Está en la página 1de 7

Published by THENATIONAL RESEARCH COUNCIL CANADA OF

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by KYOTO UNIV NOGAKU LIB on 08/22/11 For personal use only.

VOLUME 7

NOVEMBER 1970

NUMBER 4

Safety factors in soil mechanics1


G . G . MEYERHOF
Department of Engineering, Nova Scotia Technical College, Halifax, Nova Scoiia
Received March 19,1970 The influences on the safety margin in earthwork and foundation engineering are considered in relation to soil exploration and tests, analysis of stability under the applied loads, and construction and operation of the structure during its service life. Customary overall and suggested partial safety factors used in stability analyses are mainly governed by the variability and uncertainty of assessment of the soil resistance, the variability of the applied loads, approximations in the stability analyses, and the seriousness of a failure. On the basis of probability concepts of safety, the conventional overall safety factors used in earthwork and foundation engineering analyses are related to approximate probabilities of stability failure and to a range of overall coefficients of variation, which are consistent with the range of observed coefficients of variation of the soil resistance, applied loads, and stability analyses. Les influences sur la marge de securitt dans la mtcanique des sols appliqute sont consid& rtes en relation B l'exploration et des essais des sols, des calculs de stabilitt aux efforts appliqutes, la construction et l'operation des structures pendant leurs vie de service. Globals conventionnels et partials suggerts facteurs de stcuritt utilists dans les calculs de stabilitt sont gouvernts par la variabilitt et incertitude d'assessement de la resistance des sols, la variabilitt des efforts appliqutes, des approximations dans les calculs de stabilitt et la stvtritt d'un insuccts. Sur la base de la theorie de probabilitt de la stcurit6 les facteurs globals conventionnels de s k u r i t t employts dans la mtcanique des sols appliquCe sont rapportts aux probabilitks approximatives d'un insuccts de stabilitt et B une portte des cotfficients globals de variation qui conforrnent avec la portCe des cotfficients de variation observts de la rtsistance des sols, des efforts appliqutes et Aes calculs de stabilitt.

Introduction Structural design is usually based on the requirements that the structure and its components must be safe to withstand the maximum loads and forces which have a reasonable chance of occurrence during the life of the structure, and that the structure must serve the designed functions well and economically. To ensure reasonable safety of the structure, the design is usually carried out by introducing into the structural analysis a safety factor, which may be defined as the ratio of the resistance of the structure to the applied loads in order to ensure freedom from danger, loss, or risks. The
'This paper was presented orally at the Seventh International Conference on Soil Mechanics and Foundation Engineering, Mexico, 1969, and a summary of the paper was published in Volume 3 of the proceedings of that Conference.
Canadian Geotechnical Journal, 7, 349 (1970)

magnitude of the safety factor required depends mainly on the reliability of the design data and their interpretation in the assessment of the structural resistance and applied loads, the accuracy of the structural analyses, the quality of construction and maintenance, and the probability and seriousness of a failure during the service life of the structure. In the design of earthworks, earth retaining structures, and foundations there are more approximations and uncertainties than in the design of other structures, because of the complexity of soil behavior and incomplete knowledge of the subsoil conditions. These approximations and uncertainties have to be evaluated in each given case, and the total margin of safety should be reasonable and sufficient on the basis that the planning, design, construction, and operation of the particular structure

350

CANADIAN GEOTECHNICAL JOURNAL.

