pore size 8.7 ± 9.2% 14.0 ±12.5% Total pore area 6.9 ± 7.3% 17.7 ± 17.1% As can be seen from table 15, the accuracy is substantially better for the fines than for the SCF. This is mostly due to the disproportion between the sample size and the fraction size i.e. the sample size was 1-2 g and the tested semi-coarse size fraction was 0.5/1.0

90 79 mm. This equals approximately 100 grains. The sample size is same for the fines, which has size fraction <0.063 mm. The weathering of the aggregates dissolves minerals, i.e. increases particle porosity, disintegrates particles and causes mineral transformation. Hence, it is to be expected that several potential input values will correlate with the particle porosity values. The data source for the total pore area and average pore size is the same, and thus it is obvious that they display a good correlation (figure 29). 4.0 Avg pore size vs total pore area, fines 0.5 Avg pore size vs. total pore area, semi-coarse Total pore area[m 2 /g] y = x R = Average pore size [µm] Total pore area [m 2 /g] y = x R = Average pore size [µm] Figure 29. Correlations between average pore size and total pore area for the fines and semi-coarse fractions Fines The total pore area and average pore size values for the fines are presented in figure 30. The gravel products showed greater scatter in the values than did the rock products. For the total pore area, the gravel products had values between m 2 /g (B14) and m 2 /g (B10), and the rock products showed values between m 2 /g (A6) and m 2 /g (A7). The smallest and largest average pore size values for the gravel and rock products were µm (B10) µm (B14) and µm (A7) µm (A8) respectively.

91 80 Average pore size and total pore area 7 6 Avg pore size Total pore area Avg pore size [µm] Total pore area [m 2 /g] 0 B1 B2 B3 B6 B7 B8 B9 B10 B11 B12 B13 B14 A2 A3 A6 A7 A8 A15 A Fines Figure 30. Average pore size and total pore area for the fines A correlation between the total pore area and the BET surface area in the case of fines of the gravel products is also to be expected, because the molecular area of the nitrogen gas (16.2 Å 2 ) is much smaller than the smallest pore size diameter (0.006 µm = 60 Å) as measured by means of mercury intrusion (figure 31). The rock products are not weathered, and so there is no statistically significance correlation between the total pore area and the BET value (figure 31). Total pore area [m 2 /g] BET (A) vs total pore area R = BET value [m 2 /g] Total pore area [m 2 /g] BET (B) vs total pore area R = BET value [m 2 /g] Figure 31. Correlations between BET value and total pore area, fines

92 81 For powders, i.e. fines, the mercury intrusion method measures not only the particle porosity but also the pore size distribution and space/voids between the particles (see 4.4.5). The quantity of voids is dependent on the packing degree of the powder. The main factors affecting the packing degree of the powders are the grading curve, particle shape and surface adhesion (CUMBERLAND 1987, RUMPF 1990). The C u value describes the shape of the grading curve and the linear correlation between the average pore sizes and the C u value is 0.91 for the rock products and 0.89 for the gravel products. Semi-coarse fractions The highest total pore areas were m 2 /g (B1) and m 2 /g (A7) for the semicoarse fractions of the gravel and rock products respectively (figure 32). The lowest value for the rock products was m 2 /g (A8 and A10) and for the gravel products m 2 /g (B14). The highest average pore sizes were µm (A8) and 9.08 µm (B14) for the rock and gravel products respectively. The lowest values were 0.36 µm (B1) for the gravel product was and 2.94 µm (A7) for the rock product. Avg. pore size [um] Total pore area and average pore size Average pore size Total pore area Tot. pore area [m 2 /g] 0 A2 A3 A7 A8 A10 A16 B1 B2 B7 B8 B12 B13 B14 B15 REF 0.00 Semi-coarse aggregate Figure 32. Total pore area and average pore size for the semi-coarse aggregates

93 82 The total pore area of the semi-coarse fractions is divided into three categories: pore sizes > 0.09 µm, µm and µm. As can be seen in figure 33, only the gravel products have pores in the smallest category. Incremental pore area [m 2 /g] >0.09 µm µm µm Incremental pore area A2 A3 A7 A8 A10 A16 B1 B2 B7 B8 B11 B12 B13 B14 REF Semi-coarse aggregate Figure 33. Incremental pore areas for the semi-coarse aggregates What is noteworthy is that those gravel products that are partly crushed do not contain the smallest pores and only rock products that have been stored in outdoor stockpiles for years have the middle-category pores. It is quite likely that for the gravel products, the crushing process has partly shaken off the most weathered layer, and//or that this layer has partly flaked off during the crushing. The particle porosity origin from the weathered layers exist in the aggregate product even after the crushing, however, in smaller particle sizes. The particle porosity of the B8 product (sandstone) results both from weathering phenomena and from the sedimentation structure. The practical influence of aggregate particle porosity on concrete is water absorption. Normal unweathered aggregate has an absorption capacity of % and moderately weathered aggregate can easily have an absorption capacity of % or even higher. For 1 m 3 of concrete with 1750 kg of aggregate, the difference with water absorption

94 83 capacity of 0.7 % (0.3 % 1.0 %) affects 12 litres/m 3 of water when the mix design is made on a bone-dry basis. If saturated and surface dry basis is used, the influence on workability is thus insignificant, though the effect on the drying shrinkage still exists (see chapter 6.5). 5.6 Zeta potential The zeta potential measurements were made three times for samples without admixture (N) and additionally three times for samples with superplasticizer (WR). The average coefficient of variation for the zeta potential measurements was 5.5± 4.0%. The function of the superplasticizers in concrete is based on adsorption by the cement particles and the induced effect on the electrical double layer, electrostatic repulsion and enhanced dispersion of the particles (RAMACHANDRAN 1981, RUMPF 1990, TATTERSALL ET AL. 1983). The same phenomena are also involved in the way which the superplasticizer affects the aggregate fines. The higher the absolute value of the zeta potential, the greater is the particle dispersion. Figure 34 presents the zeta potential values measured for each aggregate without (N) and with the ionic surfactant, i.e. superplasticizer Mighty 150. The zeta potential measurements performed with the airentraining agent showed no difference in values when compared to N (=no admixture) levels. Hence, it was concluded that the air-entraining agent develops a lipid layer or equivalent on the aggregate surface and thus, does not affect the surface electrical stability. The range for the fines of the gravel products varied from 9.1 mv (B2) to 2.8 mv (B10) without the superplasticizer and from 18.3 mv (B14) to 5.7 mv (B10) with the superplasticizer. For the fines of the rock products the values were between 11.2 mv (A2) and 5.0 mv (A6) without the surfactant and 17.1 mv (A3) and 12.2 mv (A15) with the surfactant.

95 84 Zeta potential Zeta potential [mv] WR N & AE 2 0 B1 B2 B3 B6 B7 B8 B9 B10 B11 B12 B13 B14 A2 A3 A6 A7 A8 A15 A16 Fines Figure 34. Zeta potential values for each fines without and with superplasticizer As can be seen, the addition of the superplasticizer causes a level change in the zeta potential value. The regression equations for the level change are as follows: Zeta change N WR (A) = - 746F-LD 73F- avg. pore size +151F-BET Equation 13 Zeta change N WR (B) = - 142F-LD 49F-avg. pore size Equation 14 According to the equations, the change in the zeta potential value is greater when the average pore size increases, i.e. when the total pore area is lower. Additionally, the increase in total fineness also increases the zeta potential change. The equation for rock products also includes the effect of the BET value, i.e. surface area caused by shape properties. The greater the change in the zeta potential, thus lower the consumption of superplasticizer for adequate dispersion of the fines. The correlation between the measured and calculated values is 0.90 if all the A and B aggregates are included. If the B9 is omitted the correlation rises to 0.96 (figure 35 and table 16).

96 85 Table 16. The measured and calculated values of the percentile change in zeta potential of the fines Measured change [%] Calculated change [%] B B B B B B B B B B B B A A A A A A A Callc. Zeta pot. change [%] Zeta potential Change N->WR vs. Calculated Change N->WR without B R = 0.96 Zeta pot. change [%] Figure 35. The correlation of the measured and calculated zeta potential change-% 5.7 Resistance to fragmentation The Los Angeles values of the semi-coarse fractions (SCF) are results of one determination. According to the EN standard the reproducibility of the test is R = 0.17X (X represents the Los Angeles value). Figure 36 presents the results of the Los Angeles test done on the SCF. As can be seen from the figure, the gravel products (except the B8) had fairly constant values: between 21.5% (B2) and 29.4% (B12); the B8 had the value of 32.5%. The rock type of the B8 aggregate is sandstone, while all the other gravel products are of granitic origin. By contrast, rock products showed a large scatter in the results. The A7 had the lowest value, 19.6%, and the A8 had the highest value, 38.2%.

