Laboratory investigation of Suffusion on dame core of glacial till

Transcription

1 Laboratory investigation of Suffusion on dame core of glacial till Daniel Yadetie Tuffa Civil Engineering, master's level (1 credits) 17 Luleå University of Technology Department of Civil, Environmental and Natural Resources Engineering

2 Laboratory Investigation of Suffusion on dam cores of Glacial Till Daniel Yadetie Tuffa Master programme in Civil Engineering, with specialization in mining and Geotechnical Engineering Luleå University of Technology

3 Abstract The objective of this study is to provide a better understanding of suffusion characteristics of glacial soils and to present a simple yet reliable assessment procedure for determination of suffusion in the laboratory. Internal erosion by suffusion occurs in the core of an embankment dam when the ability of the soil to resist seepage forces is exceeded and voids are large enough to allow the transport of fine particles through the pores. Soils susceptible to suffusion are described as internally unstable. dams with core of broadly graded glacial moraines (tills) exhibit signs of internal erosion to a larger extent than dams constructed with other types of materials. The Suffusion behavior of glacial soils has been investigated through two different permeameter suffusion test have been employed, small scale permeameter and big scale permeameter. Details of the equipment along with its calibration, testing and sampling procedures are provided. The testing program were performed 9 test with different compaction degree in small scale permeameter and 2 test in big permeameter on internally stable categories of till soil. The categories are defined based on the soil grain size distribution and according to the methods developed by Kenney & Lau and Burenkova. Layers are identified with suffusion if the post-test gradation curve exhibit changes in distribution compared to the initial condition and also the tests revealed that the effect of grain size distribution and relative degree of compaction on the internal erosion susceptibility of glacial till soils at different hydraulic gradients. Key words: Internal erosion; Suffusion; Filter; Core; Glacial till; Embankment dam; Hydraulic gradients.

4 Acknowledgement This thesis would not have been possible without the encouragement and constant support of my supervisor, Professor Juan Lau. I am deeply grateful to my supervisor for his inspiration, ideas, guidance, and constructive comments. I wish to express my warm and sincere thanks to my Examiners, Ingrid Silva and Jenny Lindblom, for your broad experience, encouragement and support. I am very grateful to PhD student Ingrid Silva, research Eng. Thomas Forsberg for their assistance in setting up the laboratory equipment and sample preparation. I am obliged to many of my colleagues and classmates who supported me, especially Yonas Lemma, Dr. Stefan, Deniz Dagli and Samuel Kebed who were with me all the time, from the beginning to the end. I would like to show my gratitude to associate Professor Hans Mattsson, for his repetitive encouragement and guidance in many ways during my research. I would like thank Lulea University of Technology, department of civil environmental and Natural Resource Engineering, for organizing this research I owe my loving thanks to my friend, Sara Hagos, who provided great support and encouragement during the entire period of my study. Finally, I offer my regards and blessings to my mother Etete for her endless support. Daniel Yadetie December, 17 Luleå, Sweden

5 Contents Abstract... 2 Acknowledgement INTRODUCTION OVERVIEW AND STATEMENT OF PROBLEM Objectives and methodology Scope and limitations Thesis structure Literatures review Kenney and Lau (1984, 1985) Burenkova (1993) Skempton and Brogan (1994) Foster and Fell (1999, 1) Wan and Fell (4a, 8) Li and Fannin (8) Properties of Soil sample Experimental program Particle size analysis Wet Sieving Sedimentation Pipette Analysis Plasticity Proctor compaction Relative density and Molding water content Define the testing natural glacial soil Test program on glacial till Laboratory Apparatus Small Permeameter Big Permeameter Sample Preparation and Testing Procedure Sample Preparation Testing Procedure Result... 34

6 5.1 A curve matching technique Hydraulic gradient for suffusion Small-scale permeameter suffusion studies B-S1-8a-f B-S1-85c-f B-S1-9a-f Big-Scale Permeameter Suffusion studies Specimen B-S1-8a-F Specimen B-S1-8b-F Analysis of result Conclusion and recommendation... 5 Conclusions... 5 Recommendation... 5 REFERENCES Appendix Figure 1. Teton dam failure due to internal erosion(idaho,1976)... 8 Figure 2 gap-graded soil and coarsely graded soil which are internally unstable and susceptible to suffusion from ICOLD ( Figure 3 Grain size distribution of internally (a ) stable and (b) unstable material tested by Kenney and Lau(1984,85) Figure 4 Classification of suffusive and non-suffusive soil composition (Burenkova 1993). (Zones I and IIIsuffusive ; Zone II - non-suffusive; Zone IV Figure 5grain size distribution of eight soil samples tested by seepage test (Burenkova (1993) Figure 6 Upward flow seepage cell (Skempton and Brogan 1994) Figure 7 Gradation curve of test sample (Skempton and Brogan 1994) Figure 8 probability of internal instability for silt-sand-gravel and clay-silt-sand-gravel soils of limited clay content and plasticity (Wan and Fell 4a) Figure 9 Criteria for internal instability (Li and Fannin 8) Figure 1A) on the left side schematic diagram showing regular pipette test set used; After (Bardet, 1997), on the right side showing Pipette test for soil sample Figure 11 combined Grain size distribution curve sieve and pipette analysis of material used for the test sample Figure 12 a) a graph of the test material defined based on Kenney and Lau stability index>1. b) A graph of the test material defined showed based on Burenkova Figure 13 Outline of the test program Figure 14 Photo small permeameter seepage test and constant head apparatus assembly Figure 15 photo small scale seepage cell and type of porous disc Figure 16 photo: Apparatus, water supply system and constant head tank... 3

