THE COMPRESSIBILITY OF A CRUSHED LIMESTONE BACKFILL · The particle size distribution of the crushed limestone backfill to the culvert at Newport is reproduced from Temporal et al

TRANSPORT AND ROAD RESEARCH LABORATORY Department of Transport

RESEARCH REPORT 251

THE COMPRESSIBILITY OF A CRUSHED LIMESTONE BACKFILL

by K C Brady and G Kirk

The views expressed in this report are not necessarily those of the Department of Transport

Ground Engineering Division Structures Group Transport and Road Research Laboratory Crowthorne, Berkshire, RG11 6AU 1990

ISSN 0266-5247

Ownership of t h e Transport Research Laboratory was transferred from the Department of Transport to a subsidiary of the Transport Research Foundation on I st April 1996.

This report has been reproduced by permission of the Controller of HMSO. Extracts from the text may be reproduced, except for commercial purposes, provided the source is acknowledged.

CONTENTS

Page

Abstract 1

1 Introduction 1

2 Summary of previous test data 1

2.1 Characteristics of the backfill 1

2.2 In situ plate bearing tests 2

2.3 One metre diameter oedometer test 2

2.4 Loading tests on culvert 2

3 Details of programme of laboratory tests 2

3.1 Characteristics of the test material 2

3.2 The 0.5 metre diameter oedometer tests 3

4 .10 Implications for the design of corrugated steel culverts

5 Conclusions

6 Acknowledgements

7 References

Appendix A: The compressibil i ty of backfills

A.1

A.2

A.3

A.4

Data from oedometer tests

Data from plate bearing tests

Data from CBR tests

References for appendix

Page

24

25

25

25

25

25

26

26

27

3.2.1 Apparatus 3

3.2.2 Preparation of specimens and method of test 3

3.3

3.2.3 Results 8

California Bearing Ratio tests 14

4 Discussion 15

4.1

4.2

4.3

Time-dependent strains

Variation in stiffness and CBR value with density

Effect of moisture content

15

15

15

4.4 Changes in grading 18

4.5 Fabric of the oedometer specimens 18

4.6 Relations between stiffness and CBR value 18

4.7 Comparison of the data from plate bearing tests and laboratory tests 21

4.8 Implications for laboratory testing 21

4.9 Comparison of test data with the performance of the culvert at Newport 24

© CROWN COPYRIGHT 1990 Extracts from the text may be reproduced,

except for commercial purposes, provided the source is acknowledged

THE COMPRESSIBILITY OF A CRUSHED LIMESTONE BACKFILL

ABSTRACT

A series of 0.5 m diameter oedometer tests and a series of laboratory California Bearing Ratio tests were undertaken on the crushed l imestone aggregate used as backfil l to a corrugated steel culvert at Newport, Gwent. The method of preparing the specimens was varied to examine the influence of density and moisture content on the compressibi l i ty and CBR value of the backfi l l . The data obtained in this study are compared wi th previously published data obtained from in situ plate bearing tests and a 1.0 m diameter oedometer test; they are also compared wi th data derived from a back-analysis of the performance of the culvert.

The results show that the compressibi l i ty of the aggregate generally decreased with increasing dry density. Saturation of the specimens fol lowing preparation had l i t t le ef fect on their compressibi l i ty, but compacting specimens at moisture contents above the opt imum led to high values of compressibi l i ty being obtained. The CBR value generally increased with increasing initial dry density of the test specimens; but the relation between CBR value and sti f fness varied wi th stress level and initial dry density.

1 INTRODUCTION

The compressibi l i ty of a soil or backfil l is required as input to the design of many geotechnical structures. It is usually required when structural behaviour is strongly dependent upon the interaction between the soil and structure: for example the design of corrugated steel culverts.

The results of a detailed study at Newport, Gwent on the behaviour of a pipe arch culvert under a single axle load has been described by Temporal et al (1985). During that investigation, a series of in situ plate bearing tests was performed on the crushed l imestone backfi l l to the culvert. In addition, a laboratory test was performed on the backfil l using the 1.0 m diameter oedometer at the Building Research Station. The mean secant modulus derived from the results of the plate bearing tests was more than three t imes that obtained from the results of the oedometer test.

The objective of this study was to examine the methods used to determine the st i f fness of coarse grained backfil ls to corrugated steel culverts. Data

are presented from a series of tests undertaken using a 0.5 m diameter oedometer at Middlesex Polytechnic: these tests were undertaken on the 37.5 mm-down fract ion of the backfi l l to the culvert at Newport . The results of these tests are compared wi th those obtained from the plate bearing tests and the 1.0 m diameter oedometer test. The influence that the method of preparation of the specimens had on the results of oedometer tests is discussed, and an explanat ion for the dif ference between the laboratory and field data presented by Temporal et al (1985) is advanced.

Whereas most geotechnical laboratories are not equipped to undertake large-scale oedometer tests they are able to carry out the more routine California Bearing Ratio (CBR) tests. As there would be advantage in being able to est imate the compressib i l i ty of a backfi l l f rom the results of CBR tests, a l imited number of CBR tests were undertaken on the 37.5 mm-down and 20 ram- down fract ions of the backfi l l used at Newport . The data f rom these tests are presented and relations between CBR and compressib i l i ty are examined.

The compressib i l i ty of a soil or aggregate has been described in a number of ways; the terminology and selection of the parameter describing compressib i l i ty or st i f fness vary ing w i th the type of material and method of test . Some of the terms and methods used, and the relat ions between the various parameters are discussed in Appendix A.

2 S U M M A R Y OF PREVIOUS TEST D A T A

2.1 CHARACTERISTICS OF THE BACKFILL

The particle size distr ibut ion of the crushed l imestone backfi l l to the culver t at Newpor t is reproduced f rom Temporal et al (1985) in Figure 1. This f igure shows that about 10 per cent by weight of the aggregate supplied to site had part icles greater than 37.5 ram. The max imum dry densi ty and opt imum moisture content , determined according to Test 14 (vibrat ing hammer): BS 1377 (1975), were 2 .29 Mg/m z and 6.5 per cent respect ively; and the mean in situ dry densi ty was 2.08 Mg/m 3, some 91 per cent of the max imum dry density. The max imum dry densi ty was also determined by compact ing the aggregate to refusal, using a v ibrat ing hammer, in a 0 .34 m

diameter, 0 .47 m high mould; this gave a max imum dry densi ty of 2 .35 Mg/m 3 at a moisture content of 4 .0 per cent.

2.2 IN SITU PLATE BEARING TESTS Four plate bearing tests were undertaken at Newpor t at locations remote f rom the culvert. The tests were performed using a 300 mm diameter plate and a max imum stress of 850 kN/m 3 was applied. Temporal et al (1985) reported that the mean secant constrained modulus (M* ) for the initial loading cycles was 92.2 MN/m 2 for a stress range of 0 to 850 kN/m 2, and the corresponding value for subsequent loading cycles for the same range of stresses was 177.5 MN/m 2. In accordance wi th the Departmental Standard BD12/82 (Department of Transport , 1982) for the design of corrugated steel buried structures, a value of 0 .3 was assumed for the Poisson's ratio of the backf i l l .

2.3 ONE METRE DIAMETER OEDOMETER TEST

Details of the 1.0 m diameter oedometer apparatus at the Building Research Station have been g iven by Penman and Charles (1975). The aggregate was compacted into the oedometer in 100 mm thick layers, each layer being compacted for 1 5 minutes using a Kango vibrat ing hammer. The aggregate was compacted, as del ivered, at a moisture content of 8.5 per cent to an init ial dry densi ty of 2 .23 Mg/m3; and the value of M* over the stress range 0 to 850 kN/m 2 was 30.1 MN/m 2.