97 86 Los Angeles -test (mod.) B1 B2 B7 B8 B11 B12 Los Angeles value [%] B13 B14 REF A2 A3 A7 A8 A10 A16 Semi-coarse aggregate Figure 36. Results of the Los Angeles test for the semi-coarse aggregates As the tested samples of the aggregate products were not modified according to shape (e.g. bar sieved), the results show the true resistance to fragmentation and hence, also best describe performance in concrete castings. 5.8 Elongation, flakiness, particle volume and quantity The number of aggregate particles scanned in one image varied according to the particle size. For the size 3.15/4.0 mm the number varied between and for the 1.6/2.0 mm and 0.8/1.0 mm sizes between and respectively. The coefficient of measurement variation, as calculated from area determination, between images of two sets of particles from same aggregate product was < 5 % in each size fraction. The accuracy of the calculated particle quantity in the mix design is additionally influenced by the simplification of the particle shape. All the particles were assumed to

98 87 be spheres. This is not the case, but it allows us to perceive the quantity differences caused by the shape properties, i.e. average volume scatter. Figure 37 presents the elongation and flakiness for all three size fractions: 3.15/4.0 mm, 1.6/2.0 mm and 0.8/1.0 mm. Though the values are dimensionless, the information that they give is the extent to which the dimensions of the particles deviate from the dimensions of a sphere. The rock products are more elongated and flaky than the gravel products. The figures show clearly that the particles of smaller size are more elongated than those of larger sizes. For the rock products the tendency is seen in the case of all three size fractions, but with the gravel products the elongation values for the 1.6/2.0 mm and 3.15/4.0 mm particles are quite close to each other. The B7, B12 and B13 aggregates are party crushed gravel products and hence have increased elongation and/or flakiness values. In contrast to elongation, the relative flakiness levels are the same for all size fractions / / /4.0 Elongation / / /4.0 Flakiness Elongation Flakiness B1 B2 B7 B8 B11 B12 B13 B14 REF A2 A3 A7 Semi-coarse aggregate A8 A10 A B1 B2 B7 B8 B11 B12 B13 B14 REF A2 A3 A7 Semi-coarse aggregate A8 A10 A16 Figure 37. Elongation and flakiness values for the three size fractions Figure 38 shows scanned images of the A7 and REF aggregates. The figure makes visible the elongation variations in the size fraction 0.8/1.0 mm. The A7 aggregate has the highest value, 1.64, and the REF aggregate has the lowest value, 1.45.

99 88 A7 0.8/1.0 mm Figure 38. REF 0.8/1.0 mm Scanned images of the A7 and REF 0.8/1.0 size fractions The impression one gets from the images is that the A7 particles are bigger than the REF particles. The 2D area is, however, greater for the A7 than for the REF particles as the 1.0 mm sieve, which has been used for preparation of the narrow size fraction, allows elongated particles to pass but not those with a width greater than 1.0 mm. Due to the elongation, the density of the A7 particles (pieces/cm 2 ) is smaller in the image, and this also strengthens the impression given. The particle volume calculations have been made partly from a data source different from that for the elongation and flakiness values, though the particles used for all the measurements were the same. As can be seen from figures 39 and 37, the particle volume has the lowest values when the elongation is low and the flakiness high. The following regression equations can be calculated for the three size fractions: Volume (3.15/4.0 mm) = 76E - 35F 18(E*F) Equation 15 Volume (1.6/2.0 mm) = 10.1E - 5.1F - 2.1(E*F) Equation 16 Volume (0.8/1.0 mm) = 1.29E F (E*F) Equation 17 where E = elongation value and F = flakiness value

100 89 Particle volume B1 B2 B7 B8 B11 B12 B13 B14 REF A2 A3 A7 A8 A10 A16 Volume 0.8/1.0 [mm 3 ] 0.8/1.0 (1.6/2.0)*6 3.15/ Volume 3.15/4.0 + (1.6/2.0)*6 [mm 3 ] 17 Semi-coarse aggregate Figure 39. Particle volume for the three size fractions The equations are well in line with each other, when the theoretical volume ratios between the size fractions are taken into account. The 3.15/4.0 mm particles have a volume 2 3 * 2 3 times larger than the 0.8/1.0 mm particles and 2 3 times larger than the 1.6/2.0 mm particles. The correlation between all three equations and the measured particle volumes is The quantity data is calculated from the volume data and the quantities are dependent on mix design, i.e. aggregate volume amount. Table 17 shows the quantities for all aggregate products with the N35 mix design. The quantity difference between the most sphere and least sphere aggregate products, B1 and A8 respectively, is over 20% in each size fraction. What is also important to notice is that B7, which is a gravel product containing some crushed gravel, has more particles than two totally crushed rock products, A2 and A3. We can thus say that, at least when it comes to paste consumption due to shape properties, unprofessional production and blending of partly crushed gravel with uncrushed gravel can spoil a good raw material.

101 90 Table 17. Particle quantities for the N35 mix design QNTY QNTY QNTY 3.15/4.0 mm 1.6/2.0 mm 0.8/1.0 mm B B B B B B B B REF A A A A A A Angularity and surface texture The angularity and surface texture determinations for the SCF are based on the same scanned images used in the case of elongation and flakiness. The coefficient of measurement variation was < 4 % for surface texture and <5% for angularity. The surface texture and angularity values are presented in figure 40. As can be seen, the angularity values are higher for the rock products than for the gravel products. Only the partly crushed gravel products B7 and B12 have somewhat higher values than the average gravel products. The highest and lowest angularity values for the gravel and rock products are 4.9 (B2) (B12) and 7.1 (A2) 8.7 (A10) respectively (figure 41).

102 91 Surface texture and Angularity Surface texture Surface texture Angularity Angularity 0.6 B1 B2 B7 B8 B11 B12 B13 B14 REF A2 A3 A7 A8 A10 A Semi-coarse aggregate Figure 40. Surface texture and angularity values of the semi-coarse aggregates For the surface texture, there is no gap between the values of the crushed and uncrushed products. Those gravel products which are either crushed or weathered (high particle surface porosity) have higher surface texture values than the other gravel products. Among the rock products, the mineral size and mineralogical composition have the main effect on the surface texture. The A2 and A3 products are rich with plagioclase, feldspar and quartz, and their mineral size is fairly large, approximately 3 mm. The A8 and A10 products also have a large mineral size, approximately 3 mm, while the A7 and A16 have a mineral size smaller than 1mm. The highest and lowest surface texture values for the gravel and rock products are 0.97 (REF) (B7) and 1.01 (A3) 1.80 (A16) respectively (figure 42).

103 92 Figure 41. Scanned images of B2 and A10 with low and high angularity, SCF A2 and A16 Figure 42. Scanned images of A2, A16, B14 and B1 representing different surface texture values, SCF B14 and B1

104 93 The term surface texture describes the unevenness of the particle surface. In the case of the uncrushed gravel products, the surface texture actually characterises how much dissolving and/or disintegration of minerals has occurred; for the rock products, it describes the origin of surface roughness on the basis of the mineral and mineralogical properties, e.g. mineral size and crystal form. The correlation between the total pore area and surface texture is R = 0.98 for the gravel products (figure 43). For this correlation, the gravel products B7 and B12 are excluded, because they contain an undetermined amount of crushed gravel. For rock products the correlation is R = 0.72, and the correlation covering both the gravel and rock products is R = Total pore area [m 2 /g] Surface texture (B) vs. total pore area B7 and B12 excluded R = Surface texture Figure 43. Correlation between surface texture and total pore area for the gravel products 5.10 Discussion of the test results for aggregates The fine aggregate products produced from natural aggregate can be divided mainly into four categories as follows (the three first were studied in this work): Uncrushed gravel products Crushed rock products Partly crushed gravel products Mixture of uncrushed gravel and rock products

105 94 The test results indicate that some phenomena are more likely to be associated with gravel products and others with rock products. The partly crushed gravel products and the mixture of uncrushed gravel and rock products are combinations of the first two product types. Table 18 presents a general overview of the potential connection between product type and the quality characteristics. Table 18. General overview of the potential association between product type and quality characteristics SEMI-COARSE FRACTIONS FINES SHAPE WEATHERING STRENGTH DENSITY WEATHERING FINENESS UNCRUSHED GRAVEL CRUSHED ROCK PARTLY CRUSHED GRAVEL MIXTURE OF UNCRUSHED GRAVEL AND ROCK Semi coarse fractions X X X X X X X X X X X X X X X X X Shape: elongation, flakiness, particle volume & quantity and angularity Weathering: total pore area, incremental pore area, average pore size Strength: resistance to fragmentation (Los Angeles value) Density: particle density Fines Weathering: BET value, total pore area, average pore size Fineness: LD value, C u, 8 µm passing-%, fineness The rock products are always more elongated, flakier and more angular than the gravel products. With professional production, the differences in the elongation, flakiness and, to some extend, also the angularity between the rock and uncrushed gravel products can be minimised and controlled. However, each rock material has its own characteristic tendency for good or poor shape properties.

106 95 Weathering of the gravel is caused by a long period of physical and chemical strain. The degree of weathering can vary substantially from one gravel product to another. In the contrast, rock is not normally weathered, excluding the very surface of rock formations. Particle porosity by mercury intrusion has been used as a measure of the existing degree of weathering, but it should not be considered an evaluation of the future durability of the aggregate products. The variation in the resistance to fragmentation in the granitic gravel was found to be quite small. One of the major reasons for this is the glacial period, which was so harsh that only materials of a relatively good strength survived the abrasion effect. Additionally, the glacial period mixed the different rock types together so extensively that the variation in strength, as well as the particle density variation in the gravel, is fairy small. In contrast, the variation in the rock products for both the strength and particle density can be very large. Weathering of the fines can be also possible for the crushed rock products, if the rock has many crack joints where weathering has occurred and if this material then ends up in the crushed rock products. The dispersion effect of specific superplasticizer dosage varies appreciably with both the gravel and rock products. The dominating characteristics for the zeta potential value are the average pore size and the calculated surface area, and for rock products also the shape of the fines particles, as this influences the surface area.