7 Figure 17 big permeameter apparatus and location of relevant level for gradient analysis Figure 18 Specimen B-S1-8a-f1: particle size distributions of the post-test gradations layer 1 to Figure 19 Specimen B-S1-85c-f1: particle size distributions of the post-test gradations layer 1 to Figure Specimen B-S1-85c-f1-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles Figure 21 photo Figure 22 Specimen B-S19a-f1: particle size distributions of the post-test gradations layer 1 to Figure 23 Specimen B-S1-8a-F1: particle size distributions of the initial gradation and post-test layer.. 4 Figure 24 temporal progression of head loss profile across specimen B-S1-8a-F Figure 25 Sample B-S1-8a-F1 temporal variation of hydraulic gradient and flow rate Figure 26 Specimen B-S1-8a-F1-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles Figure 27 Specimen B-S1-8b-F1: particle size distributions of the initial gradation and post-test layer. 42 Figure 28 temporal progression of head loss profile across specimen B-S1-8b-F Figure 29 Sample B-S1-8b-F1 temporal variation of hydraulic gradient and flow rate Figure 3 Sample with porous plastic and wire mish filter Average flow velocity versus hydraulic gradient Figure 31 End-of-test head loss in terms of a) profiles across the specimens, and b) in the top, center and bottom zone Figure 32 Maximum local gradient relative average gradient at end-of-test for each layers of the specimens Figure 33 Specimen B-S1-8a-f1: particle size distributions of the post-test gradations layer 1 to Figure 34 Specimen B-S1-8b-f1: particle size distributions of the post-test gradations layer 1 to Figure 35 Specimen B-S1-8b-f1-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles Figure 36 Specimen B-S1-8c-f1: particle size distributions of the post-test gradations layer 1 to Figure 37 Specimen B-S1-8c-f1-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles Figure 38 Specimen B-S1-85a-f1: particle size distributions of the post-test gradations layer 1 to Figure 39 Specimen B-S1-85a-f1-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles Figure 4 Specimen B-S1-85b-f1: particle size distributions of the post-test gradations layer 1 to Figure 41 Specimen B-S1-85b-f1-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles Figure 42 Specimen B-S1-85c-f2: particle size distributions of the post-test gradations layer 1 to Figure 43 Specimen B-S1-85c-f2-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles Figure 44 Specimen B-S1-9a-f1: particle size distributions of the post-test gradations layer 1 to Figure 45 Specimen B-S1-9b-f1: particle size distributions of the post-test gradations layer 1 to Figure 46 Specimen B-S1-9c-f2: particle size distributions of the post-test gradations layer 1 to Figure 47 Specimen B-S1-85c-f2-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles... 6 Table 1 Properties of test sample and test result (Skempton and Brogan 1994)... 16

8 Table 2 Modified Procter data on natural gradations Table 3 Modified Proctor results on natural soils Table 4Grading characteristics of test specimen Table 5 Post grading characteristics and mass loss for unstable test specimens... 44

9 1 INTRODUCTION 1.1 OVERVIEW AND STATEMENT OF PROBLEM Evaluating internal erosion resistance glacial soil in embankment dams is of concern in many geotechnical and hydraulic engineering problems. This is mostly the case internal erosion investigations of core materials in dams and embankments which is constructed to impound water in reservoirs for various purposes stability assessment of slopes and banks, as well as in erosion control of excavation irrigation canals. Internal erosion is the second most frequent reason of failure of embankment dams for both new and existing dams after overtopping (ICOLD 16). Figure 1 shows a famous case of internal erosion failure at Teton dam in the USA (1976). Internal erosion is the phenomenon of water seepage within earth structures, such as embankments, dams or dikes, can cause a detachment and a transport of particles from the soil constituting the structure or its foundation. To assess the internal erosion and the potential failure modes, four type mechanical processes should be considered ICOLD (13) 1) Concentrated leak erosion, involves when tractive seepage forces along a surface erosion of (i.e., a crack within the soil, adjacent to a wall or conduit, along the embankment-foundation contact) are sufficient to move soil particles into an unprotected area, or at the interface of a coarse and fine layer in the embankment or foundation. 2) Backward erosion, involves detachment of soils particles when seepage exits to a free unfiltered surface 3) Contact erosion, occurs at an interface between a fine soil layer and another layer made of a coarser soil and 4) Suffusion is an internal erosion mechanism, which evolves selective erosion of fine particle are removed through the void between the larger particles by seepage force. Figure 1 Teton dam failure due to internal erosion (Idaho,1976)

10 Figure 2 gap-graded soil and coarsely graded soil which are internally unstable and susceptible to suffusion from ICOLD (13 In this last case, many attempts have been made to investigate the criteria that should be satisfied for suffusion. The identified criteria are; 1) The size of the fine soil particles must be smaller than the size of the constrictions between the coarser particles, which form the basic skeleton of the soil. 2) The amount of fine soil particles must be less than enough to fill the voids of the basic skeleton formed by the coarser particles. If there are more than enough fine soil particles for void filling, the coarser particles will be floating in the matrix of fine soil particles, instead of forming the basic soil skeleton; and 3. The velocity of flow through the soil matrix must be high enough to move the loose fine soil particles through the constrictions between the larger soil particles Significant contributions have been made on the first two the criteria by Kenney and Lau, 1985 Lafleur et al., 1989 ;), Burenkova, 1993 ;) US Army Corps of Engineers (1953). Foster, 1999; and Hunter et al., 12); Kezdi (1979), Lafleur, Mlynarek, and Rollin (1989), and also for all criteria, Skempton and Brogan (1994), Moffat (5) and Li (8). (Sherard, 1984; Lilja et al., 1998; Ravaska, 1997; Wan and Fell (8 Rönnqvist and Viklander (14). made evaluate the internal stability soil. Very little is known about the influence of suction on the hydraulic gradient on initiation of suffusion of glacial soil. There is a clear need to investigate the effect of hydraulic load.