2.4 LOADING TESTS ON CULVERT Single axle loads up to 48 .8 tonnes were applied to the culver t that had an init ial depth of cover to the c rown of 1.5m. The depth of cover was reduced incremental ly unti l the culver t fai led at a depth of cover of 0 .36 m, fai lure being def ined by the deve lopment of a permanent Iocalised buckle at the crown. A back-analysis of the fai lure was performed by Temporal et al (1985) using the design method given in BD12/82 (DTp, 1982). A reasonable agreement be tween the calculated and measured performance of the culvert , in terms of both def lect ion and buckl ing behaviour, was obtained using a value of 177.5 MN/m 2 for M * . A l though the value of M * determined f rom such a back-analysis is dependant upon the vagar ies in the method of analysis, the exercise showed that the va lue of M * determined f rom the results of the 1.0 m diameter oedometer test was unreal ist ical ly low.

3 DETAILS OF PROGRAMME OF LABORATORY TESTS

3.1 CHARACTERISTICS OF THE TEST MATERIAL

The mean particle size distr ibutions of the scalped fract ions used for the 0.5 m diameter oedometer tests and the CBR tests are shown in Figure 1.

The dif ference between the mean grading curves of the 37.5 ram-down scalped specimens was due to the inherent variabi l i ty of the crushed l imestone aggregate and segregation of the stockpiles sampled at the source. The samples for the CBR and oedometer tests were taken at di f ferent t imes and after the culvert had been constructed. Checks showed that there was litt le variat ion in the gradings of the specimens used in a particular series of tests. The apparent difference between these grading curves and that obtained from site was due, not only to the scalping out of the larger particles, but also to the different methods of test. The material taken from site was dry graded, according to Test 7B: BS 1377 (1975), whereas the scalped specimens were f irst washed to loosen the fine particles from the aggregate and then graded, according to Test 7A: BS 1377 (1975).

The maximum dry density and opt imum moisture content, determined according to Test 14: BS 1377 (1975), of the 37.5 mm-down fraction used for the 0.5 m oedometer tests were 2.41 Mg/m 3 and 4.0 per cent respectively. The maximum dry density was also determined by compact ing the scalped fraction to refusal in a CBR mould. The material was placed in f ive equal layers in a flooded CBR mould and compacted using a 750 W Kango vibrating hammer; about f ive minutes of vibration was applied to each of the layers. The mean dry density of three such determinat ions was 2.55 Mg/m 3, corresponding to a minimum voids ratio (em~,) of 0.06. The minimum dry density of the scalped fraction was determined by pouring the air-dried aggregate from a height of about 0.5 m into a CBR mould. The mean dry densi ty of six determinations was 1.67 Mg/m3: this corresponds to a maximum voids ratio (emax) of 0.62. The relative density (RD) of the test specimens at a voids ratio (e) is given by,

RD = (emax - e)/(e~ax - e~in) (1)

The 'apparent relative density ' or specif ic gravi ty (G~) of the aggregate particles was 2.7, as determined to BS 812: Part 2 (1975).

100

90

80

6O

so ==

~, 40

30

20

10

0 0001

!

B.S. TEST SIEVES

i i

I

I

0.01

~m

I I

i I

I

0.1

mm

o o ~ o ° ~ ~. ,--,. o ~. , , I

! i i 2 0 m m scalped fract ion . . . used for:

A i ~ [ CBR tests •

I I , / = °

' ' _. .-S.--: i "

2~ ... ;:_~.~ ~ ~ ' ~ , , i

Particle size (mm)

Lq

FINE I MEDIUM I COARSE I FINE I MEDIUM [ COARSE I FINE I MEDIUM I CL)ARSE I ~ C L A Y S I L T S A N D G R A V E L

100

90

80

70

6O

50 ~

E

40 .~

30

20

10

0 100

Fig.1 Particle size distribution of the backfil l to the culvert and the scalped fractions used in the laboratory tests

3.2 THE 0.5 METRE DIAMETER OEDOMETER TESTS

3.2.1 Apparatus Details of the 0.5 m diameter oedometer cell are given in Figure 2. A quadrant of the cell body could be removed to permit visual inspection of the specimen at the end of a compression test. Details of the end platens for the oedometer are shown in Figure 3; the grooves cut into the platens allowed the free entry and drainage of water. To prevent the loss of fine particles, 5 mm thick by 498 mm diameter sintered bronze plates were placed between the end platens and the test specimens. During a compression test the cell body was free to move, or 'float', between the end platens.

Loads were applied to the specimen using a hydraulically driven ram of a 200 kN capacity compression testing machine at Middlesex Polytechnic. A general view of the assembled oedometer installed beneath the machine is shown in Plate 1. The accuracy of the reported load was about _+ 0.4 kN, equivalent to _+ 2 kN/m 2 for the 0.5 m diameter oedometer specimens. As shown in Plate 1, vertical movements were recorded by dial gauges arranged at quarter-points around the

circumference of the specimen. The resolution of the dial gauges was + 0.01 ram.

3 . 2 . 2 P r e p a r a t i o n of s p e c i m e n s and m e t h o d of t es t

A schedule of the compression and grading tests is given in Table 1. Specimens were prepared in loose, medium-dense and dense conditions. The loose specimens were prepared in five equal layers by pouring a known weight of material directly into the oedometer, each layer being roughly smoothed out by hand before pouring the next layer. Apart from test 28, the medium-dense and dense specimens were compacted in five equal layers using a Kango hammer. Each layer of the medium-dense specimens was compacted for one minute whereas each layer of the dense specimens was compacted for five minutes. The thickness of the prepared specimens was about 4 0 0 mm.

Specimens were prepared at moisture content of:

(i) about 0 .5 per cent using air-dried aggregate

(ii) about 4 per cent, close to the optimum moisture content, and

(iii) about 8 per cent.

CL

Vertical stiffeners welded to cylinder

Lifting eyes for cell body on alternate vertical stiffeners

Bolted connection (20mm diam bolts)

(a) Plan

Horizontal stiffeners f " welded to cylinder

~ I 500 int diam _11 1°°1

il " ii Ii - I - II It II

= = 1 ~ - - - = 1 ~ = - - - I ~ = ' II I I It il ~1 il - i - ll II I~

= ~ - - - - ~1 . . . . I1==- II ~1 it - J r II iI Ii Ii ii ii - I -

. . . . l_l _ _ _ _ ~ , . . . . i i___,.

It ii 11 - I ! - I I i l I i

I

I ct

5OO

Cylinder and stiffener in 10mm mild steel plate

(b) Cross-section through A-A

All dimensions in mm Not to scale

F ig .2 Deta i ls o f 0 .5m d iamete r oedometer cell

Drainage grooves

(a) Plan

|A 249

230

160

6

O-ring seal , ~ r ~

Drainage channels I

I I ~ Drainage grooves

10 I 200

(b) Cross-section through B-B

(All dimensions in mm) Not to scale

80

Fig.3 Details of end platens

£_ For lift ing eye

/ (bottom platen)

T • 50

-.,.. ~-' F o r l i f t ing

e y e

C

Plate 1 View of 0.5 m diameter oedometer instal led beneath 200 kN capaci ty compression rig

5

T A B L E 1

Schedule of 0.5 m diameter oedometer tests

Test and Specimen Reference

Number

1 (L) 2 (L) 3 (L)

4 (M) 5 (M) 6 (M)

7 (D) 8 (D) 9 (D)

10 (D)

11 (L) 12 (L) 13 (L)

14 (M) 15 (M) 16 (M)

17 (D) 18 (D) 19 (D)

20 (L)

21 (M)

22 (D) 23 (D)

24 (D)

25 (D)

26 (D)

27 (D)

28 (D)

Initial Moisture Content

(per cent)

4.0 3.1 4.3

4.2 4.9 4.6

4.0 4.1 5.7 4.9

0.7 0.1 0.8

0.5 0.5 0.8

1.1 1.0 0.7

8.7

6.4

8.9 6.1

7.5

8.8

1.8

3.7

Initial Dry

Density (Mg/m 3

1.54 1.68 1.62

2.07 2.05 2.13

2.34 2.29 2.39 2.39

1.77 1.80 1.84

2.13 2.18 2.18

2.27 2.29 2.29

1 .88

2.27

2.31 2.26

2.35

2.31

1.82

2.19

Relative Density

0 0.02

0

0.56 0.54 0.63

0.83 0.79 0.88 0.88

0.17 0.21 0.27

0.63 0.68 0.68

0.77 0.79 0.79

0.33

0.77

0.81 0.76

O.84

0.81

0.8

0.8

0.8

Unsaturated

2.8 2.07

Compression Test

Saturated

/

Grading Tests Remarks

Compression tests done on specimens at or around optimum moisture content

Compression tests done on air-dried aggregate

Compression tests done on specimens at moisture contents above optimum.

grading slightly coarser than mean grading of 1 to 23.

grading slightly finer than mean grading of 1 to 23.