107 96 6. CONCRETE TEST RESULTS AND DISCUSSION 6.1 Workability The workability of the castings was measured by means of the flow and slump methods. At the time when the slump value is measured, the concrete is not moving. Thus, it can be expected that the effective average shear rate will be zero and that the slump value will correlate only with yield value. In the case of the flow method, the result is to some extent influenced by the plastic viscosity, though the correlation with the slump value is reported to be 0.92 (TATTERSALL 1983). In this work, the correlation between the slump and flow values varied between 0.93 and 0.99 with the N, AE and WR30 castings (figure 44). Slump vs. Flow Flow [mm] y = 2.5x + 80 R = 0.95 y = x R = 0.96 N30 N35 WR30 WR35 AE30 AE Slump [mm] Figure 44. Correlation between the slump and flow values The linear correlation equation varied between Flow value (N30,AE30,WR30) = [ ] * Slump value + [ ] mm Eq 18 Flow value (N35, AE35) = [ ] * Slump value + [ ] mm Eq 19

108 97 Many of the castings made with the WR35 mix design behaved differently, -more viscously - partly due to segregation and partly due to excess paste. Consequently, while the correlation equation differs significantly (Eq 15), the correlation between the slump and flow values, 0.95, is good. Flow value (WR35) = 2.5 * Slump value + 80 mm Eq 20 The repetitions of the castings showed that accuracy is better in the case of flow value than in that of slump. The repeatability errors for the slump and flow values were 12 (10 16) mm and 15 (13 19) mm respectively. The evaluation of the data was therefore conducted using the flow results. Table 19 presents the flow value repeatability errors and group average for N, WR and AE mixes, and figure 45 shows the average, minimum, maximum and standard deviation values for each mix design. Table 19. Flow repeatability error and group average median of 90% quantile (10%-90% quantiles for the median) N WR AE All Repeatability error[mm] 21 (16-31) 9 (6-13) 10 (6-19) 15 (13-19) Group average [mm] 70 (62-89) 128 ( ) 61 (48-72) 104 (97-119) The repeatability error is smallest for the WR mix designs and highest for the N mix designs. The use of admixtures with normal dosages improves the cohesion of concrete and thus advances congruent behaviour in the workability repetitions. Nevertheless, in the case of the WR mix designs, the result scatter is the largest. The WR30 mix design in particular shows a large result range, mm, which demonstrates that small/moderate changes in aggregate parameters can have a strong influence on concrete. For each mix design group the result scatter is smaller for the high paste than the low paste mixes. When the repeatability errors and group averages are compared, it can be stated that for the N mix designs less than 30%, for the AE mix designs less than 16% and for the WR mix designs less than 7% of the difference can be explained by the

109 98 repeatability error. When all the mix designs are taken into account, the repeatability error explains 14% of the result scatter. Flow 5min = Stdev Flow [mm] N30 N35 WR30 WR35 AE30 AE35 All avg [mm] max min Figure 45. Flow value statistics for each mix design Figure 46 presents measured flow table values for six WR30 and WR35 castings WR30 WR35 Flow value, 5 min Flow value [mm] B1/B1 B3/B1 B1/REF B3/REF B3/A8 B13/B13 Fines/Semi-coarse combination Figure 46. Measured flow table value for different combinations of fines and SCF (The repeatability error is indicated by bars on the columns)

110 99 Figure 46 makes it clear that the characteristics of both the fines and the SCF affect the flow value. The main aggregate characteristics in the castings are shown in table 20. Table 20. Main aggregate characteristics for the WR castings presented in figure 46 Aggregate Fines Semi-coarse fraction B1 High surface area High particle porosity B3 REF A8 Low surface area Good shape Low particle porosity Good shape Low particle porosity Poor shape B13 Medium surface area Medium particle porosity Fair shape Shape of the semi-coarse fraction The effect of poor semi-coarse fraction shape prevails when the amount of paste is low, but when enough paste is available, the shape characteristics can be mostly overcome, as can be seen from the B3/A8 castings. The influence of the shape characteristic (flakiness 3.15/4.0 mm) on the workability of the WR 30 and WR35 castings can also be seen from figure 47. The difference between the mix designs WR30 and WR35 consists of 45 litres/m 3 of paste, including 29 litres of water and 50 kg of cement. Flow (WR30) vs. Flakiness 3.15/4.0 mm Flow (WR35) vs. Flakiness 3.15/4.0 mm Flow value [mm] R = 0.73 Flow value [mm] R = Flakiness 3.15/4.0 mm Flakiness 3.15/4.0 mm Figure 47. Correlations between flow value and flakiness for the WR castings

111 100 Particle porosity of the semi-coarse fraction and surface area of the fines Both the particle porosity of the semi-coarse fraction and surface area of the fines have a clear effect on workability, mainly through water absorption. As one workability class is 60 mm in flow value (EN206), thus the effects of the fines surface area and SCF particle porosity can be calculated into changes in flow classes (table 21). Table 21. Changes in flow classes due to the surface area of the fines and pore area of the SCF; WR30 and WR35 mix designs Effect of fines surface area when SC particle porosity is high Effect of fines surface area when SC particle porosity is low Effect of SC particle porosity when fines surface area is high Effect of SC particle porostiy when fines surface area is low Effect of both SC p.porosity and fines surface area when both change from high to low WR30 Flow class change ( = 80 mm) 1.3 classes ( = 120 mm) 2.0 classes ( = 140 mm) 2.3 classes ( = 165 mm) 2.8 classes ( = 270 mm) 4.5 classes WR35 Flow class change ( = 50 mm) 0.8 classes ( = 45 mm) 0.8 classes ( = 145 mm) 2.4 classes ( = 135 mm) 2.3 classes ( = 190 mm) 3.2 classes The separate effects of the fines and SCF can best be evaluated using the WR35 castings, as the amount of paste and water are higher than those of the WR30 castings, and thus, the absorption caused by the fines and SCF does not lead to friction between aggregate particles and to decreased workability. The particle porosity of the SCF has a greater effect on the workability than does the surface area of the fines. As can be seen in table 21, the effect of the particle porosity of the SCF on the WR35 flow values is three times greater than that of the surface area of the fines. The cumulative effect of the fines and SCF is more than three flow classes, if they both change from low to high. On the other hand, because of the low surface area of the fines,

112 101 the cohesion of the B3/REF concrete was so low that the concrete became segregated and thus would not be suitable for concrete production. Medium values for surface area (fines), particle porosity and shape (SCF) The SCF B13 is a partly crushed gravel product and thus has a somewhat worse shape than the B1 and REF products. However, the shape is far better than it is for the A8. Additionally, the particle porosity of the SCF B13 falls between the A8-REF and B1. Furthermore, the surface area of the B13 fines also lies between those of the B1 and B3. The results for the B13/B13 castings are in line with the aggregate characteristics. Hence, the B13/B13 castings equal the B3/B1 and B1/REF castings respectively, thus representing average aggregate quality. 6.2 Air % Air %, fresh concrete The air in the concrete can be divided into two groups: intentionally entrained air (AE mix designs) and air entrapped because of unsuccessful compaction/low degree of (N and WR mix designs). The concrete mix designs were calculated inclusive of the air. For the AE mix designs the target value was 5.0%, and for the N and WR mix designs it was 1.0%. For the N35 and WR35 mixes the average values are close the target value; deviations + 0.1%-unit and 0.3%-unit respectively. However, for the N30 and WR30 mix designs the deviations are greater than the repeatability error (figure 48 and table 22). The AE mixes have a large variation in the air % values. For the AE30 mixes, even the average value is below the target value and the difference between the minimum and maximum values is more than 200%, i.e. 2.1% and 6.6%. For the AE35 mixes the average is within the target range, but the difference between the minimum and maximum values is more than 100%, i.e. 3.0% and 7.5%.

113 102 Air %, fresh concrete Air % =Stdev N30 N35 WR30 WR35 AE30 AE35 All avg [%] max min Figure 48. Air %, fresh concrete statistics for each mix design As the mix design calculations were performed using the target air % values it was therefore to be expected that the actual densities of the castings would deviate according to the difference between the actual and target air %. Figure 49 shows the excess density percentage for the fresh concrete and additionally the excess densities of the hardened concrete at the ages 24 h, 28d and 91d. The upper set of data is the for the AE mix designs and the lower set is for the N and WR mix designs. As can be seen from figure 49, the excess density of the fresh concrete accords quite well with the calculated density of the concrete inclusive the 1.0% and 5.0% of air. The deviation, which is approximately 0.5%-unit, is caused by the water absorption of the aggregate. Additionally, figure 49 demonstrates how the density of the concrete increases along the degree of hydration (see 24h cubes vs. 91d cubes). The difference between the densities of the 28d and 91d cubes is within the measurement accuracy. What is

114 103 noteworthy is that the excess density % scatter for the AE mix designs is much greater in the case of cubes than fresh concrete. This indirectly indicates that the air structure has not been stable in all the mixes. Air % of concrete Air % of concrete Excess density % vs. air %, fresh concrete Excess density [%] Excess density% vs. air %, 28d cubes Excess density [%] Air % of concrete Air % of concrete Excess density% vs. air %, 24 h cubes Excess density [%] Excess density% vs. air%, 91d cubes Excess density [%] Figure 49. Excess density % of fresh concrete and 24h, 28d and 91d cubes For the N mix designs the scatter of the results and the measurement accuracy are such that the expectation values of the group average model and repeatability error are virtually the same. For the WR mix designs the group average is 8 times greater than the repeatability error, and for the AE mix designs the repeatability error represents one third of the group average value (table 22).