11 A glacial soil is derived from the action of the pull, crush, mix and transport generated by the progression and regression of the glaciers, is extensively used in many parts of the world as impervious core material in embankment dams. This type of material is found in areas once glaciated and typically broadly or widely graded with a mixture of content from fines up to boulders. Coarse widely graded or gap graded soils such as those show figure 2 are susceptible to suffusion Furthermore, statistics indicate that dams with cores of glacial till are relatively more vulnerable to internal erosion compared to other soil types and its deserving more investigation. 1.2 Objectives and methodology This master thesis presents an experimental investigation suffusion and analysis of data on glacial core of embankment dams. The investigation was conducted at Geotechnical Engineering at Lulea University of Technology. The main objective of this study is to provide a better understanding of suffusion characteristics of glacial soils and to present a simple yet reliable assessment procedure for determination of suffusion in the laboratory. Suffusion characteristics of saturated tills in a big and small permeameter are investigated and compared. The findings are presented in a unified framework. More specifically, this study intends to: 1) To investigate the hydraulic gradient for suffusion to initiate in glacial material 2) Understand the suffusion behavior of saturated glacial soils using laboratory testing 3) Provide simple assessment procedures for suffusion of glacial till soils and to present key factors affecting erosion of these soils. 4) The effect filter used in small scale apparatus to the tests of specimens. 5) Compare the effect of suffusion in different compaction degree 1.3 Scope and limitations This thesis work has investigated the dam core soils which are made by glacial till. The work carried out by laboratory permeability testing. The seepage tests were conducted with unfiltered municipal tap water and average gradients applied over the specimen. The gradient has applied from 1 to. This research results are limited to the geotechnical characteristics of the glacial till soil such as relative density, grain size gradations and to the hydraulic loads subjected to the specimens investigated in the laboratory program.

12 1.4 Thesis structure The structure of the thesis is as follows. Chapter 1: Introduction to the topic of internal soil erosion and suffusion, and description of the objective and scope of this study. Chapter 2 A review of research literature dealing with internal erosion as the principal mechanism in suffusion with focus on the constant head test, Chapter 3 Documents the Properties of Soil Samples Used in the Current Study such classification of soil sample, pipet analysis, modified compaction test and defined categories of test sample Chapter 4 Description of the big and small scale permeameter test), including a seepage test review of the standard constant head, followed by details about the test program, setup, testing procedure. Chapter 5. Results Chapter 6 Result Analysis Chapter 7: Conclusion and Recommendation Chapter 8 references Appendix

13 2 Literatures review This thesis enquiry has theoretical knowledge from previous studies in similar topic. Between these previous works, some of the key works are reviewed below. These literatures refer to experiments on internal erosion performed on glacial material in many different place 2.1 Kenney and Lau (1984, 1985) They hypothesized that a soil would behave as a stable system when the size of the loose particle was larger than the controlling constriction size of the primary fabric. Kenney et al(1984) verified their hypothesis by applying downward flow test system on three mixture of gap-graded gravel and sand based on their hypothesis results of the seepage showed their prediction were correct.by extending their works (1985,86) downward flow seepage tests on14 cohesionless sand-gravel soil sample particle size up to 1mm. Kenney and Lau postulated the H: F shape curve with H/F as stability index, where F denotes mass passing (%) at grain size D and H denotes mass increment (%) between D and 4D, with D as an arbitrary particle size. The evaluation range for widely graded soils (uniformity, Cu = d6/d1 > 3) is defined by F %. A stability index less than one within the evaluation range ((H/F)min<1) indicates that a soil is deficient in the finer fraction, thus, potentially internally unstable, which means that fine particles can be washed out by seepage. Shape curves for the unstable and stable samples tested by Kenney and Lau are shown in figure 3 a and b

14 Figure 3 Grain size distribution of internally (a ) stable and (b) unstable material tested by Kenney and Lau(1984,85) 2.2 Burenkova (1993) Burenkova planned a predictive method built on the results of laboratory tests on 22 cohesion less sand gravel soils of maximum sizes up to 1mm, and coefficients of uniformity, Cu, up to. The basic assumption was that a smaller size fraction did not form part of the basic soil skeleton if it did not cause volume increase when mixed with a coarser size fraction. If the volume of the specimen increased after addition of a finer fraction this finer fraction was estimated as belonging to the soil skeleton. If additional fraction did not increase the volume of the specimen, the fraction was considered as belonging to the loose particles. According to, Burenkova (1993) proposed a geometric condition for internal stability of a soil that depends on the conditional factors of uniformity d9 /d6 and d9 /d15 ratios where d9 is the sieve size for which 9% of the sample by weight passes. The d9 /d6 ratio represents the slope of the coarse part of the particle size distribution plot. High values represent near single size coarse particles which will have large constriction spaces compared to a well graded soil. The d9 /d15 can be regarded as a measure of the filter action between the coarse fraction and the finer fraction. Burenkova (1993) defined boundaries separating suffusive soil from non suffusive soil. According to represents a zone of non suffusive compositions and Zone Boundaries were defined separating suffusive soils from non-suffusive soils. Zones I and III represent zones of suffusive compositions; Zone II represents a zone of non-suffusive compositions; and Zone IV represents a zone of artificial soils. Zone II (non-suffusive) Figure 4. Boundaries are defined as follows:.76 log (h ) 1 < h < 1.86 log (h ) Figure 4 Classification of suffusive and non-suffusive soil composition (Burenkova 1993). (Zones I and III- suffusive; Zone II - non-suffusive; Zone IV-