VOL. 7,

1970

are carried out in accordance with good practice involving reasonable skill, judgement, and experience. Influences on Safety Margin An estimate of the soil resistance to applied loads in earthwork and foundation engineering requires a site investigation and soil tests to assess the strengths of the affected soil masses during construction and service life of the structure. Based on the size of the proposed work and previous geological and other information, a soil exploration is made by boring, sounding, sampling, and in-place testing of strength and permeability. Since such investigations may not disclose the most unfavorable subsurface conditions, a margin of safety is required to cover important weak zones, erratic discontinuities, uncertainties associated with interpolation between test locations, and unavoidable sample disturbance. Given an adequate number and size of representative samples, the significant index properties, mechanical and hydraulic properties of the various soils can be determined in the laboratory. The appropriate shearing strength parameters, pore pressures, and volume change characteristics are usually obtained from standard triaxial compression tests on cylindrical specimens (Bishop and Bjerrum 1960). The probable effects of other loading, drainage, and environmental conditions on the test results are often assessed from previous experience, and the margin of safety must also provide for differences of soil properties obtained from small samples in the laboratory compared with the in-place properties of large soil masses. On the basis of the results of the site investigation and soil tests simplified soil profiles are constructed, and any differences between the real and approximated soil conditions have to be covered by the margin of safety. Analyses of the proposed structure to determine its stability against failure under the applied loads at the critical stages during the construction and service life are usually carried out by semiempirical methods based on plastic theory. In many cases these methods assume homogeneous soils and simplified failure surfaces leading to upper bound solutions, which overestimate the degree of stability and the internal

resistance at some locations within the structure and subsoil. Moreover, the real mechanism may include local, progressive, discontinuous, and other features, and the failure pattern depends on the true shearing characteristics of the soil and thus differs from the assumed simplified pattern based on effective or total strength parameters. Even greater uncertainties arise in estimates of pore pressures, especially when they depend mainly on stress changes in the soil mass rather than on the level or flow pattern of ground-water. The design of the proposed works should also allow for normal deviations and tolerances of the dimensions and behavior of individual components and the whole structure during the construction. While defects and accidental errors in the execution of the work can be reduced by adequate inspection and control of the quality of materials and workmanship, the construction procedures should be flexible to allow changes to be made in the design or construction if the results of performance observations of movements, earth pressures, and water pressures during the construction make it necessary. Although it is customary to use lower margins of safety in the case of performance observations for temporary works on the basis of calculated risks (Casagrande 1965), all necessary precautions against failure of the structure and any adjacent works have to be taken. Some margin of safety must also be provided for changes in climatic and other environmental conditions and for critical construction stages due to limitations of the construction season and work stoppages. Satisfactory operation of the structure for its intended use should be ensured with a minimum of maintenance or remedial measures during the expected or economic service life, which usually varies from about 50 to 200 years, except for temporary works which may have a service life of a few years only. The initial margin of safety against failure of the structure can be allowed to decrease if necessary to a small value at the end of the design life, and thereby provide for some deterioration of the resistance of the structure and its support under long-term and fluctuating loading conditions. Overall and Partial Safety Factors The order of magnitude of the various in-

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by KYOTO UNIV NOGAKU LIB on 08/22/11 For personal use only.

MEYERHOF: SAFETY FACTORS

351

TABLE I Values of minimum overall safety factors Failure type Item Earthworks Earth retaining structures Foundations Uplift, heave Exit gradient, piping Safety factor
1 . 3 to 1 . 5 1 . 5 to 2 2 to 3 1 . 5 to 2 . 5 3 to 5

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by KYOTO UNIV NOGAKU LIB on 08/22/11 For personal use only.

Shearing Seepage

fluences affecting the margin of safety in soil mechanics may be assessed on the basis of the values of the present customary safety factors used in the stability analyses of earthworks, earth retaining structures, and foundations. The range of the conventional minimum overall safety factors (Terzaghi and Peck 1967) are given in Table I. While the upper values of these safety factors are used in stability analyses of structures under normal, service conditions, the lower limits are used in analyses based on maximum loading conditions (excluding earthquakes) and for temporary works. The lower values are also sometimes used when analyses are based on the results of full-scale field tests or failure analyses of similar structures on the same subsoil conditions. When extreme conditions of loading (including earthquakes) or soil resistance (e.g. full-softened or residual strength) are considered or when the design life of the structure is approached, a marginal safety factor against stability failure of about 1.1 is often accepted, which represents the limit of accuracy of stability analyses discussed below. The values of conventional overall safety factors against instability of various types of structures and against seepage developed through judgement and experience associated with good practice. The magnitude of these safety factors increases mainly with greater variability and degree of uncertainty of the assessment of the soil resistance, due to greater involvement of subsoil conditions, and with greater variability of the loading. Further, the safety factors in relation to shearing failure increase with the need to restrict plastic zones and shearing deformations of soils, especially to prevent damage of associated other structural materials or components, and as the potential losses resulting from a partial or complete failure increase. On the other hand, the