25 mm to 12.5 mm aggregate size.

12.5 mm-down aggregate size.

Alternative layers of 25 mm to 12.5 mm-down fractions.

KEY: L = Loose specimen M = Medium-dense specimen D = Dense specimen

6

Three of the specimens were not tested in the oedometer, but following their preparation a quadrant of the oedometer was removed and the fabric examined and photographed; a sample was then taken to determine the particle size distribution according to Test 7A: BS 1377 (1975).

After compaction into the oedometer cell, four specimens were saturated by allowing water to enter under a low hydraulic gradient through the bottom of the specimen.

To investigate the effect of small changes in grading on compressibility, specimens 24 and 25 were prepared slightly coarser and slightly finer respectively than the mean grading of specimens 1 to 23. The particle size distributions of the specimens before compaction and compression testing are given in Figure 4. To examine the effects of particle size and layering on compressibility, the aggregate was graded into two fractions: one was 25 mm to 12.5 mm and the other was 12.5 mm-down. A compression test was carried out on each fraction and also on a specimen made up of three alternate layers of each fraction. The particle size distributions of

these specimens before compaction and compression testing are also given in Figure 4.

The specimens were usually loaded in f ive increments to a maximum stress of about 820 kN/m 2 with two cycles of loading and unloading generally applied in each test. The top platen assembly imposed an initial surcharge of 7.5 kN/m 2 to the specimens. The compression generated by this surcharge was not measured and the quoted stresses do not include this initial surcharge stress. Also, the strains generated by saturating the specimens were not recorded. The compression of the specimens due to a change in load was recorded from the t ime of application and a steady state was usually achieved within a few minutes of the change in load; consequently most of the tests were completed within a working day. A few longer-term tests were undertaken to investigate whether any time- dependent strains developed under sustained loading.

Following compression testing, a quadrant of the oedometer cell was removed and the structure of the specimens examined and photographed. Samples were then taken as required from the centre of the specimens to determine the particle size distribution.

10C

90

80

70

60 == 5 § 50 E

~_ 40

30

20

10

0 O001 0.01

B.S. TEST SIEVES ~ ~ ~ oo ~ ~ co

01

Particle size (ram)

1 10

mm

m ¢o Ln Lo ~ ~

100

90

80

70

6O

5 50 ~-

E

40

30

20

10

0 100

FINE ] MEDIUM J COARSE I FINE I MEDIUM I COARSE I FINE J MEDIUM C L A Y Sl LT S A N D G R A V E L

J COARSE

Fig.4 Particle size distribution of specimens used in the O.5m diameter oedometer tests

3 . 2 . 3 R e s u l t s

In some of the tests the compressions recorded by the four dial gauges varied wide ly , part icular ly at low stress levels during the f i rs t compression cycle. The data show that the top platen t i l ted during the init ial appl icat ion of the load, the degree of t i l t ing was not changed much by subsequent loading. The var iat ion in the recorded movements , in absolute or percentage terms, was least w i th the loose specimens. At the max imum stress of around 820 kN/m 2, the strains recorded by mos t of the dial gauges were wi th in 10 per cent of the mean strain, but deviat ions of more than 20 per cent were recorded in some tests on the denser specimens.

Most of the specimens gave a rapid response to the changes in applied load and a steady state was achieved wi th in a f ew minutes. Insigni f icant t ime-dependent strains were recorded in most of the long-term tests , and all the specimens gave an immedia te response to unloading. But increases in

strain were recorded for some t ime after an increase in load during the f i rst loading cycle on some of the dense specimens compacted at moisture contents above the opt imum. About an hour was required before a steady state was achieved during the initial loading cycle on tests 22 and 25. Typical relations between vert ical strain and t ime for the initial loading cycle on test 22 are shown in Figure 5.

The typical relations between applied vert ical stress and mean strain in Figure 6, show that the stress-strain relations were non-linear, part icularly for the initial loading cycle. Values of the constrained modulus (M*) for the loading and unloading cycles are given in Table 2 and relations between M* and the initial dry density are given in Figure 7. The data from the 1.0 m oedometer test and the in situ plate bearing tests are also shown in Table 2 and Figure 7. Data from the particle size analyses of the test specimens are given in Table 3.

1.0 - E

E

0 E

2.0 - >

3.0 0

I 0

- range of stresses

410 to 820 kN/m 2

I I I I I ! 1 10 25 50 100 150

Elapsed t ime (minutes)

I I ! 1 2 3 4 5 6 7 8 9 10 11 12

Square root of t ime (minutes) ½

13

F ig .5 Rela t ions b e t w e e n vert ical m o v e m e n t and t i m e fo r 0 . b m d i a m e t e r o e d o m e t e r test 22

900

800

700

600

E z

500 =_

400

0. .

3OO

200

100

F

0.005 0.01 0.015

Vertical strain (per cent)

/ Jl0

I 0.02

Fig.6 Typical relat ions between appl ied vert ical stress and vert ical strain

9

100

60

30

A CN

E z ~ 1 0 -

6 i

3 -

1 1.5

3 0

i 1.6

(A) Initial loading cycle, nominal stress range 25 to 100 kN/m 2

26 O Field 2 O Field 3 O Field 1

120

11 • ,,***'"°°° °° 130

O Field 4

6 ,.7 . . . . .

5 ~ . . o ° " ~ 0 15 _ • 19 " " I , ° ° ° . ° • 16 ~21

o°°.°°.o°°Oo,,,.oO-°°* 14 • • 22 23 • 25 13 BRE

. o o *

2 0

• 2 0

Key: • 0.5m diam oedometer test, see Table 1 [] 1.0m diam oedometer test, BRE O In situ plate bearing test

I I I I 1.7 1.8 1.9 2.0

Initial dry density (Mg/m 3)

100

60

30

10

- 3

I I I 1 2.1 2.2 2.3 2.4

E z

Fig.7 Re la t ions between M* and in i t ia l d ry dens i ty

10

100

A ¢N

E z

_io

1 1.5

(B) Init ial loading cycle, nominal stress range 25 to 250 kN/m

26 O Field 2 O Field 4 O Field 3

3 0

5 0 14 0 . . - ' 1 6

e • o *

tl, t l,

ql,~, ~ql,

f * 12 .°°" • o o *

o* " o* 110 ..

o° 20 ~* •

• "* 13

**o

.@ @

/

2 0

O Field 1 24 9 • • 27 . . . . • 1 5 / / . . . , . , , u . . " 28, ....-"1% Io •

0 - - " • 1 9

21

23 • 25 • 22

[] BRE

Key: • 0.5m diam oedometer test, see Table 1 [] 1.0m diam oedorneter test, BRE O In situ plate bearing test

I I

1.6 1.7

I I I

1.8 1.9 2.0

Initial dry density (Mg/m 3)

I I I

2.1 2.2 2.3 2.4

- 1 0 0

A

E z

10

Fig.7 C o n t i n u e d

11

1000

100

A

E z

10

.5

(C) Nominal stress range 205 to 820 kN/m 2

27 26 • 10 • 28 23 8 •

• o o o . . . . . . . . "-" Second loading cycle 6 _ . .17 ' " -~=

\ . ; ....... . 11 ~ . \ . . . . . . . . . . . . . • 14 • • 22 • 21 m l l l l l . o I I-wm