15 Burenkova (1993) also carried out a series of seepage test to study the effects of suffusion the size of eroded particle. The eight test confirmed 4 suffusive and 4 non suffusive soil sample whose grain size distribution curve are shown in figure 5.the seepage tests were carried out at hydraulic gradient up to 2,5. Figure 5 grain size distribution of eight soil samples tested by seepage test (Burenkova (1993) 2.3 Skempton and Brogan (1994) They attempted to verify the critical hydraulic gradient at which wash out of the fine particle would start after carried out laboratory seepage test to investigate the internal instability in sandy gravels. Skempton and Brogan use upward flow seepage cell of 139mm diameter and 155mm internal length, saturated in low hydraulic gradient before the test began, the test set up show in figure 6. They plotted critical hydraulic gradient contrary to stability index(h/f) and recognized that the critical hydraulic gradient increase rapidly from the low value to high value over the line represented by H/F=1, which is the boundary between stable and unstable grading (Kenney and Lau 1985,1986) however they recommended that the relationship required further investigation. Skempton and Brogan also observed, erosion of sand grains might happen at hydraulic gradient. 3 to.2 of the theoretical critical gradient for a homogenous granular material of the same porosity unstable sandy gravel (sample A and B) figure 7. The critical hydraulic gradient, corresponding to zero effective stress, is defined as:

16 Where ic: Critical hydraulic gradient, Ƞ: Porosity of the material, Gs: Specific gravity of the soil grains, Ƴ: Submerged unit weight of soil, Ƴw: Unit weight water. ii cc = (1 ηη)(gg ss 1) = γγ γγ ww (1) They suggested that, in an internally unstable soil, the overburden load was probably carried on a primary fabric so that sand was relatively under small pressure. Table 1 summarizes the result of four seepage tests carried out by Skempton and Brogan (1994) and the proprieties of the soil sample. Figure 6 Upward flow seepage cell (Skempton and Brogan 1994)

17 Figure 7 Gradation curve of test sample (Skempton and Brogan 1994) Table 1 Properties of test sample and test result (Skempton and Brogan 1994) Test sample A B C D Porosity,ƞ (%) D15 (mm) Cu Permeability(cm/s) Filter ratio component, dc15/df85 Stability index,(h/f)min Critical gradient,ic,in test

18 Note sample A and B, with (H/F) min <1, were assessed as internal unstable, whereas sample C and D were assessed as stable 2.4 Foster and Fell (1999, 1) They presented the boundary of no erosion and continues erosion filter test behavior by studied analyzing of experimental data on several filter tests carried by other and by performing the continual erosion filter test against bases which included a non-plastic glacial till sourced from the Australian. Foster and fell (1999,1) also reviewed the performance of the filter in existing dam. Based on the result of their investigation, they proposed the boundary of the filter behavior related to Df15 of filter which is a concentrated leak was simulated by punching a 1mm or 2mm (Continuing erosion tests) diameter stiff wire through the compacted base and some characteristic of base material related to Db85Db9 and Db95, and fine contentment of base material, in terms of broadly graded base soils, these eroded at filter opening sizes much smaller than that of the fine-grained bases. Foster and Fell attributed this to the shape of the gradation curve and its fine to medium sand size range (.75mm to 1.18mm). Thus, the lower the amount of fine to medium sand sizes in the base, the finer the filter needed to arrest erosion Generally, Foster and Fell investigation is not related to the study of internal stability of soil. Their investigation, even though provide useful information to help assessing the likelihood of moving fine particle through the void of coarser soil Skelton, as happen in suffusion process 2.5 Wan and Fell (4a, 8) They postulated by extending Burenkova (1993) work, who did not put a clear a clear boundary between internally stable and unstable soils in the data set hence, Wan and Fell (4,8) developed contours for forecasting the probability of internal instability by logistic regression of h and h. Their modified Burenkova method for broadly graded and gap-graded soils is shown in Figure (8) the probability contours are represented by the following equations (Wan and Fell 4a): PPPP = eeee 1 + eeee eeeeee 2 For silt-sand-gravel soils and clay-silt-sand-gravel soils percent of limited clay content and plasticity ZZ = log( ) 3.648(h )3.71 equ.3 For sand-gravel soils with less than 1 percent non-plastic fines zz = 3.875llllll(h ) 3591(h ) eeeeee. 4