safety factors in seepage considerations are based mainly on the uncertainties involved and the seriousness of sudden failures without much warning, and these safety factors are greater when local soil conditions are involved. Since both the resistance of the soil and the applied loads are to some extent independent variables, a more uniform margin of safety for different types and components of structures may be obtained by separate safety factors on the resistance and the loads, as commonly used in structural limit design (Freudenthal 1956 and 1968; Pugsley 1955 and 1966). Such partial safety factors have to ensure the same minimum margin of safety as the corresponding overall safety factors based on experience. Partial safety factors on shearing strength may differ for the cohesion and friction parameters (Taylor 1948; Brinch Hansen 1956 and 1965; Hueckel 1964) or these factors may be common (Janbu et al. 1956) and allow also for uncertainties in effective stress and pore pressure predictions associated with the estimates of shearing strength. Similarly, the partial safety factors on loads may differ for dead loads and various types of live loads. The range of these suggested minimum partial safety factors for stability analyses are given in Table 11. The magnitude of the partial safety factors depends on the same considerations as the overall factors so that the partial safety factors for foundations are somewhat greater than those for earthworks and earth retaining structures, and the factors for live loads are greater than for dead loads. Since partial safety factors have so far been used only to a limited extent in earthwork and foundation engineering, their effectiveness in providing a more uniform margin of safety in practice cannot be fully assessed at present.
TABLE I1 Values of minimum partial safety factors Item Earthworks and earth retaining structures Foundations Dead loads Water pressures Live loads Shearing strength parameter Cohesion (c) Friction (tan +) Cohesion and friction Cohesion (c) Friction (tan +) Safety factor
1.5 1.2 1 . 3 to 1 . 5 2 to2.5 1 . 2 to 1 . 3 1.0 1 . 0 to 1 . 2 1 . 2 to 1 . 5

352

CANADIAN GEOTECHNICAL

JOURNAL. VOL. 7,

1970

I
-

&(a,

LOADS
RESISTANCE

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by KYOTO UNIV NOGAKU LIB on 08/22/11 For personal use only.

aC

Id

u
0
Q, Q;R,

R,

LOAD O R R E S I S T A N C E
FIG. 1. Frequency distributions of loads and resistance.

Probability Concepts of Structural Safety Concepts and criteria of the safety of structures have been studied in a number of papers (e.g. Freudenthal 1956 and 1968; Pugsley 1955) and in the first book on this subject (Pugsley 1966). The probabilistic interpretation of safety factors requires a knowledge of the statistical distributions of the resistance of the structure and of the applied loads during the service life. This approach to safety analysis can strictly be used only for sets of independent, random variables which can be observed within the range of interest. It also requires the assessment of a numerical value for the acceptable risk of failure on the basis of personal, economic, or other considerations (Levi 1958). Figure 1 shows the frequency distribution curves for the applied loads ( Q ) and the soil resistance ( R ) with mean values of Qo and Ro, respectively, corresponding to the overall safety factor Fo = Ro/Q,,. The point of intersection of the two curves represents the limit state Ql = R1 when the structure fails. This state corresponds to the partial safety factors F, = Ql/Qo and F, = Ro/R, on the loads and resistance, respectively. If the resistance is timedependent, the position of curve R changes with respect to curve Q in time and thus both affects the overall and partial safety factors. In the small region where the two curves overlap there is a small probability of failure ( P , ) and if the shapes of the distribution curves are known, the value of Pf associated with the safety factors can be estimated. In some cases the frequency distribution curves can be ap-