115 104 Table 22. Air % repeatability error and group average median of 90% quantile (10%-90% quantiles for the median) N WR AE All Repeatability error[%] 0.3 ( ) 0.1 ( ) 0.4 ( ) 0.3 ( ) Group Average [%] 0.4 ( ) 0.8 ( ) 1.5 ( ) 0.9 ( ) For the N and WR mix designs, a strong inverse correlation, R = -0.91, between the entrapped air and flow value can be detected, i.e. when the workability is poor, the normal compaction energy is not sufficient to force the air out of the concrete (figure 50). The reverse phenomenon applies to the AE mix designs, R = 0.89 and R = 0.52 for the AE30 and AE35 respectively. When the workability is good, the entrained air % also tends to be higher (figure 50). Air % (N & WR) vs. Flow 5 min Air % (AE) vs. Flow 5 min Flow [mm] R = Flow [mm] AE30 AE35 R = 0.52 R = Air %, fresh concrete Air %, fresh concrete Figure 50. Correlation between entrapped and entrained air % and flow values As the correlations are so strong, we can conclude that fine aggregate characteristics that affect the workability also affect the entrapped air %. When the amount of paste is low, the entrained air % is also influenced by aggregate characteristics that affect the workability. With higher amounts of paste, other mix design and/or aggregate characteristics begin to compensate for the influence.

116 Air %, hardened concrete Thin sections of all the AE castings were made for air analysis. The factors determined were: the entrained and entrapped air %, the specific surface of the entrained air voids, and the spacing factor, i.e. the thickness of the hardened cement paste between adjacent air voids. Tables 23 and 24 present the thin-section statistics for the AE30 and AE35 mix designs. Additionally, for comparison purposes, the tables show the air % statistics, with the volumetric method from the fresh concrete. Table 23. Thin-section statistics from the AE30 mix design Entrained air Entrapped air Total air [%] Specific surface of Spacing Volumetric AE30 AE30 [%] [%] voids [mm2] Factor method avg avg 3.9 stdev stdev 0.9 max max 6.6 min min 2.1 Total air [%] Table 24. Thin-section statistics from the AE35 mix design Entrained air Entrapped air Total air [%] Specific surface of Spacing Volumetric AE35 [%] [%] voids [mm2] Factor method avg avg 5.4 stdev stdev 0.9 max max 7.5 min min 3.0 AE35 Total air [%] As can be seen from tables 23 and 24, the AE30 castings are more homogenous with the air quality. Furthermore, the deviations between the volumetric air % and thin-section air % are smaller for the AE30 than for the AE35 castings. These results reveal that the repeatability of the thin-section is two sided. The results were relatively constant if the air void system was stable and if not, they could deviate strongly from each other (table 25 and figures 51). The repeatability error for the volumetric method is covered in chapter 6.2.

117 106 Table 25. Examples of thin-section analysis results from repeated castings Recipe # Entrained air [%] Entrapped air [%] Total air [%] Specific surface of Spacing voids [mm2] Factor AE35: B9 (50%)+ A3 (50%) / B7 (50%) + A2 (50%) AE35: B7 (100%) / B7 (100%) Casting # 64 Casting # 58 Figure 51. Pictures of repetition castings of the same mix with unstable void structure. The castings had very similar fresh concrete values; air 6.0% and 5.6%, flow value 440 mm and 435 mm, bleeding 60 min 1.0 g/cm 3 and 1.1 g/cm 3 for castings #58 and #64 respectively.

Geology 229 Engineering Geology. Lecture 7. Rocks and Concrete as Engineering Material (West, Ch. 6)

Geology 229 Engineering Geology. Lecture 7. Rocks and Concrete as Engineering Material (West, Ch. 6) Geology 229 Engineering Geology Lecture 7 Rocks and Concrete as Engineering Material (West, Ch. 6) Outline of this Lecture 1. Rock mass properties Weakness planes control rock mass strength; Rock textures;

More information

Concrete Technology Prof. B. Bhattacharjee Department of Civil Engineering Indian Institute of Science IIT Delhi. Lecture - 6 Aggregates (Size, Shape)

Concrete Technology Prof. B. Bhattacharjee Department of Civil Engineering Indian Institute of Science IIT Delhi. Lecture - 6 Aggregates (Size, Shape) Concrete Technology Prof. B. Bhattacharjee Department of Civil Engineering Indian Institute of Science IIT Delhi Lecture - 6 Aggregates (Size, Shape) Welcome to concrete technology module 2. Module 2 deals

More information

TESTING of AGGREGATES for CONCRETE

TESTING of AGGREGATES for CONCRETE TESTING of AGGREGATES for CONCRETE The properties of the aggregates affect both the fresh and hardened properties of concrete. It is crucial to know the properties of the aggregates to be used in the making

More information

Chapter 2. The Ideal Aggregate. Aggregates

Chapter 2. The Ideal Aggregate. Aggregates Chapter 2 Aggregates The Ideal Aggregate Strong and resists loads applied Chemically inert so it is not broken down by reactions with substances it comes in contact with Has a stable volume so that it

More information

Aggregates for Concrete

Aggregates for Concrete Fine Aggregate Sand and/or crushed stone < 5 mm (0.2 in.) F.A. content usually 35% to 45% by mass or volume of total aggregate Coarse Aggregate Gravel and crushed stone 5 mm (0.2 in.) typically between

More information

AIR BUBBLE STABILITY MECHANISM OF AIR-ENTRAINING ADMIXTURES AND AIR VOID ANALYSIS OF HARDENED CONCRETE

AIR BUBBLE STABILITY MECHANISM OF AIR-ENTRAINING ADMIXTURES AND AIR VOID ANALYSIS OF HARDENED CONCRETE AIR BUBBLE STABILITY MECHANISM OF AIR-ENTRAINING ADMIXTURES AND AIR VOID ANALYSIS OF HARDENED CONCRETE Bei Ding, Jiaping Liu, Jianzhong Liu Jiangsu Academy of Building Science Co., Ltd, Nanjing, China

More information

Copyright SOIL STRUCTURE and CLAY MINERALS

Copyright SOIL STRUCTURE and CLAY MINERALS SOIL STRUCTURE and CLAY MINERALS Soil Structure Structure of a soil may be defined as the mode of arrangement of soil grains relative to each other and the forces acting between them to hold them in their

More information

Geotechnical Engineering I CE 341

Geotechnical Engineering I CE 341 Geotechnical Engineering I CE 341 What do we learn in this course? Introduction to Geotechnical Engineering (1) Formation, Soil Composition, Type and Identification of Soils (2) Soil Structure and Fabric

More information

Chapter. Materials. 1.1 Notations Used in This Chapter

Chapter. Materials. 1.1 Notations Used in This Chapter Chapter 1 Materials 1.1 Notations Used in This Chapter A Area of concrete cross-section C s Constant depending on the type of curing C t Creep coefficient (C t = ε sp /ε i ) C u Ultimate creep coefficient

More information

USE OF DUNE SAND AS AN ALTERNATIVE FINE AGGREGATE IN CONCRETE AND MORTAR. Department of civil Engineering, The Open University Sri Lanka

USE OF DUNE SAND AS AN ALTERNATIVE FINE AGGREGATE IN CONCRETE AND MORTAR. Department of civil Engineering, The Open University Sri Lanka USE OF DUNE SAND AS AN ALTERNATIVE FINE AGGREGATE IN CONCRETE AND MORTAR R. Sanjeevan 1, S. Kavitha 2, T.C. Ekneligoda 3 and D.A.R. Dolage 4 1,2,3,4 Department of civil Engineering, The Open University

More information

CHEMICAL ADMIXTURES FOR CONCRETE

CHEMICAL ADMIXTURES FOR CONCRETE CHEMICAL ADMIXTURES FOR CONCRETE Definition: what are chemical admixtures? The definition of RILEM (International Union of Testing and Research Laboratories for Materials and Structures) is: Admixtures

More information

Engineering materials

Engineering materials 1 Engineering materials Lecture 9 Aggregates 2 Composition and structure Natural aggregates are derived from rocks. Classification: Igneous( 火成岩 ), sedimentary ( 沉積岩 ) and metamorphic ( 變質岩 ) Fine or coarse

More information

CEEN Laboratory 4 Aggregates for Base Layers, Portland Cement Concrete & Hot Mix Asphalt