19 The probabilities should not be used directly in a risk assessment, but rather used to help develop a list of more likely and less likely factors during an elicitation of probability estimates. Figure 8 probability of internal instability for silt-sand-gravel and clay-silt-sand-gravel soils of limited clay content and plasticity (Wan and Fell 4a) 2.5 Li and Fannin (8) Li and Fannin (8) studied two commonly used methods to define the susceptibility to internal instability: Kézdi (1979) and Kenney and Lau (1985, 1986). Kézdi allocated a soil into a coarse fraction and a fine fraction at one point along its grain -size distribution curve and applied Tirzah s (1939) rule for designing protective filters (D 15/d 85) to the two fractions, with the fine fraction as the base and the coarse fraction as the filter, to assess if the soil would self-filter and be internally stable. The mass increment (H) over D 15 and d 85 is constant and equal to 15 percent, resulting in a criterion for instability of H less than 15 percent. Kenney and Lau postulated the H: F shape curve with H/F as stability index, where F denotes mass passing (%) at grain size D and H denotes mass increment (%) between D and 4D, with D as an arbitrary particle size. They originally proposed a criterion in 1985 for internal instability of H/F < 1.3, applicable within F 3 percent (and cu 3) for narrowly graded soils and within F percent (and cu > 3) for widely graded soils. This criterion was subsequently revised in 1986 to H/F < 1.. This method is commonly used for cohesion less sand-gravel soils. Li and Fannin (8) shared aspects of these two methods for evaluating the susceptibility to internal instability. They concluded that the Kenney and Lau criterion is more conservative at F > 15 percent, but the Kézdi criterion is more conservative at F < 15 percent. The combined criteria are shown in Figure 9, where the respective values of H and F are plotted at (H/F) min.

20 Figure 9 Criteria for internal instability (Li and Fannin 8)

21 3 Properties of Soil sample 3.1 Experimental program Glacial core sample that are from dam site were provided by dam owner to geotechnical laboratory of Luleå University of technology for physical index and seepage test. Before starting the seepage tests, a series of pre-test (index) were performed to know the classification and some geotechnical properties of the material. The pre-test done are summarized as follow 3.2 Particle size analysis The particle size distribution of the glacial core material is primarily used for classification purposes and to evaluate the gradation carachetrsitcs suffusion after the test. Figure 11 shows the particle size distribution curve of the natural glacial soil before the test have done.it can represent the pre-sieve gradation curve for small permeameter test specimens. The distribution of particle sizes larger than.63 mm is determined by sieve, while distribution of particles sizes smaller than.63 mm is determined by sedimentation process using a pipette analysis Wet Sieving The sieve analysis determines the grain size distribution curve of soil samples based on the available mass of fines, grains smaller than 63 μm in the samples, either hydrometer or pipette analysis has been adopted Sedimentation The theory of sedimentation is since large particles suspended in a liquid settle more quickly than small particles, if all particles have similar densities and shapes. By assuming that, particles are approximately spherical, the relation between the velocity and particle diameter is given by Stokes law, which is stated as: ννννdd 2 equ.5 A total of 8 samples like the sieving were gone through sedimentation test. Pipette analysis were employed based on the mass of the soil passing sieve size.63mm. For those samples that have more than 5 grams of fine particles (finer than.63mm),

22 3.2.3 Pipette Analysis This is for the determination of the sieve of fine particle distribution (smaller than 63µm) in a soil sample by mechanical analysis. An analysis of this kind expresses quantity the proportions by weight of various sizes of particles present in the soil. Like hydrometer test, it is recommended as a standard procedure to use dispersion agent to avoid flocculation. The apparatus used consists of regular sampling pipette, capable of measuring 1 ±.2 ml of liquid, with a lowering and raising support. (Figure 1), Dispersion apparatus (1ml), 5 ml of stock solution of sodium hexa-metaphosphate prepared as in the hydrometer test, many sedimentation cylinders, thermometer, ranging from to 5 C, accurate to.5 C, stopwatch, and balance which is accurate to.1 g. During the test, it is observed that Pipette analysis has several advantages over hydrometer analysis which is also supported by (Bardet, 1997). It takes less time because the sampling depth is adjustable, whereas it is fixed in hydrometer analysis. The calculations are also simpler and there is no need to account for the correction of meniscus or hydrometer dilation. However, compared to hydrometer analysis, pipette analysis is less adapted to the conditions encountered in a field laboratory. It requires accurate weight measurement Figure 1A) on the left side schematic diagram showing regular pipette test set used; After (Bardet, 1997), on the right side showing Pipette test for soil sample

23 Procedure and analysis of this test were adopted from (Bardet, 1997), which is based on British Standard PARTICLE SIZE DISTRIBUTION -natural glacial soil Mass passing [%] ,1,1, Grain size [mm] Tes t 1 Tes t 2 Tes t 3 Ave rag e Figure 11 combined Grain size distribution curve sieve and pipette analysis of material used for the test sample 3.2 Plasticity The Atterberg limit test were carried out on the soil sample according to (ASTM D4318) reveled that soil have non plastic fines. Generally, cohesionless soils are less resistant to erosion than plastic soils (ICOLD, 13; Sherard, 1953) and, in terms of glacial till soils, PI > 4 inhibits internal instability (Crawford-Flett, 14). 3.3 Proctor compaction Modified proctor compaction was conducted on a glacial core material, to obtain information maximum dry density and optimum moisture content. It was conducted according to ASTM standard for laboratory compaction characteristics of soil using modified effort (2,7KN/m3). (ASTM D1557) on D < mm. The compaction tests are summarized in Table 2 with complete density curves in Figure3.3. This information was essential for controlling the dry density and the molding water content of the specimens.