2 . . . . . . . . . . . • 13 • Field 2 3 • . . . . . 12 • • F ie ld3

20 • Field 4 • Field 1 9

0 27 8 1~0 0 0 ,°o°°" 26 [] Field 2

2801:3= O Field 4 23 ,~.19

Field 3 [] 6 t~_ .*~Or~ 0 Field 1 [] ~ oO*°. 17"" 1 8 2 4

o 1 4 o o 5 °o.°° 15 0 25

o°o° • . ° - O 22 init ial loading cycle o°O°

\ o.° • oO°O°°° BRE

~ ° o * ° °

12 o-'°°" 11 O ,_O~°°__ 20

°°.o°°°°°°° 130 0

,o..;;. @°

Key: 1st load cycle 0.5m diam oedometer test, see Table 1 O 1.0mdiamoedometer test, BRE • In situ plate bearing test []

2nd load cycle

I 1.6

! I I I I I 1.7 1.8 1.9 2.0 2.1 2.2

Initial dry density (Mg/m 3)

I 2.3

1000

100

E z

10

1 2.4

F ig .7 C o n t i n u e d

12

O O O ~ O ~ O ~ O O O ~ O ~ O ~

0

0

0

0

I 0 ~N

0 ~N

0

0 CN

I 0

A ,

o

z -8 o ~

cO

m O O cN

~n ~- I o

¢.- , _ a ~ m

' - 0 I

°

I

9 N o

o ~N

I

0

N ' • "~-- I.~ _~ o

O4 I

O O

I IX) ¢',1

LO C'4

I

O

~ o ~ o ~ I I ~ I

o o o o o o o I I o O O O O O O ~ O O ~ O ~

I O O O O IX) O)

O O O O O O O O O O O O O O O O O O

~ O ~ O O O ~ O O O I . f ) L o r ' - ,~'UO

~ ~ O O ~

O ~ O ~ O ~ O ~ O O LOI.~

O O ~ O O O O ~ O O ~ ~ ~

m m o m m m o I I m I O U O O I " - L~LO

O ~ O ~ O ~ O O ~ ~ O ~

o o o o o o o I I o I I o o O O O O O O O O O O ~ ~ O ~ ~ ~ O

O O O O ~ , - -

I o o o o o o O O O O O O

~ ~ ~ ~ O O O ~ ~ O ~ O ~ ~ ~ ~ ~ ~ O ~ ~ ~ ~ O ~ ~ ~ ~

O O ~ O

O O O O 9 1 / ) 1 / )

~ O ~

~ O

~ O ~ ~ O ~

~ ~ 0 ~ ~ 0 ~ ~ ~ ~ 0 0 ~ ~ ~ 0 ~ 0 ~ O ~ ~ ~ ~ O ~ ~ ~ ~ ~ ~ ~

d ~ ~ o o ~ ~ o ~ ~ ~ o o ~ ~ ~ o ~ ~ ~ O ~ ~ ~ ~ ~ ~ ~ ! ~ ~ O

~ O ~ ~ O ~ ~ ~ ~ O ~ ~ ~ ~ O ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ O

* I

,-- , - - ¢ 4 , - - , - ,-- O') O~l

A A A A

O ~ O ~

~ ~ 0 ~ ~ ~ 0 ~ ~ ~ ~

~ a D ~

W W W W U- I/. I/. I.I.

6 . I 4-,

o E 09 .-- (n O

a- O0

4--

O °~

> ' ~ °~

( D

( D 09

13

TABLE 3

Particle size distribution tests on oedometer specimens

IJescnpt~on oT specimen ana [es~ reference number

Prior to compaction and compression test (Mean of four tests)

Following compaction, no compression test (mean of two tests on each specimen

Following compression test (mean of two tests on each specimen)

1 (L) 4 (M) 7 (D)

2 (L) 3 (L) 5 (M) 6 (M) 8 (D) 9 (D)

13 (L) 16 (M) 17 (D)

20 21 22 23

24 25

26 27 28

25 mm

91

I 96 (+5) 94 (+3) 91 (0)

I 96 (+5) 92 (+1) 95 (+4) 95 (+4) 95 (+4) 93 (+2)

93 (+2) 94 (+3) 96 (+5)

(L) 94 ( + 3) (M) 89 ( - 2) (D) 93 ( + 2) (D) 88 ( - 3)

(D)** 91 (+3) (D) * * 93 (0)

(D)** (D)** (D) * *

Percentage passing sieve size (change in per cent passing*)

12.5 mm

63

75 (+12) 70 (+7) 69 (+6)

74 (+11) 66 (+3) 71 (+8) 72 (+9) 75 (+12) 71 (+8)

66 (+3) 70 (+7) 68 (+5)

73 (+10) 51 ( - 1 2 ) 68 (+5) 54 ( - 9)

58 (0) 69 (+2)

16 (+16)

68 (0)

1.00 mm

16

20 (+4) 18 (+2) 20 (+4)

22 (+6) 21 (+5) 18 (+2) 21 (+5) 21 (+5) 22 (+6)

19 (+3) 20 (+4) 17 (+1)

22 (+6) 15 ( - 1 ) 24 (+8) 13 ( - 3 )

13 (+1) 19 (+1)

2 (+2) 28 (+2) 16 (0)

0.2 mm

10

12 (+2) 12 (+2) 13 (+3)

14 (+4) 14 (+4) 12 (+2) 16 (+6) 13 (+3) 15 (+5)

11 (+1) 12 (+2) 11 (+1)

16 (+6) 14 (+4) 20 (+10) 12 (+2)

11 (+3) 14 (+3)

2 (+2) 19 (+3) 12 (0)

*Dif ference between mean grading of bulk sample. **Di f ference from the results of grading tests performed on subsamples taken prior to compaction. L =Loose specimen M = Medium-dense specimen D = Dense specimen

3 . 3 C A L I F O R N I A B E A R I N G R A T I O T E S T S

The CBR tests were performed on both 37.5 mm- down and 20 mm-down scalped fractions of the original backfill. Test specimens were prepared at nominal moisture contents of 4 and 8 per cent.

It was intended to test specimens at three dif ferent relative densities: dense and medium- dense specimens were prepared by compacting the material using a vibrating hammer, and loose specimens were prepared by pouring the material direct ly into the mould. With the densest specimens, each of the three layers was compacted for a minute, whereas for the medium- dense specimens the layers were vibrated for 30

seconds. In the event there was little difference between the densities achieved with the different times of vibration.

It proved diff icult to obtain a smooth testing surface to the CBR specimens: even with the compacted specimens there were a few loose particles at the surface. Consequently the initial part of some of the force/penetration curves were concave upwards and required correction. It also meant that the densities of the specimens could not be determined to any great accuracy. With the wetter specimens, there was a tendency for some of the finer particles to adhere to the base of the vibrating plate, and also to be washed out with the excess water draining from them.

14

The results of the CBR tests are given in Table 4, and the relation between initial dry density and CBR value is shown in Figure 8.

After each test, a sub-sample was taken from the centre of the CBR specimen and the particle size determined in accordance with Test 7A: BS 1377 (1975). A comparison of the results of these and the initial grading tests is given in Table 5.

4 D I S C U S S I O N

4.1 TIME-DEPENDENT STRAINS The data obtained from oedometer tests 22 and 25, which were compacted at moisture contents above optimum, showed time-dependent strains persisting over an hour or so. Thus pore water pressures were sustained for some time within these test specimens.

Oedometer tests 9 and 10 were conducted on specimens having similar dry densities as tests 22 and 25, but they were compacted at moisture contents at or around optimum and subsequently saturated. The results from tests 9 and 10 showed an "immediate' response to loading, indicating that the time-dependent strains recorded in tests 22 and 25 were a function of the method of preparation of the specimens.

4 .2 VARIATION IN STIFFNESS AND CBR VALUE WITH DENSITY

The relations given in Figure 7 show that the sti f fness of the aggregate generally increased wi th the initial dry density of the specimens: this dependency was retained to a lesser degree for the second loading cycle and the unloading cycles.