CEEN Laboratory 4 Aggregates for Base Layers, Portland Cement Concrete & Hot Mix Asphalt CEEN 3320 - Laboratory 4 Aggregates for Base Layers, Portland Cement Concrete & Hot Mix Asphalt INTRODUCTION Civil Engineering projects utilize aggregates for a variety of purposes, including unbound base

More information

CIVE 2700: Civil Engineering Materials Fall Lab 2: Concrete. Ayebabomo Dambo

CIVE 2700: Civil Engineering Materials Fall Lab 2: Concrete. Ayebabomo Dambo CIVE 2700: Civil Engineering Materials Fall 2017 Lab 2: Concrete Ayebabomo Dambo Lab Date: 7th November, 2017 CARLETON UNIVERSITY ABSTRACT Concrete is a versatile construction material used in bridges,

More information

Asphalt Mix Designer. Module 2 Physical Properties of Aggregate. Specification Year: July Release 4, July

Asphalt Mix Designer. Module 2 Physical Properties of Aggregate. Specification Year: July Release 4, July Specification Year: July 2005 Release 4, July 2005 2-1 The first step in the development of an HMA mix design is to identify the materials that will be used in the pavement. In Florida the asphalt binder

More information

Prof. B V S Viswanadham, Department of Civil Engineering, IIT Bombay

Prof. B V S Viswanadham, Department of Civil Engineering, IIT Bombay 05 Clay particle-water interaction & Index properties Electrical nature of clay particles a) Electrical charges i) The two faces of all platy particles have a negative charge. Resulting due to isomorphous

More information

Wikipedia.org BUILDING STONES. Chapter 4. Materials of Construction-Building Stones 1

Wikipedia.org BUILDING STONES. Chapter 4. Materials of Construction-Building Stones 1 Wikipedia.org BUILDING STONES Chapter 4 Materials of Construction-Building Stones 1 What is Stone? Stone is a concretion of mineral matter. Used either as a; Construction material, Manufacture of other

More information

Geotechnical Properties of Soil

Geotechnical Properties of Soil Geotechnical Properties of Soil 1 Soil Texture Particle size, shape and size distribution Coarse-textured (Gravel, Sand) Fine-textured (Silt, Clay) Visibility by the naked eye (0.05 mm is the approximate

More information

Soil Mechanics Brief Review. Presented by: Gary L. Seider, P.E.

Soil Mechanics Brief Review. Presented by: Gary L. Seider, P.E. Soil Mechanics Brief Review Presented by: Gary L. Seider, P.E. 1 BASIC ROCK TYPES Igneous Rock (e.g. granite, basalt) Rock formed in place by cooling from magma Generally very stiff/strong and often abrasive

More information

REGRESSION MODELING FOR STRENGTH AND TOUGHNESS EVALUATION OF HYBRID FIBRE REINFORCED CONCRETE

REGRESSION MODELING FOR STRENGTH AND TOUGHNESS EVALUATION OF HYBRID FIBRE REINFORCED CONCRETE REGRESSION MODELING FOR STRENGTH AND TOUGHNESS EVALUATION OF HYBRID FIBRE REINFORCED CONCRETE S. Eswari 1, P. N. Raghunath and S. Kothandaraman 1 1 Department of Civil Engineering, Pondicherry Engineering

More information

Construction aggregates : evaluation and specification Clive Mitchell Industrial Minerals Specialist

Construction aggregates : evaluation and specification Clive Mitchell Industrial Minerals Specialist Construction aggregates : evaluation and specification Clive Mitchell Industrial Minerals Specialist Outline of presentation Minerals at the British Geological Survey Particle size, shape & density Strength

More information

LECTURE NO. 4-5 INTRODUCTION ULTRASONIC * PULSE VELOCITY METHODS

LECTURE NO. 4-5 INTRODUCTION ULTRASONIC * PULSE VELOCITY METHODS LECTURE NO. 4-5 ULTRASONIC * PULSE VELOCITY METHODS Objectives: To introduce the UPV methods To briefly explain the theory of pulse propagation through concrete To explain equipments, procedures, calibrations,

More information

Module 5: Failure Criteria of Rock and Rock masses. Contents Hydrostatic compression Deviatoric compression

Module 5: Failure Criteria of Rock and Rock masses. Contents Hydrostatic compression Deviatoric compression FAILURE CRITERIA OF ROCK AND ROCK MASSES Contents 5.1 Failure in rocks 5.1.1 Hydrostatic compression 5.1.2 Deviatoric compression 5.1.3 Effect of confining pressure 5.2 Failure modes in rocks 5.3 Complete

More information

THE VALUE OF COLLOIDAL SILICA FOR ENHANCED DURABILITY IN HIGH FLUIDITY CEMENT BASED MIXES

THE VALUE OF COLLOIDAL SILICA FOR ENHANCED DURABILITY IN HIGH FLUIDITY CEMENT BASED MIXES THE VALUE OF COLLOIDAL SILICA FOR ENHANCED DURABILITY IN HIGH FLUIDITY CEMENT BASED MIXES Jansson, Inger (1), Skarp, Ulf (1) and Bigley, Carl (2) (1) Eka Chemicals AB, Sweden (2) Consultancy, New Zealand

More information

SOIL STRUCTURE AND FABRIC

SOIL STRUCTURE AND FABRIC SOIL STRUCTURE AND FABRIC The structure of a soil is taken to mean both the geometric arrangement of the particles or mineral grains as well as the interparticle forces which may act between them. Soil

More information

CHAPTER 3.3: METAMORPHIC ROCKS

CHAPTER 3.3: METAMORPHIC ROCKS CHAPTER 3.3: METAMORPHIC ROCKS Introduction Metamorphism - the process of changes in texture and mineralogy of pre-existing rock due to changes in temperature and/or pressure. Metamorphic means change

More information

Introduction to Soil Mechanics Geotechnical Engineering-II

Introduction to Soil Mechanics Geotechnical Engineering-II Introduction to Soil Mechanics Geotechnical Engineering-II ground SIVA Dr. Attaullah Shah 1 Soil Formation Soil derives from Latin word Solum having same meanings as our modern world. From Geologist point

More information

Soil Mechanics. Chapter # 1. Prepared By Mr. Ashok Kumar Lecturer in Civil Engineering Gpes Meham Rohtak INTRODUCTION TO SOIL MECHANICS AND ITS TYPES

Soil Mechanics. Chapter # 1. Prepared By Mr. Ashok Kumar Lecturer in Civil Engineering Gpes Meham Rohtak INTRODUCTION TO SOIL MECHANICS AND ITS TYPES Soil Mechanics Chapter # 1 INTRODUCTION TO SOIL MECHANICS AND ITS TYPES Prepared By Mr. Ashok Kumar Lecturer in Civil Engineering Gpes Meham Rohtak Chapter Outlines Introduction to Soil Mechanics, Soil

More information

The development of a new method for the proportioning of high-performance concrete mixtures

The development of a new method for the proportioning of high-performance concrete mixtures Cement & Concrete Composites 26 (2004) 901 907 www.elsevier.com/locate/cemconcomp The development of a new method for the proportioning of high-performance concrete mixtures Konstantin Sobolev Facultad

More information

NHBRA SOIL LABORATORY SECTION INTERIM REVISED TEST RATES FOR THE MATERIALS LABORATORY

NHBRA SOIL LABORATORY SECTION INTERIM REVISED TEST RATES FOR THE MATERIALS LABORATORY NHBRA SOIL LABORATORY SECTION INTERIM REVISED TEST RATES FOR THE MATERIALS LABORATORY DATE OF REVISION: ) 10th October 2010 S/N TEST DESCRIPTION UNIT RATE REMARKS 1 Sieve Analysis (wet and dry methods

More information

Aggregates. AAPA training

Aggregates. AAPA training Aggregates AAPA training Topics Aggregate sources and rock types Aggregate Production Aggregate Properties Coarse and fine aggregates in Asphalt Mixes Aggregates in Sprayed Seals Filler in asphalt mixes

More information

Lecture # 02 DEPARTMENT OF CIVIL ENGINEERING SWEDISH COLLEGE OF ENGINEERING & TECHNOLOGY, WAH CANTT. 14th December,

Lecture # 02 DEPARTMENT OF CIVIL ENGINEERING SWEDISH COLLEGE OF ENGINEERING & TECHNOLOGY, WAH CANTT. 14th December, Lecture # 02 DEPARTMENT OF CIVIL ENGINEERING SWEDISH COLLEGE OF ENGINEERING & TECHNOLOGY, WAH CANTT Instructor: Date: Engr. Imran Mehmood 14th December, 2011 1 SEDIMENTARY ROCKS SEDIMENTARY ROCKS The rocks

More information

Chapter -4 GRAIN SIZE PROPERTIES V_V

Chapter -4 GRAIN SIZE PROPERTIES V_V Chapter -4 GRAIN SIZE PROPERTIES Q V_V Chapter - 4 GRAIN SIZE PROPERTIES 4.1 Introduction The size of soil materials in a soil mass may range from the finest (colloidal size) to the coarsest (boulders).