proximated by normal or logarithmic-normal curves of probability theory (Langejan 1965). It follows that the safety factors do not imply absolute structural safety, but they include a small acceptable risk of accident or probability of failure of the structure during its service life. Thus, surveys of failures of earth dams constructed since soil mechanics became established in engineering practice about 30 years ago (Middlebrooks 1953; Feld 1965) indicate that roughly 1% of earth dams have failed partly or completely by either sliding (about one-third) or piping (about two-thirds). Most of these failures occurred during the first few years of reservoir o~erationand took place in cohesive soils. Since recent improvements in design and construction are probably offset by an increasing size and boldness of structures and the use of less favorable sites, the conventional minimum safety factor of about 1.5 against sliding failure of earth dams would be associated with a probability of stability failure of roughly 4% during the first years of service life, which is greater than can reasonably be tolerated. Moreover, since current effectivestress stability analyses including interslice forces frequently indicate safety factors of about 10 to 20% greater than previously used methods ignoring such forces, it may be suggested that the minimum overall safety factor should be increased from about 1.5 to about 1.7, especially in connection with cohesive soils, unless careful performance observations are made both during and after construction for a period of several years. No surveys of other types of earthworks such as cuts and fills are available, but their probability of failure is likely to be smaller (perhaps roughly one per thousand) due to the absence of impounded water effects on these works. A similar or somewhat smaller probability of stability failure may also be associated with the design of large earth retaining structures, such as retaining walls and shcet pile walls of heights exceeding about 20 ft ( 6 m ) . These structures fail usually by unfavorable foundation conditions, such as clay subsoils, and less frequently by excessive lateral Fressure from cohesive soils (Peck et al. 1948). The corresponding overall safety factors may be judged, therefore, to be fairly adequate, especially if the upper limit of customary safety factors is approached.

MEYERHOF: SAFETY FACTORS


fi

a u e
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by KYOTO UNIV NOGAKU LIB on 08/22/11 For personal use only.

OVERALL 5 A F E T Y F A C T O R ( , F)

LEGEND

E = EARTHWORKS F = FOUNDATIONS R = EARTH RETAINING


STRUCTURES
V

COEFFICIENT OF
VAR\ATION

FIG.2. Comparison between safety factor and probability of stability failure.

In spite of combined uncertainties of applied loading and subsoil resistance, the numerical percentage of foundation failures is very small (Szechy 1961; Feld 1968); however, the importance of variability of soil conditions increases as the size of foundations decreases, and foundation failures often affect the stability of neighboring areas. Further, the uncertainties of subsoil and foundation behavior associated with pile foundations are greater than with spread foundations, and the ultimate bearing capacity of foundations may be reduced by changes of the soil and water conditions and by climatic and other environmental influences. Although the probability of stability failure of foundations associated with the conventional overall safety factors is not known at present, it is probably greater than that of superstructures designed for the same range of safety factors. The latter probability of failure has been estimated to be of the order of to 10-"or concrete structures and about to for steel structures for a single application of the design load (Freudenthal 1968). The present overall safety factors used in stability analyses of foundations correspond, therefore, perhaps to a probability of the order of one foundation failure in about 5 000 to 10 000 structures, which may be just acceptable. The results of these safety analyses are shown in Fig. 2, which indicates the order of magnitude of the range of probability of stabil-

ity failure of earthworks, earth retaining structures and foundations designed for the upper limits of the corresponding conventional minimum overall safety factors. The present semiprobabilistic approach to safety analysis is limited by numerous non-random effects in the applied loads and soil resistance in earthwork and foundation engineering problems. However, it is of interest to note that the results of the safety analyses shown in Fig. 2 are consistent with overall coefficients of variation (ratio of standard deviation to mean resistance) in the range of about 0.1 to 0.2 on the basis of a normal frequency distribution or in the range of about 0.1 to 0.3 on the basis of a logarithmicnormal distribution of the soil resistance. if in both cases the relatively small variation of the applied loads is ignored. The lower limit of these coefficients governs the stability of earthworks and the upper limit !governs that of foundations, as would be expected. Lack of sufficient statistical data permits only an indication of the order of magnitude of some of the parameters contributing to this random variability. Thus, in well-controlled embankment construction (Davis 1953; Turnbull et al. 1966) the coefficient of variation of the placement dry density of fill material ranged between about 0.02 and 0.04, while the coefficient of variation of the placement water content varied between about 0.05 and 0.15. both soil properties havinz a normal frequency dis-

3 54

CANADIAN GEOTECHNICAL

JOURNAL. V L 7. O.