In v iew of the similar methods of preparation, the relative densities (RD) of oedometer specimens 26, 27 and 28 should have been about the same as specimens 17, 18 and 19 at around 0.8. The values of M* obtained from tests 27 and 28 were broadly similar to those obtained in tests 17, 18 and 19, but the data from tests 26, 27 and 28 do not f i t in particularly well with the general trend given in Figure 7. This suggests that relative density rather than initial dry density is a better determinant of st i f fness and that the grading and particle size are not significant influences. The data given in Figure 8 show that the CBR value increased with increasing dry density of the test specimens, but there was no clear difference between the results obtained from the 20 ram- down and 37.5 ram-down scalped fractions.

4.3 EFFECT OF MOISTURE CONTENT The effects of saturation on compressibi l i ty can be gauged by comparing the data obtained from oedometer test 2 with test 3, test 5 wi th 6, and

T A B L E 4

Results of CBR tests

Test and Specimen Reference Number

1 (L) 2 (M) 3 (D)

4 (L) 5 (M) 6 (D)

7 (L) 8 (M) 9 (D)

10 (L) 11 (M) 12 (D)

Maximum particle

size (mm)

37.5 37.5 37.5

37.5 37.5 37.5

20 20 20

20 20 20

Nominal moisture content

(per cent)

' 8 8 8

4 4 4

Actual moisture content

(per cent)

3.5 3.5 3.4

7.6 3.6 6.9

3.6 3.4 3.6

5.3 6.0 7.7

Dry Density (Mg/m 3)

1.68 2.04 1.94

1.63 1.94 1.97

1.54 2.08 2.11

1 . 5 8 2.02 2.12

CBR (per cent)

Top

23 70

110*

10 85* 60*

4 100

85*

11 55*

120"

Base

12 90

105

16 100" 110

5 120 150

7 100 140

D = Dense specimen M = Medium-dense specimen L = Loose specimen

* =Displacement of more than 1 mm required to correct the position of the origin of the load-penetration curve.

15

160

140

120

~1 O0 E

c~

60

40

20

0 ! 1.4 1.5

20mm - down scalped specimens

~ . ~ 3p7ecSmmmen- s down scalped

# o

o*

.o

I I I I 1.6 1.7 1.8 1.9

Initial dry density (Mg/m 3)

I I I ,J 2.0 2.1 2.2 2.3

Fig.8 Relat ion between CBR value and initial dry density

16

TABLE 5

Particle size distribution tests on CBR specimens

Description and test reference number

37.5 mm-down scalped specimens 20 mm

Before testing (mean of 2 tests) 72

1 (L) 74 (+ 2) Following preparation and CBR test

Percentage passing sieve size (change in percentage passing)

10 mm 1.0 mm

34 9

33 ( - 1 ) 8 ( - 1 )

2 (M) 73 (+1) 37 (+3 ) 10 (+1 ) I

3 (D) 72 (0) 31 ( - 3 ) 7 ( - 2 )

4(L) 59 ( - 1 3 ) 21 ( - 1 3 ) 5 ( - 4 )

20 ram-down scalped specimens

Before testing (mean of 2 tests)

Following preparation and CBR test

5 (M) 82 (+ 10)

68 ( - 4)

10 mm

49

6 (D)

7 (L) 49 (0)

8 (M) 52 (+ 3)

9 (D) 55 (+ 6)

40 ( - 9)

54 (+5)

50 (+1)

10 (L)

11 (M)

12 (D)

36 (+2 ) 5 ( - 4 )

27 ( - 7 ) 6 ( - 3 )

5 mm 0.5 mm

28 12

21 ( - 7 ) 9 ( - 3 )

26 ( - 2 ) 11 ( - 1 )

29 (+1 ) 12 (0)

16 ( - 1 2 ) 6 ( - 6 )

25 ( - 3 ) 8 ( - 4 )

24 ( - 4 ) 10 ( - 2 )

test 8 with 9 and 10. The results from tests 2 and 3 are similar, and after making allowance for the differences in density, the data from tests 5 and 6 are in reasonable agreement, as are the results from tests 8, 9 and 10. Terzaghi (1960) stated that the compressive strength of rocks was reduced by saturation, but it did not appea r from these tests to lead to a reduction in stiffness by increasing the breakdown of particles.

The results in Figure 7 show that in general the values of M* obtained from oedometer tests 20 to 25 are somewhat lower than the values recorded in tests 2 to 19 for the same initial dry densities. This is particularly true for the results of tests 22 and 25 which had the highest moisture contents. The high moisture content of these two specimens may have led to high pore water pressures being generated by the vibration applied during compaction. In turn, this could have caused fine material to migrate towards the top of a layer leaving a coarser fraction at the bottom. Under compression, the coarser particles could then penetrate into the loose fines-rich zone at the top

of the underlying layer. Such a phenomenon led Terzaghi (1960) to state that it was preferable to construct low height rockfill dams in a single operation rather than in multiple l ifts. The thickness of a fine-grained layer would reduce with consolidation and the presence of it would have less effect wi th increasing stress. Such a layer would also have little ef fect on the strains developed during unloading and reloading cycles. This type of mechanism is consistent with the data given in Table 2.

There was a wide variation in the CBR values from the top and bottom of the specimens, but no extremely low CBR values were recorded in tests on dense specimens prepared at moisture contents above the optimum. However many of the load- penetration curves derived from tests on the top of the specimens were init ially concave upwards, and were corrected according to the procedure laid down in BS 1377: 1975. The correction procedure would have reduced the effects of the presence of a loose or fines-rich layer at the top of the specimens. Thus there is some evidence from

17

the load-penetration curves to suggest that there was some movement of the fines within the denser CBR specimens.

4.4 CHANGES IN GRADING The data given in Table 3 show that the changes in grading caused by preparation and test ing of the oedometer specimens were modest. The data given in Table 5 show that there was some breakdown of the larger particles of the CBR specimens, and also an apparent loss of f ines particularly from the loose specimens prepared at a nominal moisture content of 8 per cent. The wider variabi l i ty of the data obtained for the dense oedometer and CBR specimens compacted at moisture contents above the opt imum was probably due to segregation of the aggregate, which made it more dif f icult to obtain a representative sample for the grading tests.

4.5 FABRIC OF THE OEDOMETER SPECIMENS

The fabric of most of the oedometer specimens was examined and photographed after preparation or compression testing. All the loose specimens had a uniform appearance and it was di f f icul t to discern any structure in the medium-dense specimens, although in a few cases the interface between the individual layers of the specimen could be identified. It was possible to distinguish the various layers of the dense specimens prepared at low moisture contents. As shown in Plate 2, the interface between the layers were

characterised by a concentrat ion of f ine part ic les around the c i rcumference of the specimens.

A thin layer of f ines was also observed around the dense specimens prepared at high moisture contents. Plate 3 shows a v iew of test specimen 22 immediate ly fo l low ing the removal of the quadrant of the oedometer cell; some f ines adhered to the removed quadrant. Plate 4 shows a close-up v iew of tes t specimen 22 after the face, had been washed and l ight ly brushed to remove the thin layer of f ines, which was about 1 mm thick. It was d i f f icu l t to ident i fy the indiv idual layers of the dense specimens af ter the f ines had been cleaned f rom the exposed face, but it did seem that the grading of some of the specimens varied w i th posit ion.

Plates 5 and 6 show v iews of test specimen 28 which was composed of al ternate layers of 25 mm to 12.5 ram, and 12.5 ram-down f ract ions of the aggregate. A l though there was some mix ing of the f ract ions at the interfaces, the f ines did not appear to have penetrated far into the layers of the coarser fract ion. The values of M* determined for this specimen were not that dissimi lar to those obtained for the const i tuent f ract ions.