More information

Your teacher will show you a sample or diagram of each, and show you a settling column. Draw these, and label your diagrams (8 pts) Ungraded:

Your teacher will show you a sample or diagram of each, and show you a settling column. Draw these, and label your diagrams (8 pts) Ungraded: From Sand to Stone: How do we recognize and interpret sedimentary rocks in the rock record? (Based closely on the University of Washington ESS 101 Lab 5: Sedimentary Rocks) Introduction: This lab consists

More information

Prof. B V S Viswanadham, Department of Civil Engineering, IIT Bombay

Prof. B V S Viswanadham, Department of Civil Engineering, IIT Bombay 06 Index properties Review Clay particle-water interaction Identification of clay minerals Sedimentation analysis Hydrometer analysis 0.995 20-40 Hydrometer is a device which is used to measure the specific

More information

Assistant Prof., Department of Civil Engineering Bhagwant University,Ajmer,Rajasthan,India ABSTRACT

Assistant Prof., Department of Civil Engineering Bhagwant University,Ajmer,Rajasthan,India ABSTRACT Study of Index Properties of the Soil 1 Mr Utkarsh Mathur 2 Mr Nitin Kumar 3 Mr Trimurti Narayan Pandey 4 Mr.Amit Choudhary 1 PG Scholar, Department of Civil Engineering Bhagwant University,Ajmer,Rajasthan,India

More information

What factors affect the angle of a slope?

What factors affect the angle of a slope? Climate Rock type and Structure What factors affect the angle of a slope? Aspect Fast mass movements Slides: Slides are movements along the SLIP PLANE, i.e. a line of weakness in the rock or soil structure.

More information

SOIL STRUCTURE AND FABRIC

SOIL STRUCTURE AND FABRIC SOIL STRUCTURE AND FABRIC The structure of a soil is taken to mean both the geometric arrangement of the particles or mineral grains as well as the interparticle forces which may act between them. Soil

More information

Non-Destructive Electrical Methods to Determine the Quality of Concrete

Non-Destructive Electrical Methods to Determine the Quality of Concrete Athens Journal of Technology & Engineering X Y Non-Destructive Electrical Methods to Determine the Quality of Concrete By Sreekanta Das William Clements Govinda Raju There is a great need to explore and

More information

Chapter 2 Concrete with Recycled Aggregates: Experimental Investigations

Chapter 2 Concrete with Recycled Aggregates: Experimental Investigations Chapter 2 Concrete with Recycled Aggregates: Experimental Investigations Carmine Lima, Marco Pepe, Ciro Faella and Enzo Martinelli Abstract The mechanical behaviour of Recycled Aggregate Concrete (RAC)

More information

Minerals. What are minerals and how do we classify them?

Minerals. What are minerals and how do we classify them? Minerals What are minerals and how do we classify them? 1 Minerals! Minerals are the ingredients needed to form the different types of rocks! Rock - is any naturally formed solid that is part of Earth

More information

The physical breakdown and chemical alteration of rocks and minerals at or near Earth s surface.

The physical breakdown and chemical alteration of rocks and minerals at or near Earth s surface. The physical breakdown and chemical alteration of rocks and minerals at or near Earth s surface. The material that is chemically and mechanically weathered to yield sediment and soil. Regolith consisting

More information

UNIT TOPICS TOPIC 1: MINERALS TOPIC 2: IGNEOUS ROCKS TOPIC 3: SEDIMENTARY ROCKS TOPIC 4: METAMORPHIC ROCKS TOPIC 5: THE ROCK CYCLE

UNIT TOPICS TOPIC 1: MINERALS TOPIC 2: IGNEOUS ROCKS TOPIC 3: SEDIMENTARY ROCKS TOPIC 4: METAMORPHIC ROCKS TOPIC 5: THE ROCK CYCLE UNIT TOPICS TOPIC 1: MINERALS TOPIC 2: IGNEOUS ROCKS TOPIC 3: SEDIMENTARY ROCKS TOPIC 4: METAMORPHIC ROCKS TOPIC 5: THE ROCK CYCLE TOPIC 1: MINERALS ESSENTIAL QUESTION: WHAT ARE MINERALS AND HOW DO WE

More information

Schedule of Accreditation issued by United Kingdom Accreditation Service 2 Pine Trees, Chertsey Lane, Staines-upon-Thames, TW18 3HR, UK

Schedule of Accreditation issued by United Kingdom Accreditation Service 2 Pine Trees, Chertsey Lane, Staines-upon-Thames, TW18 3HR, UK Unit 4 Heol Aur Dafen Industrial Estate Dafen Carmarthenshire SA14 8QN Contact: Mr P Evans Tel: +44 (0)1554 784040 Fax: +44 (0)1554 784041 E-Mail: pevans@gstl.co.uk Website: www.gstl.co.uk locations: Testing

More information

Effect of glass powder instead of mineral powder on asphalt mixture

Effect of glass powder instead of mineral powder on asphalt mixture 5th International Conference on Civil Engineering and Transportation (ICCET 2015) Effect of glass powder instead of mineral powder on asphalt mixture Binkai Tang 1, a *, Hengshan Wu1,b, Yiping Liao 1,c

More information

Influence of various acids on the physico mechanical properties of pozzolanic cement mortars

Influence of various acids on the physico mechanical properties of pozzolanic cement mortars Sādhanā Vol. 32, Part 6, December 2007, pp. 683 691. Printed in India Influence of various acids on the physico mechanical properties of pozzolanic cement mortars STÜRKEL, B FELEKOǦLU and S DULLUÇ Department

More information

Keywords: Zeolite powder, High-strength concrete, Plastic viscosity, Chloride-penetration resistance, Self-shrinkage.

Keywords: Zeolite powder, High-strength concrete, Plastic viscosity, Chloride-penetration resistance, Self-shrinkage. Advanced Materials Research Submitted: 2014-06-16 ISSN: 1662-8985, Vols. 1030-1032, pp 1003-1009 Accepted: 2014-07-10 doi:10.4028/www.scientific.net/amr.1030-1032.1003 Online: 2014-09-22 2014 Trans Tech

More information

GeoShanghai 2010 International Conference Paving Materials and Pavement Analysis

GeoShanghai 2010 International Conference Paving Materials and Pavement Analysis Particle Shape, Type and Amount of Fines, and Moisture Affecting Resilient Modulus Behavior of Unbound Aggregates Debakanta Mishra 1, Erol Tutumluer 2, M. ASCE, Yuanjie Xiao 3 1 Graduate Research Assistant,

More information

20/10/2015. Results: Part 1. Elucidation of the molecular architecture of the SPs

20/10/2015. Results: Part 1. Elucidation of the molecular architecture of the SPs Introduction Lime mortars used in the Built Heritage over centuries lime, usually air lime, as the binding material renders, repair mortars and other mixes. ROLE OF DIFFERENT SUPERPLASTICIZERS ON HYDRATED

More information

Studies on Furan Polymer Concrete

Studies on Furan Polymer Concrete Studies on Furan Polymer Concrete Rajesh Katiyar 1, Shobhit Shukla 2 1Associate Professor, Department of Chemical engineering, H.B.T.U., Kanpur-208002, India 2Research Scholar, Department of Chemical engineering

More information

Table of Contents. Foreword... xiii Introduction... xv

Table of Contents. Foreword... xiii Introduction... xv Foreword.... xiii Introduction.... xv Chapter 1. Controllability of Geotechnical Tests and their Relationship to the Instability of Soils... 1 Roberto NOVA 1.1. Introduction... 1 1.2. Load control... 2

More information

The Rheological and Mechanical Properties of Self-Compacting Concrete with High Calcium Fly Ash

The Rheological and Mechanical Properties of Self-Compacting Concrete with High Calcium Fly Ash The Rheological and Mechanical Properties of Self-Compacting Concrete with High Calcium Fly Ash Tomasz Ponikiewski 1, Jacek Gołaszewski 2* 1 Silesian University of Technology, Poland 2 Silesian University

More information

CONCRETE IN THE MIDDLE EAST

CONCRETE IN THE MIDDLE EAST CONCRETE IN THE MIDDLE EAST ALKALI REACTIVITY IN CONCRETE STRUCTURES Presented by : Eng. ELIE J. SFEIR INTRODUCTION What is the Alkali-Reactivity? The alkali reaction is a chemical reaction between some

More information

PHYSICO-MECHANICAL PROPERTIES OF ROCKS LECTURE 2. Contents

PHYSICO-MECHANICAL PROPERTIES OF ROCKS LECTURE 2. Contents PHYSICO-MECHANICAL PROPERTIES OF ROCKS LECTURE 2 Contents 2.1 Introduction 2.2 Rock coring and logging 2.3 Physico-mechanical properties 2.3.1 Physical Properties 2.3.1.1 Density, unit weight and specific

More information

Effect of Lime on the Compressibility Characteristics of a Highly Plastic Clay

Effect of Lime on the Compressibility Characteristics of a Highly Plastic Clay Effect of Lime on the Compressibility Characteristics of a Highly Plastic Clay Abstract İnci Süt-Ünver Ph.D. Candidate Istanbul Technical University Istanbul - Turkey Musaffa Ayşen Lav Prof. Dr. Istanbul

More information

BIO & PHARMA ANALYTICAL TECHNIQUES. Chapter 5 Particle Size Analysis

BIO & PHARMA ANALYTICAL TECHNIQUES. Chapter 5 Particle Size Analysis BIO & PHARMA ANALYTICAL TECHNIQUES Chapter 5 by Dr Siti Umairah Mokhtar Faculty of Engineering Technology umairah@ump.edu.my Chapter Description Aims Discuss theory, principles and application of analytical