24 3.3.1 Relative density and Molding water content Test sample were prepared at the modified optimum water content (OWC). To attained desire molding water content, the appropriate amount of water was added to soil sample, which was cured for at least one day before a test sample was prepared. Water content test were carried out on remaining soil trimmed from compaction to find out the actual molding water content of the test sample. The Samples were prepared at three degree of compaction defined respect to the modified Proctor test. The degrees of compaction considered are: a) 9%, well-compacted representing a material on the borderline of acceptance based on the recommendation of the current Swedish dam safety guidelines (Svensk Energy, 12). The well-compacted specimen is to create a dense state according to standard, thus representing a well-engineered homogenous filling with acceptable erosion resistance b) 85% representing low compacted material; and c) 8% representing poorly compacted material. Table 2 Modified Procter data on natural gradations Sample Water content Dry density Test-1 5 2,7 Test-2 6 2,1 Test-3 6, Test-4 7, Maximum Dry Density 2.11 (g/cm3) Optimum Moisture (%) 6.5

25 2,14 Moisture Density Test Results 2,12 Dry Density (g/cm3) 2,1 2,8 2,6 2,4 2,2 4,7 5,7 6,7 7,7 Moisture Content (%) Calculated Curve Points Data Points Figure 3.3 Modified Proctor results on natural soils 3.4 Define the testing natural glacial soil The testing program includes three categories of till soil: i) internally stable, ii) internally unstable; and iii) soils in the transition zone between the two first categories. The categories are defined by applying the current available methods developed by Kenney & Lau (1985, 1986), Burenkova (1993). In this master thesis, the tests have been done only till soil internally stable. Based on Kenney and Lau (1985,1986) the method described in section (2.1) the test material has stability index greater than one which is defined a stable show in the figure 12a The Burenkova (1993) method is based on d9 /d6 and d9 /d15 ratios. The d9 /d6 ratio represents the slope of the coarse part of the particle size distribution plot, whilst d9 /d15 ratio is regarded as a measure of the filter action between the coarse fraction and the fine fraction. The results are reported in Figures 12b based on Burenkove (1993) a test material showed in zone 4 which define not a clear.

26 Table 4Grading characteristics of test specimen Test sample Clay size fractions (%) (<.2mm) Fine Content (<.75mm) Gravel Fraction (>4.75mm) Sand fraction( mm) plasticity Cu=d6/d1 Coefficient of uniformity Cc=(d3) 2 /(d1*d6) Coefficient of Curvature Soil Classification(AS TMD2488) Finer fraction estimation (%) B-S Noneplastic 5.9 SW 53 Mass Incriment,H(%) H= Mass passing Diamater D,F(%) Natural galacial soil Kennya-Lau limit curve H/F=1 Kenney-Lau Evaluation Range F<% a) D9/D , D9/D15 b) Figure 12 a) a graph of the test material defined based on Kenney and Lau stability index>1. b) A graph of the test material defined showed based on Burenkova

27 4 Test program on glacial till A glacial till soils, named dam-b, was provided by dam owners from dam sites. In total 9 tests have been performed in small scale and 2 tests in big scale permeameter in this study according to the scheme in Fig13. 9 tests were conducted on different Dam-B till; three on 8% relative density; three on 85% and three 9% on Dam-B till and two on 8% relative density for big permeameter. The specimens are identifiable by their denotation: in falling order by i) the source soil (e.g. Dam-B), ii) category of soil (stable) iii) relative density, iv) test number and, Vi) filter type Figure 13 Outline of the test program

28 4.1 Laboratory Apparatus Small Permeameter The apparatus consists of a cylindrical stainless-steel seepage cell of 1mm internal diameter and height of cylinder is 115mm mounted on a detachable plastic cap plate on the top and bottom with inlet and outlet, which have to be connected to the constant level tank and seepage out flow collecter. A different porous disc used such as wire-mesh and plastic-prous disc placed in the top and the bottom of soil sample for tests to check the effect.the objective of the porous disc used to prevent migration of material through valves and tubing during test Figure15 Figure 14 Photo small permeameter seepage test and constant head apparatus assembly.

29 Figure 15 photo small scale seepage cell and type of porous disc

30 4.1.2 Big Permeameter The big permeameter apparatus, designed and build to accommodate sample contain particle size up to 3mm. The permeameter cell was comprises rigid wall stainless steel cylinder that is 3mm internal diameter and internal height the cylinder is 45mm long containing the soil sample to be tested. Figure 16 The cell was fitted with lockable silicon O-rings with through an inlet tube of 15mm diameter on the top plate of loading position and a shaft passing through a pre-drilled hole around the external wall part of permeameter of the sample. The sample shaft is which allow a force to hold down the upper part of the O- rings plate. The bottom plate is mounted on a square steel and through the bottom plate the outlet tube of 15mm diameter connect to the lower tank to facilitate measurement the rate of flow through the system. The higher hydraulic gradient can have achieved by changing the position of the constant head tank manually, however, head roughly 26mm for test sample B-S1-8a-F1and mm test sample B-S1-8b-F1 which generate 13 and 1 average hydraulic gradient respectively The transparent manometer (piezometer) mounted on a stand with gradual scales. Seven piezometer points embedded at different depth of soil sample to provide the water pressure within soil sample. The drainage layer, located on the top of the sample, is mm height and its maximum grain size is mm. This layer serves to disperse the incoming flow to ensure more uniform water pressure on the upper surface of the soil sample. The filter layer at the bottom most of the permeameter cell is 5mm long a filter to the soil tested based on sherard and dunningan (1989) filter criteria provided the soil are internally stable. The objective of to deliver filtering against the bottom of the specimen meanwhile it allows for an open system for unhindered seepage The piezometers consisted of ordinary transparent tube, piezometer 1, 3,5,6,7 are in the same side and piezometer 2, 4 on their opposite side. Piezometer one and two gives the pore water pressure in the drainage. It is used for defining any losses in the supply system from the constant head reservoir, and for confirmation on the actual head applied on the top surface of the specimen. Piezometer 3,4,5,6 indicate on head losses through the specimen and piezometer 7 gives the pore water pressure in the filter layer. The seepage outflow was measured manually with a bucket and stopwatch at the overflow chute in the tub wall.