4.6 RELATIONS BETWEEN STIFFNESS A N D CBR VALUE

The relat ions between the init ial dry densi ty and the values of M * , for the f i rs t loading cycle over the stress range of 0 to 820 kN/m 2, and the CBR values are shown in Figure 9. Values of M * and CBR for part icular init ial dry densit ies have been derived f rom relat ions such as shown in Figure 9 and are given in Table 6. For specimens having initial dry densit ies greater than 1.75 Mg/m 3, the

Plate 2 View of dense oedometer specimen 17 prepared with air-dried aggregate

18

Plate 3 V iew of dense oedometer spec imen 22 prepared at a moisture content above op t imum

Plate 4 Close-up v iew of dense oedometer spec imen 22 af ter face had been washed and l ight ly brushed

19

Plate 5 View of layered oedometer specimen 28

Plate 6 Close-up view of layered oedometer specimen 28

20

following empirical relations can be derived for the first loading cycle:

CBR= 5M*, for stress range 25 to 205 kN/m 2 (2a)

CBR=2.5M*, for stress range 0 to 820 kN/m 2 (2b)

CBR= 2M*, for stress range 205 to 820 kN/m2

where the CBR value is quoted in per cent and M* is measured in MN/m 2.

A theoretical approach given in Appendix A, based on the use of elasticity equations, produced the following relation;

CBR = M* (2d)

for a penetration of 5 mm, M* again being measured in MN/m2.

An empirical relation presented by Croney (1977) suggests that an in situ CBR value of 100 per cent is approximately equal to an M* value of 183 MN/m2, ie

CBR=0.55M* (2e)

As the stress range imposed in a CBR test varies with the measured CBR value, and also because aggregates are not linear elastic materials, a simple relation between CBR and M* that covers a wide range of dry densities and applied stress cannot be expected. Nonetheless, empirical relations such as those given above may be of some value at the design stage of a project.

4.7 COMPARISON OF THE DATA FROM PLATE BEARING TESTS AND LABORATORY TESTS

Values of M* were derived from the load- penetration curves of the CBR tests for a stress range of 0 to 820 kN/m 2 using equation (6) given in Appendix A1 and an assumed value of 0.3 for Poisson's ratio. A stress of 820 kN/m 2 was recorded at penetrations of less than 0.2 mm with the dense specimens, consequently the highest values of M* are not particularly accurate. Values of M* were derived from both the measured and corrected load-penetration curves. For comparative purposes, these values of M* were divided by a factor of 10 to account for scale effects; this factor was inferred from Figure A1. Similarly values of M* were derived from the plate bearing tests, these values were divided by a factor of 2.2, again taken from Figure A1, to account for scale effects. A comparison between these values of M* and those derived from the oedometer tests is given in Figure 10. Given the assumptions and limitations inherent in the theoretical approach, and also the variation in the test data, there is a reasonably consistent trend in the results obtained

from the medium-dense specimens. The good agreement between the field and laboratory data may be fortuitous; but the data given in Figure 10 confirm that the values of M* derived from the 1 .O m diameter oedometer test undertaken at BRE, and also from the 0.5 m diameter oedometer tests numbered 22 and 25 undertaken at Middlesex Polytechnic, were about half that anticipated from the general trend.

The mean values of M* determined for the unloading and reloading cycles of the plate bearing tests, over the stress range O to 820 kN/m 2 were about 200 MN/m 2 and 185MN/m 2 respectively. The corresponding values determined from the oedometer tests undertaken on medium dense specimens were about 450 MN/m2 and 300 MN/m 2. Corrections to account for the effects of scale in the plate bearing tests would widen the difference between the two sets of values. The discrepancy is probably associated wi th the different stress paths fol lowed in the two types of test: high radial stresses locked into the oedometer specimens from the f irst loading cycle could substantial ly increase the st i f fness recorded in subsequent unloading and loading cycles, and the stress distribution beneath a loaded plate may vary with previous loading history. This phenomenon warrants further analytical and experimental investigation.

Although the photographic evidence of fines-rich layers within the oedometer specimens compacted at moisture contents greater than the optimum is inconclusive, it should be borne in mind that only a thin layer is necessary to give a large reduction in M*. For example, the values of M* for the f irst loading cycle in tests 22 and 25, over the stress range from 25 to 100 kN/m 2, were about half those expected from the general trend of data given in Figure 9. The measured and expected values of M* of 14 and 28 MN/m2 correspond to strains of 0.54 and 0.27 per cent respectively. For a 400 mm high specimen, a strain of 0.27 per cent corresponds to a settlement of just over 1 mm and a reduction in st i f fness by a factor of two would only require an additional sett lement of about 0.2 mm at each of the interfaces between the layers of the specimen and at the top platen. Such a small sett lement could occur in a thin interface layer and would be dif f icult to detect by relatively crude photographic observations.

4 .8 IMPLICATIONS FOR LABORATORY TESTING

The data given in Figure 9 show that compressibi l i ty and CBR value are functions of the initial dry density of the aggregate. In addition, Brady et al (1983) and Bolton (1986) have also shown that the peak angle of fr ict ion is dependent upon dry density. Thus for comparative purposes

21

100 -

z

1 1.5

9 ~ r 1 2 Field 2 ~ IL.--

~ i 1-"'1 t_ OF~eld3 8 . . . . T . ' I 1111 / z ~ .=, w . - ' " 10-

..- ' / 1 I Field 1 • 6 o1711D • 19 • 24 ..-'" ! / ' 1 4 o.O;°-*" • 18

. . ~11 • I - - . * I 1 ~ • *" - - I I • 5 ~ ° • 23 •21 .o .J. 15

. . I I ° . o " I - - 25 °* I .L ° ° I

°°° I . , , , P ° I • 22 o ° °

°°°° °°ooJll" .i. 13 B'R E o o °°*

°° °*

1 * 12 °°°° 1/~ 1 1 0 ° ° °°°

A - .* / • .-° ~ l : / .-" I I 20

° ° ° v . . ,

/1 | .." 13 ( ~ ' ~ O e d ° m e t e r test data f°r first I°adlng 2 10 ~ I I 7 °°°° cycle over stress range 0 to 820kN/m

J 3 - o . "

" * ' 1 0 2

,i Key: • 0.5 diem oedometer test, see Table 1 13 1.0m diem oedometer test, BRE

v& CBR test, seeITable 4, corrected data points • CBR test, see Table 4, uncorrected data points O In situ plate bearing test

I 1.6

I I I I I I I 1.7 1.8 1.9 2.0 2.1 2.2 2.3

Initial dry density (Mg/m 3)

100

1 2.4

A +~ E

Q.

lO ~ t ~

Fig.9 Relations between M * and CBR with initial dry density

T A B L E 6

Comparison of CBR and M* values

INITIAL DRY

DENSITY

(Mg/m 3)

1.55 1.65 1.75 1.85 1.95 2.05 2.15

CBR VALUE (from

Figure 9)

(1)

7 17 32 55 80

105 130

(2) 25 to

205 kN/m 2

2 4.5 7.5

11 15.5 21 26

M* VALUES (MN/m 2)

1st loading cycle

(3) 0 to

820 kN/m 2

5.3

(4) 205 to

820 kN/m 2

10

2nd loading cycle

(5) 205 to

820 kN/m 2

Ratio of CBR/M*

1/2 1/3 1/4 1/5

9 14 21 29 41 53

14 19 26 36 50 70

225 250 290 320 360 410 450

3.5 1.3 0.7 0.03 3.8 1.9 1.2 0.07 4.3 2.3 1.7 0.11 5.0 2.6 2.1 0.17 5.2 2.8 2.2 0.22 5.0 2.6 2.1 0.26 5.0 2.5 1.9 0.2£

22

100

10 -

E z

1.0 - 10 ]

0.1 1.5

3 0

2 e

411[

1.6

110

1.7

12

130

20

I

5

3; I I 16 I I I I

i

8 9= ] 100 • 10 :

1 7 0 0 1 9 • 8~bField 2 : 16 2300 O18 24

5 0 r i 1 4 F i e l d 4 ? 112 15 21 ~ 2 5 Field 3 ( ~ Field 1 ~1~ 22

11

I n I BRE

II I2 II II II II II i I ! Stress range 0 to 820 kN/m 2

I III llill' ° r t ° 8 5 0 k N / m 2

i

Key: • 0.Sm diam oedometer test, seeTable 1 I:3 1.0m diam oedorneter test, BRE

• 'A CBR test, see,Table 4, corrected data points • CBR test, see'Fable 4, uncorrected data points O In situ plate bearing test

I I I I I I 1.8 1.9 2.0 2.1 2.2 2.3

Initial dry density (Mg/m 3)

10

E z

1.0

0.1 2.4

Fig.lO Relations between M* and init ial d ry density

23

it is imperative to use the same initial dry density in all laboratory tests, and wherever possible this should replicate the in situ dry density. Relative density is probably more appropriate when comparing test results obtained from dif ferent soils or graded fractions of the same soil.