More information

Science and technology of concrete admixtures / edited by Pierre-Claude Aïtcin and Robert J. Flatt. Amsterdam [etc.], cop

Science and technology of concrete admixtures / edited by Pierre-Claude Aïtcin and Robert J. Flatt. Amsterdam [etc.], cop Science and technology of concrete admixtures / edited by Pierre-Claude Aïtcin and Robert J. Flatt. Amsterdam [etc.], cop. 2016 Spis treści About the contributors Woodhead Publishing Series in Civil and

More information

Soil structure Classification

Soil structure Classification Soil structure Classification Soil conditions and characteristics such as water movement, heat transfer, aeration, and porosity are much influenced by structure. In fact, the important physical changes

More information

INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING Volume 1, No 4, 2011

INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING Volume 1, No 4, 2011 Undrained response of mining sand with fines contents Thian S. Y, Lee C.Y Associate Professor, Department of Civil Engineering, Universiti Tenaga Nasional, Malaysia siawyin_thian@yahoo.com ABSTRACT This

More information

Rheological studies on the flow behavior of twophase solid-liquid materials

Rheological studies on the flow behavior of twophase solid-liquid materials Retrospective Theses and Dissertations 2008 Rheological studies on the flow behavior of twophase solid-liquid materials Gang Lu Iowa State University Follow this and additional works at: http://lib.dr.iastate.edu/rtd

More information

Tikrit University College of Engineering Civil engineering Department

Tikrit University College of Engineering Civil engineering Department Tikrit University SOIL CLASSIFICATION College of Engineering Civil engineering Department Soil Mechanics 3 rd Class Lecture notes Up Copyrights 2016 Classification of soil is the separation of soil into

More information

Washing of Aggregates - Influence on Aggregate Properties and Mortar Rheology

Washing of Aggregates - Influence on Aggregate Properties and Mortar Rheology 1 Washing of Aggregates - Influence on Aggregate Properties and Mortar Rheology Mikael Westerholm Tech. Lic., M.Sc. (Chem. Eng) Swedish Cement and Concrete Research Institute, CBI SE-1 44 Stockholm, Sweden

More information

Impact of some parameters on rheological properties of cement paste in combination with PCE-based Plasticizers

Impact of some parameters on rheological properties of cement paste in combination with PCE-based Plasticizers Impact of some parameters on rheological properties of cement paste in combination with PCE-based Plasticizers Ameneh Schneider 1, Heinrich Bruckner 2 1 Smart Minerals GmbH, Vienna, Austria 2 Vienna University

More information

SIMULATING FRESH CONCRETE BEHAVIOUR ESTABLISHING A LINK BETWEEN THE BINGHAM MODEL AND PARAMETERS OF A DEM-BASED NUMERICAL MODEL

SIMULATING FRESH CONCRETE BEHAVIOUR ESTABLISHING A LINK BETWEEN THE BINGHAM MODEL AND PARAMETERS OF A DEM-BASED NUMERICAL MODEL International RILEM Conference on Material Science MATSCI, Aachen 2010 Vol. II, HetMat 211 SIMULATING FRESH CONCRETE BEHAVIOUR ESTABLISHING A LINK BETWEEN THE BINGHAM MODEL AND PARAMETERS OF A DEM-BASED

More information

Course Scheme -UCE501: SOIL MECHANICS L T P Cr

Course Scheme -UCE501: SOIL MECHANICS L T P Cr Course Scheme -UCE501: SOIL MECHANICS L T P Cr 3 1 2 4.5 Course Objective: To expose the students about the various index and engineering properties of soil. Introduction: Soil formation, various soil

More information

Prediction of torsion shear tests based on results from triaxial compression tests

Prediction of torsion shear tests based on results from triaxial compression tests Prediction of torsion shear tests based on results from triaxial compression tests P.L. Smith 1 and N. Jones *2 1 Catholic University of America, Washington, USA 2 Geo, Lyngby, Denmark * Corresponding

More information

Concrete Technology Prof. B. Bhattacharjee Department of Civil Engineering Indian Institute of Technology, Delhi

Concrete Technology Prof. B. Bhattacharjee Department of Civil Engineering Indian Institute of Technology, Delhi Concrete Technology Prof. B. Bhattacharjee Department of Civil Engineering Indian Institute of Technology, Delhi Lecture - 25 Strength of Concrete: Factors Affecting Test Results Welcome to module 6, lecture

More information

Lecture 13 Portland Cement Based Paste Systems

Lecture 13 Portland Cement Based Paste Systems Hydration, Porosity and Strength of Cementitious Materials Prof. Sudhir Mishra and Prof. K. V. Harish Department of Civil Engineering Indian Institute of Technology, Kanpur Lecture 13 Portland Cement Based

More information

SAND. By A S M Fahad Hossain Assistant Professor Department of Civil Engineering, AUST

SAND. By A S M Fahad Hossain Assistant Professor Department of Civil Engineering, AUST SAND By A S M Fahad Hossain Assistant Professor Department of Civil Engineering, AUST Definition Sand is a loose, fragmented, naturally-occurring material consisting of vary small particle (fine to medium

More information

Geology 229 Engineering Geology. Lecture 6. Basic Rock Classification and Engineering Considerations (West, Chs. 2, 3, 4, 5)

Geology 229 Engineering Geology. Lecture 6. Basic Rock Classification and Engineering Considerations (West, Chs. 2, 3, 4, 5) Geology 229 Engineering Geology Lecture 6 Basic Rock Classification and Engineering Considerations (West, Chs. 2, 3, 4, 5) Outline of this Lecture 1. Rock types and rock cycle 2. Geological and engineering

More information

Soil Mechanics/Geotechnical Engineering I Prof. Dilip Kumar Baidya Department of Civil Engineering Indian Institute of Technology, Kharagpur

Soil Mechanics/Geotechnical Engineering I Prof. Dilip Kumar Baidya Department of Civil Engineering Indian Institute of Technology, Kharagpur Soil Mechanics/Geotechnical Engineering I Prof. Dilip Kumar Baidya Department of Civil Engineering Indian Institute of Technology, Kharagpur Lecture - 01 Rock Cycle Good morning. I welcome you to this

More information

CHAPTER 4 STATISTICAL MODELS FOR STRENGTH USING SPSS

CHAPTER 4 STATISTICAL MODELS FOR STRENGTH USING SPSS 69 CHAPTER 4 STATISTICAL MODELS FOR STRENGTH USING SPSS 4.1 INTRODUCTION Mix design for concrete is a process of search for a mixture satisfying the required performance of concrete, such as workability,

More information

EPS 50 Lab 4: Sedimentary Rocks

EPS 50 Lab 4: Sedimentary Rocks Name: EPS 50 Lab 4: Sedimentary Rocks Grotzinger and Jordan, Chapter 5 Introduction In this lab we will classify sedimentary rocks and investigate the relationship between environmental conditions and

More information

MECHANISMS FOR THE CHANGES IN FLUIDITY AND HYDRATION KINETICS OF GROUTS AFTER MIXING

MECHANISMS FOR THE CHANGES IN FLUIDITY AND HYDRATION KINETICS OF GROUTS AFTER MIXING MECHANISMS FOR THE CHANGES IN FLUIDITY AND HYDRATION KINETICS OF GROUTS AFTER MIXING Keisuke Takahashi 1* and Thomas Bier 2 1 UBE Industries, Ltd., Research & Development Department, JAPAN. 2 TU Bergakademie

More information

International journal of Engineering Research-Online A Peer Reviewed International Journal Articles available online

International journal of Engineering Research-Online A Peer Reviewed International Journal Articles available online RESEARCH ARTICLE ISSN: 2321-7758 AN INVESTIGATION ON STRENGTH CHARACTERISTICS OF BASALT FIBRE REINFORCED CONCRETE SANGAMESH UPASI 1, SUNIL KUMAR H.S 1, MANJUNATHA. H 2, DR.K.B.PRAKASH 3 1 UG Students,

More information

Geology and Soil Mechanics /1A ( ) Mark the best answer on the multiple choice answer sheet.