1970

tribution. The corresponding dead loads have thus a coefficient of variation of at least 0.02, while live loads, such as floor and wind loads, have a coefficient of variation of about 0.2 (Freudenthal 1968). The friction parameter (tan 4 ) of cohesionless soils varies roughly with the dry unit weight of the material and, on the basis of the above findings, would show a normal frequency distribution. The corresponding coefficient of variation may be in the range of about 0.1 to 0.15, which has also been found for the effective friction parameter of some cohesive soils, while the effective cohesion parameter was generally more variable (Lumb 1966). Since the undrained shearing strength of normallyconsolidated clays has a roughly linear logarithmic variation with the water content, a logarithmic - normal frequency distribution may be expected for the undrained strength of such clays, while the undrained strength of heavilyoverconsolidated clays would show a normal frequency distribution. The corresponding coefficient of variation was between about 0.2 and 0.3 for some lightly-overconsolidated clays (Lumb 1966; Wu and Kraft 1967) and for large-scale undrained tests on heavily-overconsolidated stiff-fissured London Clay (Ward et al. 1965; Hooper and Butler 1966). Some indication of the degree of accuracy of customary methods of stability analyses can be obtained by comparing estimated overall safety factors with the real value of unity in the case of failures when particularly detailed soil investigations reduce other uncertainties to a minimum, although such comparisons underestimate the full effect of inaccuracies in design assumptions and calculations. These comparisons indicate (Bishop and Bjerrum 1960) that for stability failures of earthworks, earth retaining structures and foundations at the end of the construction in saturated intact clays the average value of the estimated safety factors was very close to unity. Moreover, these comparisons show that the estimated safety factors had a coefficient of variation of about 0.1 and could thus lead to an overestimate of the stability by about 15% for a 95% confidence limit. Similarly, for long-term stability failures of natural slopes in intact clays the limited data available at present indicate that the safety

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by KYOTO UNIV NOGAKU LIB on 08/22/11 For personal use only.

factors may be overestimated by up to about 15%, while in stiff-fissured clays, the safety factors may be overestimated by about 10% if the residual friction only is used (Skempton 1964). Although more field data are required, it may be concluded that the range of overall coefficients of variation indicated by the safety analyses of earthworks, earth retaining structures, and foundations (Fig. 2) is consistent with the range of observed coefficients of variation of the soil resistance, applied loads, and stability analyses; for it is unlikely that these variations act together at their maximum and most unfavorable values and at the worst location in the structure at a given time.

Conclusions The margin of safety in earthwork and foundation engineering depends mainly on the uncertainties and variability of the soil conditions and approximations in the stability analyses. Careful soil explorations and tests and their interpretation are, therefore, essential to assess the resistance of the affected soil masses during construction and service life of the structure. Customary overall safety factors used in stability analyses relating to shearing failure and seepage considerations are, therefore, mainly governed by the degree of variability and uncertainty of assessment of the soil resistance, the need to restrict shearing deformations to prevent damage of other associated structural materials, and the seriousness of a failure. Partial safety factors on the soil resistance and applied loads may lead to a more uniform margin of safety, but limited use hitherto prevents a full assessment of their effectiveness at present. Safety analysis of earthworks indicates that earth dams designed for a minimum overall safety factor of about 1.5 have a probability of sliding failure of roughly 4% during the first few years of service life. It is suggested, therefore, that the minimum safety factor be increased to about 1.7, unless careful performance observations are made during and after construction. The customary overall safety factors used in the design of large earth retaining structures and foundations correspond perhaps to a probability of failure of the order of lo-"

MEYERHOF: SAFETY FACTORS

355

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by KYOTO UNIV NOGAKU LIB on 08/22/11 For personal use only.