31 Figure 16 photo: Apparatus, water supply system and constant head tank

32 4.2 Sample Preparation and Testing Procedure Sample Preparation Compacted glacial soil samples were tested in this study, for both big and small permeameter. To make homogenization and to avoid segregation the samples were manually compacted to relative density and a desire molding moisture content Particles > mm and > 1mm were removed by hand (limiting D > mm and > 1mm to only infrequent particle) for both permeameter respectively. The specimen was thereafter divided into five equal sized streaks for small permeameter whereas for big permeameter the specimen divided in five equals sized including filter layer. The compaction was performed for both permeameter using 2,5kg steal cylinder rammer dropped from 5cm, subjected the soil to achieved target relative density. The target relative density was found from the maximum Modified Proctor density which is considered on tests performed on D < mm (see section 3.3: compaction). For big permeameter the first layer is compacted above the filter layer. The dry density of a test sample to ensure known mass of soil was compacted to precalculated thickness corresponding to the desired dry density. Lastly the amount of soil and the molding water content of the tested sample were measured so as actual dry density of the soil Testing Procedure Suffusion test procedure included the following stage 1) De-air the sample using CO2 (carbon die oxide): before upward saturation of the sample with water. The air content in the gaseous phase is replaced by upward incorporation of CO2 (carbon dioxide) by connecting in the bottom inlet of the sample cell. The aim of this procedure is to enhance quicker saturation of the sample. 2) Place the tested sample in the constant headsets. The connection tube which is subjected to the test is connect to test sample and allow the saturation in upward system in the low hydraulic gradient 3) Change the system to downward seepage to start the suffusion test once it has saturated 4) increase the hydraulic gradient by increasing the level of the constant head tank for big permeameter and whereas the small permeameter by changing the cell and allow water to flow through the sample until the condition appear to steady and the water levels in the manometer become recorded and observe if there is any erosion of the base that is transported through the filter into the collector can. 5) Dismantling cell: when several consistent sets of readings seepage in different head position have been obtained and any eroded particles was recovered and analyzed and finally the sample is take out from the permeameter in layers for gradation analysis.

33 6) Calculation of hydraulic gradient and flow rate For small permeameter the hydraulic conductivity can be determined as the difference between the constant head height and the height of spacemen placed over the sample length (eq. a Darcy law) thus for big apparatus there are seven piezometers each of them give the water pressure at different depth of the test sample which was manually recorded. Average gradient (iavg) is defined as H/ (p3-p7) and local gradients are defined as icore-1top = (p3-p4)/ (Z3-Z4), icore-2 = (p4-p5)/ (Z4-Z5), icore-3 = (p5-p6)/ (Z5-Z6), and icore-4bottom = (p6-p7)/(z6-z7), similarly for the head loss profiles of the ii = h LL (eeee 8) Where Q=flow kk = QQ (EEEE 9) AAii K=hydraulic conductivity I=hydraulic gradient A=area of spaceman L=length of spaceman H=head The rate of the flow inside the test sample was determined at regular time interval by taking measurement of the volume of water collected from the overflow chute of the lower reservoir with specified period 7) Report results

34 Figure 17 big permeameter apparatus and location of relevant level for gradient analysis

35 5 Result This chapter presents the results of the small and big permeameter suffusion tests. Aiming used for defining internal instability, including Kenney and Lau (1985) proposed a curve matching technique, to evaluate the size of the largest eroded particle and estimated mass loss in which effect of grain size distributions of layers related to an initial gradation is used to evidence internal instability and suffusion. 5.1 A curve matching technique It is a graphical technique proposed by Kenney and Lau (1985, 1986) can be used to investigate and estimate size of larger particles eroded by suffusion process and approximately fraction of materials eroded by the process. The techniques involve extending initial grain size distribution curve of the test sample to matches the grain size distribution curve of the same sample after the test or by extending the bottom layer of grain size distribution curve of the tested sample to matches the grain size distribution curve of each layer of the tested specimen specially the top layers after exhumed sublayer of tested specimens, Fig (). In this thesis, the comparison of curve matching techniques checked by the both. The application the curve matching illustrated by a selection of results and graphs. The complete results are compiled in Appendix1 5.2 Hydraulic gradient for suffusion To analysis suffusion to begin the hydraulic gradient on tested sample, by recording the change in the flow rate and hydraulic gradient across the test sample and to record onset of erosion indicated washout of fine particles from the samples. The temporal variation of hydraulic gradient within the test sample, and the seepage rate through the test sample are plotted big permeameter test found fig (4.4). Thus, refer to the gradient and seepage velocity over core-1(p3-p4), core-2(p4-p5), core- 3(P5-P6), and core-4(p6-p7). The gradual increase in slope of the curve implies the permeability of the sample is increasing until some point of hydraulic gradient when the sample probable starts to initiate. The technique was introduced by Chi Fai Wan and Robin Fell 4.