Geology and Soil Mechanics /1A ( ) Mark the best answer on the multiple choice answer sheet. Geology and Soil Mechanics 55401 /1A (2003-2004) Mark the best answer on the multiple choice answer sheet. 1. Soil mechanics is the application of hydraulics, geology and mechanics to problems relating

More information

Moulding and Casting Methods 2. Greensand system

Moulding and Casting Methods 2. Greensand system MME 345, Lecture 22 Moulding and Casting Methods 2. Greensand system Ref: [1] Heine, Loper & Rosenthal, Principles of Metal Casting, McGraw-Hill, 1976 [2] Beeley, Foundry Technology, Butterworth-Heinemann,

More information

Chapter 1 - Soil Mechanics Review Part A

Chapter 1 - Soil Mechanics Review Part A Chapter 1 - Soil Mechanics Review Part A 1.1 Introduction Geotechnical Engineer is concerned with predicting / controlling Failure/Stability Deformations Influence of water (Seepage etc.) Soil behavour

More information

Geology 229 Engineering and Environmental Geology. Lecture 5. Engineering Properties of Rocks (West, Ch. 6)

Geology 229 Engineering and Environmental Geology. Lecture 5. Engineering Properties of Rocks (West, Ch. 6) Geology 229 Engineering and Environmental Geology Lecture 5 Engineering Properties of Rocks (West, Ch. 6) Outline of this Lecture 1. Triaxial rock mechanics test Mohr circle Combination of Coulomb shear

More information

Engineering Properties of Soil-Fly Ash Subgrade Mixtures

Engineering Properties of Soil-Fly Ash Subgrade Mixtures Engineering Properties of Soil-Fly Ash Subgrade Mixtures Zachary G. Thomas Graduate Research Assistant Iowa State University Department of Civil and Construction Engineering 394 Town Engineering Building

More information

MAINTENANCE AND REHABILITATION OF LOW COST SURFACE DRESSING FOR LOW VOLUME ROADS EXPERIMENTAL ROAD SITES

MAINTENANCE AND REHABILITATION OF LOW COST SURFACE DRESSING FOR LOW VOLUME ROADS EXPERIMENTAL ROAD SITES 1 MAINTENANCE AND REHABILITATION OF LOW COST SURFACE DRESSING FOR LOW VOLUME ROADS EXPERIMENTAL ROAD SITES Pétur Pétursson, Icelandic Building Research Institute Reykjavik, Iceland e-mail: petur.p@rabygg.is

More information

Analysis of soil failure modes using flume tests

Analysis of soil failure modes using flume tests Analysis of soil failure modes using flume tests A. Spickermann & J.-P. Malet Institute of Earth Physics, CNRS UMR 751, University of Strasbourg, Strasbourg, France Th.W.J. van Asch, M.C.G. van Maarseveen,

More information

Physical and chemical characteristics of natural limestone fillers: mix properties and packing density

Physical and chemical characteristics of natural limestone fillers: mix properties and packing density 3/3/ Physical and chemical characteristics of natural limestone fillers: mix properties and packing density Luc COURARD, Eric PIRARD and Huan HE Université de Liège, Belgium TC-SCM WORKSHOP, CYPRUS, 9-3

More information

Earth Science Chapter 6 Rocks

Earth Science Chapter 6 Rocks Earth Science Chapter 6 Rocks I. Rocks and the Rock Cycle * Material that makes up the solid part of the Earth. * Made of a variety of different combinations of minerals and organic matter. A. Three Major

More information

Rock Identification. invisible rhyolite andesite basalt komatiite. visible granite diorite gabbro peridotite

Rock Identification. invisible rhyolite andesite basalt komatiite. visible granite diorite gabbro peridotite Rock Identification The samples in this lab are arranged into four groups: igneous, sedimentary, metamorphic, and unknown. Study the igneous, sedimentary, and metamorphic collections to get an idea of

More information

Effects of Basalt Fibres on Mechanical Properties of Concrete

Effects of Basalt Fibres on Mechanical Properties of Concrete Effects of Basalt Fibres on Mechanical Properties of Concrete A. M. El-Gelani 1, C.M. High 2, S. H. Rizkalla 3 and E. A. Abdalla 4 1,4 University of Tripoli, Civil Engineering Department, Tripoli, Libya

More information

GEOL Lab 9 (Carbonate Sedimentary Rocks in Hand Sample and Thin Section)

GEOL Lab 9 (Carbonate Sedimentary Rocks in Hand Sample and Thin Section) GEOL 333 - Lab 9 (Carbonate Sedimentary Rocks in Hand Sample and Thin Section) Sedimentary Rock Classification - As we learned last week, sedimentary rock, which forms by accumulation and lithification

More information

PCE WITH WELL-DEFINED STRUCTURES AS POWERFUL CONCRETE SUPERPLASTICIZERS FOR ALKALI-ACTIVATED BINDERS

PCE WITH WELL-DEFINED STRUCTURES AS POWERFUL CONCRETE SUPERPLASTICIZERS FOR ALKALI-ACTIVATED BINDERS PCE WITH WELL-DEFINED STRUCTURES AS POWERFUL CONCRETE SUPERPLASTICIZERS FOR ALKALI-ACTIVATED BINDERS 2 ND INTERNATIONAL CONFERENCE ON POLYCARBOXYLATE SUPERPLASTICIZERS 28. SEPTEMBER 2017 SIKA TECHNOLOGY

More information

Mar 1, 2018 LAB MANUAL INDEX 1. Table of Contents Laboratory Testing Methods Reducing Aggregate Field Samples to Testing Size (Ver.

Mar 1, 2018 LAB MANUAL INDEX 1. Table of Contents Laboratory Testing Methods Reducing Aggregate Field Samples to Testing Size (Ver. Mar 1, 2018 LAB MANUAL INDEX 1 Table of Contents Laboratory Testing Methods 1000 Standard Practices (Ver. Sep 23, 2014) 1001 Receiving and Identifying Samples (Ver. Mar 1, 2018) 1002 Reducing Aggregate

More information

Chapter I Basic Characteristics of Soils

Chapter I Basic Characteristics of Soils Chapter I Basic Characteristics of Soils Outline 1. The Nature of Soils (section 1.1 Craig) 2. Soil Texture (section 1.1 Craig) 3. Grain Size and Grain Size Distribution (section 1.2 Craig) 4. Particle

More information

Cubzac-les-Ponts Experimental Embankments on Soft Clay

Cubzac-les-Ponts Experimental Embankments on Soft Clay Cubzac-les-Ponts Experimental Embankments on Soft Clay 1 Introduction In the 197 s, a series of test embankments were constructed on soft clay at Cubzac-les-Ponts in France. These full-scale field tests

More information

SHEAR STRENGTH OF SOIL

SHEAR STRENGTH OF SOIL SHEAR STRENGTH OF SOIL Necessity of studying Shear Strength of soils : Soil failure usually occurs in the form of shearing along internal surface within the soil. Shear Strength: Thus, structural strength

More information

Changes in soil deformation and shear strength by internal erosion

Changes in soil deformation and shear strength by internal erosion Changes in soil deformation and shear strength by internal erosion C. Chen & L. M. Zhang The Hong Kong University of Science and Technology, Hong Kong, China D. S. Chang AECOM Asia Company Ltd., Hong Kong,

More information

LAB 2 IDENTIFYING MATERIALS FOR MAKING SOILS: ROCK AND PARENT MATERIALS

LAB 2 IDENTIFYING MATERIALS FOR MAKING SOILS: ROCK AND PARENT MATERIALS LAB 2 IDENTIFYING MATERIALS FOR MAKING SOILS: ROCK AND PARENT MATERIALS Learning outcomes The student is able to: 1. understand and identify rocks 2. understand and identify parent materials 3. recognize

More information

Bowen s Chemical Stability Series

Bowen s Chemical Stability Series Lab 5 - Identification of Sedimentary Rocks Page - Introduction Sedimentary rocks are the second great rock group. Although they make up only a small percentage of the rocks in the earth s crust (~5%)

More information

Dry mix design. Lecture Notes in Transportation Systems Engineering. Prof. Tom V. Mathew. 1 Overview 1. 2 Selection of aggregates 1

Dry mix design. Lecture Notes in Transportation Systems Engineering. Prof. Tom V. Mathew. 1 Overview 1. 2 Selection of aggregates 1 ry mix design Lecture Notes in Transportation Systems Engineering Prof. Tom V. Mathew Contents 1 Overview 1 2 Selection of aggregates 1 3 Aggregate gradation 2 4 Proportioning of aggregates 2 5 Example

More information

COMPARISONS OF LINEAR REGRESSION MODELS FOR PROPERTIES OF ALKALI- ACTIVATED BINDER CONCRETE

COMPARISONS OF LINEAR REGRESSION MODELS FOR PROPERTIES OF ALKALI- ACTIVATED BINDER CONCRETE COMPARISONS OF LINEAR REGRESSION MODELS FOR PROPERTIES OF ALKALI- ACTIVATED BINDER CONCRETE Arkamitra Kar Birla Institute of Technology and Science - Pilani, Telangana, India Udaya B. Halabe West Virginia

More information

Non-Destructive Assessment of Residual Strength of Thermally Damaged Concrete Made with Different Aggregate Types

Non-Destructive Assessment of Residual Strength of Thermally Damaged Concrete Made with Different Aggregate Types IOP Conference Series: Materials Science and Engineering PAPER OPEN ACCESS NonDestructive Assessment of Residual Strength of Thermally Damaged Concrete Made with Different Aggregate Types To cite this

More information

4. Objectives of Research work

4. Objectives of Research work 4. Objectives of Research work 4.1 Objectives of Study: The design of bellows is challenging looking to varieties of applications and evaluation of stresses is further difficult to approximate due to its

More information

Chemical and Mechanical Mechanisms of Moisture Damage in Hot Mix Asphalt Pavements

Chemical and Mechanical Mechanisms of Moisture Damage in Hot Mix Asphalt Pavements Chemical and Mechanical Mechanisms of Moisture Damage in Hot Mix Asphalt Pavements Dallas N. Little Texas A&M University David R. Jones Owens Corning Moisture Damage Loss of strength and durability due

More information