and they appear to be fairly adequate. to A semi-probabilistic approach to safety analyses indicates that the conventional safety factors in stability analyses in earthwork and foundation engineering are consistent with overall coefficients of variation in the range of about 0.1 to 0.3, which is supported by the range of observed coefficients of variation of the sail resistance, applied loads, and stability analyses.
BISHOP, W. and BJERRUM, 1960. The relevance A. L. of the triaxial test to the solution of stability problems. Amer. Soc. Civil Eng. Conf. Shear Strength of Cohesive Soils, Boulder, Colorado. p. 437. CASAGRANDE. 1965. Role of calculated risk in A. earthwdrk and foundation engineering. J. Soil Mech. Found. Div., Proc. Amer. Soc. Civil Eng. 91, SM4, p. 1. DAVIS, J. 1953. Quality control of earth embankF. ments. Proc. Int. Conf. Soil Mech., 3rd. Zurich, 1, p. 218. FELD,J. 1965. The factor of safety in soil and rock mechanics. Proc. Int. Conf. Soil Mech. Found. Eng., 6th. Montreal, Quebec. 3, p. 185. 1968. Construction failure. John Wiley and Sons, New York. FREUDENTHAL, M. 1956. Safety and the probability A. of structural failure. Trans. Amer. Soc. Civil Eng. 121, p. 1337. 1968. Critical appraisal of safety criteria and their basic concepts. Prel. Publ. 8th Congr. Int. Ass. Bridge and Struct. Eng., New York. p. 13. HANSEN, BRINCH.1956. Limit design and partial J. safety factors in soil mechanics. Danish Geot. Inst., Bull. 1, p. 1. 1965. The philosophy of foundation design: design criteria, safety factors and settlement limits. Symp. Bearing Capacity and Settl. of Found., Duke University, Durham, North Carolina. p. 9. HOOPER, A. and BUTLER, A. 1966. Some numerJ. F. ical results concerning the shear strength of London clay. GCotechnique, 16, p. 282. HUECKEL, 1964. The problem of safety in soil S.

mechanics and foundation engineering. Proc. Sem. Soil Mech., Lodz, p. 137. JANBU, N., BJERRUM, and KJAERNSLI, 1956. L., B. Soil mechanics applied to some engineering problems. Norweg. Geot. Inst., Publ. 16, p. 1. LANGEJAN, 1965. Some aspects of the safety factor A. in soil mechanics considered as a problem of probability. Proc. Int. Conf. Soil Mech. Found. Eng., 6th, Montreal, Quebec. 2, p. 500. LEVI, R. 1958. Les calculs de sCcuritC en matiere de fondations. Ann. Inst. Tech. Bat. et Trav. Publ. 11, p. 541. LUMB,P. 1966. The variability of natural soils. Can. Geotech. J. 3. u. 74. T.. MIDDLEBROOKS, A. 1953. Earth-dam practice in the United States. Cent. Trans. Amer. Soc. Civil Eng. CT, p. 697. H. PECK,R. B., IRELAND, O.,and TENG,C. Y. 1948. A study of retaining wall failures. Proc. Int. Conf. Soil Mech. Found. Eng. 2nd. Rotterdam. 3, p 296. PUGSLEY, 1955. Report on structural safety. Struct. A. Engineer, 34, p. 141. 1966. The safety of structures. Arnold, London. SKEMPTON, W. 1964. Long-term stability of clay A. slopes. GCotechnique, 14, p. 77. SZECHY, 1961. Foundation failures. Concrete Publ., K. London. TAYLOR, W. 1948. Fundamentals of soil mechanics. D. John Wiley and Sons, New York. p. 414. TERZAGHI, and PECK,R. B. 1967. Soil mechanics K. in engineering practice. John Wiley and Sons, New York. THORNBURN, H. and LARSEN, R. 1959. A statisT. W. tical stuudy of soil sampling. J. Soil Mech. Found. Div., Proc. Amer. Soc. Civil Eng. 85, No. SM5, p. 1. TURNBULL, J., COMPTON, R., and AHLVIN, G. W. J. R. 1966. Quality control of compacted earthwork. J. Soil Mech. Found. Div., Proc. Amer. Soc. Civil Eng. 92, No. SM1, p. 93. A., S. WARD,W. H., MARSLAND, and SAMUELS, G. 1965. Properties of the London clay at the Ashford common shaft. Gtotechnique, 15, p. 321. Wu, T. H. and KRAFT,L. M. 1967. The probability of foundation safety. J. Soil Mech. Found. Div., Proc. Amer. Soc. Civil Eng. 93, No. SMS, p. 213.

También podría gustarte