36 Table 5 Laboratory program schedule Specimen Compaction (relative Modified Proctor max. Dry density) (%) Average gradient (end-of-test) B-S1-9a-f B-S1-9b-f B-S1-9c-f B-S1-85a-f B-S1-85b-f B-S1-85c-f B-S1-8a-f B-S1-8b-f B-S1-8c-f B-S1-8a-F B-S1-8b-F Test duration (hours) 5.3 Small-scale permeameter suffusion studies A total of 9 tests have been performed with 8%,85% and9 % degree of compaction and with two types of filter, from the sample name f1 represent the pores stone whereas f2 represent the wire mesh. Results and graphs each of the specimen take out in layers. The layers are sequenced as follows: I) L 1 (the top most layer) ii) L 2, L3 and L 4 iii) L 5 (Lower layer) onto which the either porous stone or wire mesh filter layer was placed. These subsamples are subsequently compared to layer-1 and layer-5 that has been subjected to testing, layer-1 representative of the suffusion may experience by Compared to layer-5, the diagnosis criterion was that any coarsening of the top(layer-1) transition zone relative the layer-5 proved the existence of loose movable particles B-S1-8a-f1 Test B-S1-8a-f1 ran for 336 hours at an average gradient of 21.7 end-of-test (Table 5) and Compacted to a relative density of 8% of maximum Modified Proctor. The post-test gradation curves show layer-5 more courser than layer-1 which indicates that the erosion happened on the bottom layer also seen the effect clearly in the remaining layers, but a change in the bottom layer was dependent on backward erosion and not to internal instability (Fig 18)

37 mass passing(%) B-S1-8- a-f1-l1 B-S1-8- a-f1-l2 B-S1-8- a-f1-l3 B-S1-8- a-f1-l4 B-S1-8- a-f1-l5,1,1 1 1 Grain size(mm) Figure 18 Specimen B-S1-8a-f1: particle size distributions of the post-test gradations layer 1 to B-S1-85c-f2 Test B-S1-85c-f2 was performed over the length 9.5 hours at an average gradient of 1 (End-of-test) (Table 5) and compacted to a relative density of 85% of maximum Proctor. The posttest gradation curves show the top layer-1 coarse than other layers (Fig 19.), indicating most mass loss in the topmost layer -1 and Applying the curve matching technique (described in section) to the unstable specimen B-S1-85c-f2, as seen in Fig., reveals a mass loss in the range of 2.9% to 5% and a largest eroded particle size of mm based on curve matching techniques. When dismantling the apparatus, a soil open channel formation in the bottom layer was observed. Fig (21).

38 1 Mass passing [%] ,1,1 1 1 Grain Size[mm] 1 9 B-S cf2-L1 7 6 B-S cf2-L5 at 1.5mm erossion loss= /(3.+1)*1=2.9% 1 At 3 mm eerosion loss=,1,1 1 1 B-S1-85-cf2-L1 B-S1-85-cf2-L2 B-S1-85-bf2-L3 B-S1-85-cf2-L4 B-S1-85-cf2-L5 Figure 19 Specimen B-S1-85c-f1: particle size distributions of the post-test gradations layer 1 to5 Mass passing [%] Grain Size[mm] Larger eroded particle mm Figure Specimen B-S1-85c-f1-Curve matching for estimated the fraction of materials loss by suffusion and the largest erodible particles

39 Figure 21 open channel surface post-test in the bottom layer(5) B-S1-9a-f1 Test B-S1-9a-f1 had duration of hours at an average gradient of 21.7 (end-of-test) (Table 5), compacted to a relative density of 9% of maximum Proctor. There is no obvious change in post gradation to each layer (Fig 22).

40 Mass passing [%] ,1,1 1 1 B-S1-9- a-f1-l1 B-S1-9- a-f1-l2 B-S1-9- a-f1-l3 B-S1-9- a-f1-l4 B-S1-9- a-f1-l5 Figure 22 Specimen B-S19a-f1: particle size distributions of the post-test gradations layer 1 to5 5.4 Big-Scale Permeameter Suffusion studies In this permeameter there were two tests have been performed on B-S1 till on the natural soil. Each of specimen release in layers. The layers arranged as follow Figure Layer-1 underlain drainage layer, followed layer-2, layer-3 and finally layer -4 against the filter layer, those sample layer compare to pre-grain size distribution that has not been subjected to test. To evaluate the stability of the sample based on the curve matching technique described in section and the also using a head loss profile, the technique was introduced by Lafleur and Nguyen (7) and hydraulic gradient profile through the sample Specimen B-S1-8a-F1 The specimen compacted to relative density of 8% test of maximum Modified Proctor and ran for 392 hours at an average gradient of 1(end-of-test) which is remarkable to present in dam core. The post-test gradation curves showed obvious shift in distribution from the initial gradation figure (23). The head loss in the top part at beginning test and significant increase in seepage and at the end-of-test, it progressed to uniform shape figure24, Curve matching technique (described in section 5.1) indicates 17.5% to 16.6% mass loss and erosion of particles up to the size of 1.2mm to 4.mm (Fig26).

41 Mass passing [%] ,1,1, Grain size [mm] B-S1-8-b- F1-L1-Post B-S1-8-a- F1-L3-Post B-S1-8-a- F1-L2-Post B-S1-8-a- F1-L4-Post B-S1-8-a- F1-L1-Pre Figure 23 Specimen B-S1-8a-F1: particle size distributions of the initial gradation and post-test layer Top 14 Location [mm] Center Bottom Head loss[%] 3 1 Figure 24 temporal progression of head loss profile across specimen B-S1-8a-F1.not the drainage and the filter part as well