The experience gained from this study shows that little if any increase in dry density is achieved by using the Kango hammer for more than about two minutes to compact 100 mm thick layers in a 0 .5 m diameter cell, or for more than about 30 seconds to compact 50 mm thick layers in a 1 52 mm diameter CBR mould. Any extra increase in dry density for longer periods of vibration is accompanied by the production of fine-grained particles. The overall change in grading may be small in terms of the percentage passing a particular sieve size, but this does not give a good picture of the large increase in the number of fine particles. The unnecessary production of fines during preparation of the test specimens should be avoided as this can change the property that is being assessed. The availability of excess water may enable the fines to migrate to the boundaries of the specimens and substantially modify the inherent properties. As granular materials are rarely compacted at moisture contents above opt imum, it would seem unnecessary to prepare specimens in such a manner for routine testing.

4 .9 COMPARISON OF TEST DATA WITH THE PERFORMANCE OF THE CULVERT AT NEWPORT

An examination of the data provided by Temporal et al (1985) indicates that a good fit between the predicted and measured increase in horizontal span of the culvert required, a M* value ranging from about 100 MN/m 2 at a depth of cover to the c rown of the culvert of 1.5 metres, to a value of about 200 MN/m 2 at a cover depth of 0 .5 metres.

It could be argued that the backfill around the culvert would be loaded during its compaction and so the modulus determined from reloading cycles would be appropriate for back-analysis purposes. This is supported by the reasonable agreement between the calculated and measured performance of the culvert obtained by Temporal et al (1985) when using a M* value of 177.5 MN/m2 obtained from the reloading cycles of the plate bearing tests; but the effects of scale were not taken into account when selecting this value.

The in situ behaviour of the culvert under load falls between, a low value of aggregate stif fness of about 85 MN/m 2, based on the scale-adjusted values of M * determined from the reloading cycles of the plate bearing test, and a high value of about 300 MN/m 2, determined from the reloading cycles of the oedometer tests. It could be

concluded that the restraint on the movement of soil in the field lay between that imposed in the plate bearing and oedometer tests. However the fol lowing points should be noted:

(i) The magnitude of the stresses acting in the backfill adjacent to the culvert would have varied with position, both during construction and throughout loading with the HB axle.

(ii) Similarly, the direction of the principal stresses would have also varied with position and loading, and the stress path imposed in the plate bearing and oedometer tests would not necessarily be followed in the field.

(iii) The loading imposed by the wheels of the HB axle was theoretically equivalent to a contact stress of 1.1 MN/m 2 acting over a square area of side 0.32 m. The maximum stresses imposed in plate bearing and oedometer tests were about 30 per cent lower than the tyre pressures, but the data given in Table 2 indicate that this would have had a small influence on the determined values of M*. More importantly, the 300 mm diameter plate was not much smaller than the equivalent area stressed by the tyres, and it may not therefore be appropriate to factor down the values of M* determined from the bearing tests when back-analysing the performance of the culvert.

The back-analysis of the performance of the culvert is clearly not straightforward, and it is too optimistic to expect the complex soil-structure behaviour to be described by a single value of stiffness. Nonetheless, a simplistic approach, such as given in BD12/82 (DTp, 1982), is required for the design of such structures. Temporal et al (1985) concluded that BD 12/82 (DTp, 1982) produced safe, if somewhat conservative designs for corrugated steel culverts.

4 .10 IMPLICATIONS FOR THE DESIGN OF CORRUGATED STEEL CULVERTS

The design of corrugated steel culverts is not critically dependant upon the value of M* of the backfill; nevertheless, a reasonably accurate value, say within 25 per cent, is required to ensure safety and a reasonable level of economy in the quantity of steel used. The cost of the steel plate is typically between 25 and 50 per cent of the total cost of installing a culvert; and the effect of changing M* from the minimum value of 33 MN/m2 permitted in BD12/82 (DTp, 1 982), to a value of 177.5 MN/m 2, determined from a back- analysis of the performance of the culvert at Newport, may have produced overall savings in steel costs of up to 25 per cent, Temporal et al (1985). However more substantial savings may accrue from the use of cheaper and more widely available backfills. To this end, it seems necessary to improve the means of determining values of M*.

24

5 CONCLUSIONS

1. The CBR value and the compressibility of a crushed limestone backfill were both found to be dependent upon the initial dry density of the backfill. For the first loading cycle over the stress range 0 to 820 kN/m 2 the following relation was found to hold reasonably well for medium dense and dense specimens;

CBR=2 .5M* (2b)

where CBR is quoted in per cent and M* in MN/m 2.

2. At a particular initial dry density, the compressibility of the backfill did not appear to be significantly influenced by particle size or grading, or to be reduced by saturation following preparation.

3. No significant difference in stiffness was recorded for samples prepared from air-dried material compared to samples with a moisture content at or around the optimum.

4. When specimens were prepared with a moisture content higher than optimum, pore water pressures were generated in the material which allowed the fine particles, originally present in the aggregate or generated by compaction, to migrate through the specimen and form loose layers or pockets within the sample. The presence of these fines-rich zones gave rise to time-dependent strains and substantially reduced the overall stiffness of the material.

5. For comparative purposes, it is necessary to use the same initial dry density and moisture content for laboratory and in situ tests and to ensure that excessive fines are not generated during the preparation of the test specimens.

7 REFERENCES BOLTON, M D (1986). The strength and di latancy of sands. Geotechnique, Vol. 36, No. 1.

BRADY, K C, I AWCOCK and N R WlGHTMAN (1983). A comparison of shear strength measurements using two sizes of sh:earbox. Department of the Environment, Department of Transport, TRRL Laboratory Report 1105, Crowthorne.

BRITISH STANDARDS INSTITUTION (1975). BS 812, Part 2: Methods for sampling and test ing of mineral aggregates, sands and filters, London.

BRITISH STANDARDS INSTITUTION (1975). BS 1377: Methods of test for soils for Civil Engineering purposes, London.

CRONEY, D (1977). The design and performance of road pavements. Department of the Environment, Department of Transport, TRRL. HMSO (London).

DEPARTMENT OF TRANSPORT (1982). , Departmental Standard BD 12/82: Corrugated steel buried structures, London.

PENMAN, A D M and J A CHARLES (1975). The quality and suitability of rockfill used in dam construction. Department of the Environment. Building Research Establishment, Current paper CP87/75.

TEMPORAL, J, D A BARRATT and B E F HUNNIBELL (1985). Loading tests on an Armco pipe arch culvert. Department of Transport, TRRL Research Report 32, Crowthorne.

TERZAGHI, K (1960). Discussion: Salt Springs and Lower Bear River concrete face dams. Trans. ASCE, 125, Part II, pp. 1 3 9 - 1 4 8 .

6 ACKNOWLEDGEMENTS Most of the laboratory tests were undertaken by Middlesex Polytechnic Geotechnical Laboratories acting under contract to TRRL. The California Bearing Ratio tests were performed by Mr A Everett, and the 0.5 m oedometer tests by Mr T Williams supervised by Mr G Kirk. The 1.0 m diameter oedometer test was undertaken by Dr J A Charles of the Building Research Station.

This study forms part of the research programme of the Ground Engineering Division (Division Head: Dr M P O'Reilly) of the Structures Group of TRRL.

A P P E N D I X A

THE COMPRESSIBILITY OF BACKFILLS A,1 Data from oedometer tests

The compressibi l i ty of clayey soils is readily determined from laboratory oedometer tests, as defined by Test 17: BS 1377 (1975); in this case, values for the coefficient of volume compressibi l i ty (mv) are usually quoted where

mv=(eo -e l ) / ( 1 +eo). (o1' -oo' ) (1)

and eo and e~ are the initial and final voids ratios corresponding to Oo' and o~', the initial and final effective normal stresses.

25

As the va lue of my varies w i th the level of stress, it is necessary to quote the range of stresses over wh ich the determinat ion was made. A l te rnat ive ly , the slope of the compression line plot ted in voids rat io and log pressure space may be used. The slope of the semi- logar i thmic plot is def ined as the compress ion index (Co), where

Cc = (eo - el) / Iog (o1'/o0') (2)

Because of the practical problems of sampling and test ing coarse grained backf i l ls, their compress ib i l i t ies are usual ly est imated or determined f rom in situ tes ts such as the plate bearing tes t described in the code of pract ice for si te invest igat ions, BS 5930 (1981). On the other hand, laboratory compression tests have been under taken on coarse grained backf i l ls for the design or back-analysis of impor tant in-service s t ructures such as rockfi l l dams; see for example Charles (1973) . Wi th coarse grained soils it is usual to quote ei ther values of the constrained modulus M * (this is the reciprocal of mv) or a combinat ion of Young's modulus (E) and Poisson's rat io (u). Again, it is necessary to quote the range of stress over wh ich the determinat ion of compress ib i l i t y was made. For a homogeneous isotropic elast ic material compressed under condi t ions of no lateral strain, as occurs in oedometer tests, it can be shown that:

(Aox/~x) =(1 --u). E/(1 +u).(1 - 2 u ) = M * = 1/my (3)

where Aox and ~x are respect ive ly the change in appl ied ver t ica l stress and the result ing vert ical strain.

A . 2 Data from plate bearing tests In si tu plate bearing tests are carried out in sha l low pits and trenches, and most rarely at the bo t tom of boreholes. The depth to wh ich the soil is s ign i f i cant ly stressed during these tests is l imi ted to about 1.5 t imes the d iameter or w id th of the plate. It is therefore necessary to carry out a series of tes ts to determine the var ia t ion of compress ib i l i t y w i th depth. It is usual ly assumed in the analysis of plate bearing tests that the classical e last ic equation for the penetrat ion of a r igid plate into a semi- inf in i te plane surface appl ies. For a c i rcular plate of d iameter D,

E =lz.q.D.(1 -u2 ) /4 .s (4)

where q is the mean applied stress on the plate, and s is the ver t ica l def lect ion of the plate.

It should be appreciated tha t the values of E obta ined using the above equat ion are only approx imate measures of Young 's modulus. Some author i t ies , for example Burland and Lord (1970) , descr ibe the value of E der ived f rom plate bearing tes ts as an 'equ iva lent modulus of e las t ic i ty ' , and denote it by a symbol other than E.

Westergaard (1962) defined the modulus of subgrade reaction (k) as the secant slope of the pressure (q) versus mean sett lement (s) relation obtained from a 760 mm (30 inch) diameter plate bearing test. Thus,

k = q/s = 4E/riD.( 1 - u 2) (5)

The modulus (k) was normally calculated for a mean plate deflect ion of 1.25 mm (0.05 inch). The equipment required for carrying out 760 mm diameter plate bearing tests is cumbersome and not readily available on site. When plates smaller than the standard 760 mm diameter are used, the relation between modulus and diameter given in Figure A1 can be used to derive values of k. This f igure, reproduced from Croney (1977), was based on the original experimental data of Stratton (1944).

600

cO

E 500 -oE

X e- ~, 400

e - . . o o ~

~ 300 = E

-Q >

~ .~ 200

0 ~ _ ~ 0

o ~ 100

! 0 0 200 400 600 800 1000

Plate diameter (mm)

Fig.A1 Effect of plate size on apparent modulus of subgrade reaction

(Reproduced f rom Croney,1977)

A.3 Data from CBR tests In a CBR test the relation between force and penetration of a cyl indrical plunger is determined. The force, expressed as a percentage of a standard force, for penetrations of 2.5 mm and 5.0 mm are recorded and the higher value is defined as the CBR value. The CBR test can thus be v iewed as a srnallscale plate bearing test. For the standard aggregate, BS 1377: 1975, a CBR value of 100 per cent corresponds to forces of 8.7, 13.2 and 20 kN at respective penetrations of 1.25, 2.5 and 5.0 ram; for a 50 mm diameter plunger these forces correspond to stresses of 4.5, 6.8 and 10.3 MN/m 2 respectively. Thus from

26

E

E --~ 106

Combining and eliminating E from equations (3) and (5) gives the following relation;

k=q /s=4M* . (1 - 2u)hzD(1 -u ) 2 (6)

For a CBR test, at penetrations of 2.5 and 5.0 mm respectively;

CBR = 10OF/13.24 (7a)

CBR = 100F/19.96 (7b)

where F is the force in kN.

As q=4F/nD 2, then from equations (6) and (7a)

q = (4)(13.24)(CBR)/100nO 2 = 4M *.s.( 1 - 2u)/nD( 1 - u) 2 (8)

which on substituting for n, D = 50 mm, s=2 .5 mm, u=0.3 , yields the following relation;

CRB=0.77M* (9a)

where M* is in MN/m 2.

Similarly from equations (6) and (7b) for a penetration of 5 mm,

CBR= 1.0 M* (9b)

And using the force developed at a penetration of 1.25 mm, in comparison to the standard soil;

CBR=0.6M* (9c)

Thus from the above, a soil with a CBR value of 100 per cent should have a value of M* of about 88 MN/m 2.

The empirical relation between in situ CBR and k, reproduced from Croney (1977) in Figure A2, indicates that an in situ CBR value of 100 per cent corresponds to a modulus value (k) of about 250 MN/m2/m. Assuming that this value of k applies for a standard 760 mm diameter plate; then substituting for ~, D=0 .76 m, and v = 0 . 3 in equation (6) gives, for a CBR value of 100 per cent, M* = 183 MN/m 2.

~ 105

"5

-~ 10 4 0

I I 10 100

California bearingratio(percent)

equation (5) the value of k for the standard aggregate at a penetration of 1.25 mm with a CBR plunger is 3.6 GN/m2/m. Similarly at penetrations of 2.5 and 5.0 mm the values of k for the standard soil are 2.7 and 2.0 GN/m2/m respectively. Extrapolation of the relation given in Figure A1 indicates that the value of k for a standard 760 mm diameter plate and a penetration of 1.25 mm would be perhaps an order of magnitude less than for a 50 mm diameter CBR plunger, ie around 350 MN/m 2.

1000

Fig.A2 Empirical relation between k-value and CBR (Reproduced from Croney, 1977)

A . 4 References for Appendix BRITISH STANDARDS INSTITUTION (1975). BS 1377: Methods of test for soils for Civil Engineering purposes, London.

BRITISH STANDARDS INSTITUTION (1981). BS 5930: Code of Practice for site investigations, London.

BURLAND J B and J A LORD (1970). The load deformation behaviour of Middle Chalk at Mundford, Norfolk: a comparison between full- scale performance and in situ and laboratory measurements. Proc.Conf. In situ investigations in soils and rocks, British Geotechnical Society, London.

CHARLES, J A (1973). Correlation between laboratory behaviour of rockfill and field performance with particular reference to Scammonden Dam. Ph.D Thesis, London University.

CRONEY, D (1977). The design and performance of road pavements. Department of the Environment, Department of Transport, TRRL. HMSO (London).

STRATTON, J H (1944). Construction and design problems. Mil i tary airfields, a symposium. Proc. ASCE Vol 70 Part 1 pp 28 -54 .

WESTERGAARD, H M (1962). Stresses in concrete pavements computed by theoretical analysis. Public Rds., Washington, Vol 7, No 2, pp 25 -35 .

Printed in the United Kingdom for HMSO Dd8222792 8/90 C5 G476 10170