[elsevier] reliability of reinforced concrete structures under fatigue

Upload: pauldelgado

Post on 04-Jun-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    1/9

    Reliability of reinforced concrete structures under fatigue

    Y.S. Petryna*, D. Pfanner, F. Stangenberg, W.B. Kratzig

    Department of Civil Engineering, Ruhr-University Bochum, Universitatsstr. 150, D-44780 Bochum, Germany

    Abstract

    This paper focuses on time-variant reliability assessment of deteriorating reinforced concrete structures under fatigue conditions. Astrategy combining two time scales, namely the micro-scale of instantaneous structural dynamics (or statics) and the macro-scale of structural

    lifetime, is proposed. Non-linear response of reinforced concrete structures is simulated by means of the finite element method with adequate

    material model. A phenomenological fatigue damage model of reinforced concrete is developed and calibrated against experimental results

    available in the literature. Reliability estimates are obtained within the response surface method using the importance/adaptive sampling

    techniques and the time-integrated approach. The proposed assessment strategy is illustrated by an example of a concrete arch under fatigue

    loading. The obtained results show a general inapplicability of local and linear fatigue models to system level of structures.q 2002 Elsevier

    Science Ltd. All rights reserved.

    Keywords:Structural reliability; Reinforced concrete; Fatigue; Damage

    1. Introduction

    The necessity of time-variant reliability assessment of

    deteriorating structures is becoming increasingly recog-

    nized [l4]. However, technical service life limited by

    structural failure is mainly addressed by defining either

    critical load or material parameters [4,5]. Degradation

    analysis of structural capacity is often wrongly con-

    sidered as durability assessment of constructional

    materials or single components overlooking the impact

    of local damage on global failure conditions.

    A similar situation may also be observed in fatigue of

    reinforced concrete structures focused mainly on the

    local material level [68]. Fatigue failure itself plays in

    concrete engineering a secondary role, but it caninfluence usual failure mechanisms and accelerate

    structural degradation. Such effects can be studied solely

    on the system level taking interaction of damage

    mechanisms and stress redistribution within entire

    structure into account.

    Response of reinforced concrete as composite construc-

    tional material exhibits strongly non-linear character of

    response and corresponding damage processes [9]. Struc-

    tural reliability assessment under such conditions implies a

    system approach, which combines damage mechanics, non-

    linear structural analysis and probabilistic methods in a

    rational way. In our opinion, such a general and unified

    approach is still missing.This work proposes an assessment strategy for structural

    reliability problems under fatigue conditions and illustrates

    this assessment by a test numerical example. The results

    show a fundamental inapplicability of local and linear

    fatigue damage models to structural level.

    2. Material model of reinforced concrete

    Realistic material models of reinforced concrete suitable

    to be used in structural analysis must correctly simulate at

    least four different components: a ductile concrete responseto compression and a brittle response to tension, a ductile

    behavior of steel reinforcement and highly non-linear bond

    effects[9]. The material model applied in the present study

    has recently been developed by Polling[10]and bases on

    the uniform elasto-plastic continuum damage theory [11,

    12]. It describes the time-independent static material

    response to monotonic or cyclic loads.

    For the model formulation, the hardening and softening

    functions of inelastic concrete response are of importance

    (Fig. 1). The model fits the ascending branches of the

    uniaxial stressstrain relationships stated in the Model

    Code 90 [13], whereas the applied softening laws in the

    post-peak ranges depend on the dissipated energy in the

    localized process zones. The descending branch of

    0951-8320/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.

    PII: S 0 9 5 1 - 8 3 2 0 ( 0 2 ) 0 0 0 5 8 - 3

    Reliability Engineering and System Safety 77 (2002) 253261www.elsevier.com/locate/ress

    * Corresponding author. Fax: 49-234-32-14370.

    E-mail address:[email protected] (Y.S. Petryna).

    http://www.elsevier.com/locate/resshttp://www.elsevier.com/locate/ress
  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    2/9

    the stress strain curve for uniaxial compression reads

    sc1 21

    2gcfc1c2fc

    gc1 gc

    21c12

    1

    with

    gc p

    2fc1c

    2 gpcl 21

    2fc 1c1 2 b b

    fc

    Ec

    2 ; gpcl Gclleq :

    Herein,1crepresents the peak strain and fcthe compressive

    strength,gccontrols the area under the curve and therewith

    the dissipated energy depending on the initial Youngs

    modulusEc, the crushing energy Gcl, the equivalent length

    leq of the finite element and on the ratio b of plastic and

    damage components of inelastic strains.

    For tensile stresses, the softening law reads

    sct1 fctexp 1

    gt

    fct

    Ec2 1

    ; gt

    Gf

    leqfct2

    1

    2

    fct

    Ec;

    2

    where fct denotes the tensile strength, gt controls the area

    under the stressstrain curve and Gf indicates the fracture

    energy.

    According to the envelope concept confirmed byexperiments, the monotonic stressstrain curves provide a

    bound for all hysteretic loops. Cyclic un- and reloading

    paths of the material are modeled linearly with a gradient

    corresponding to the damaged stiffness.

    The uniaxial material model for reinforcing steelcorresponds generally to the classical elasto-plasticity.

    The yield surface has been modified according to Ref. [14]

    in order to correctly describe the Bauschingers effect. The

    stressstrain relationship in the plastic hardening domain

    reads

    s1 fy Bs 1 2 1h 21 2 1h

    gs

    gs1f 2 1hgs 21

    3

    with

    gs Bs1f 2 1h

    fs 2fy 2Bs1f 2 1h

    ;

    where fy denotes the yield stress, fs the tensile strength, 1hthe strain at the beginning of plastic hardening, Bs, the

    plastic hardening modulus and 1f the constriction strain

    [10].

    3. Fatigue damage model of reinforced concrete

    General strategy. For complex structures under high-

    cycle fatigue loading (N . 103 load repetitions), an explicit

    tracing of each single load cycle becomes unacceptable due

    to enormous computational efforts. An alternative way

    consists in the application of prognostic models predicting

    the increments of fatigue damage, similar to the increments

    of creep strains according toRef. [15]. Starting with a stress

    state after a single cycle of fatigue loading, the model

    computes the degradation of material properties for n1repetitions. Structural simulation with damaged parameters,

    which are evidently different in each material point,

    provides the estimate of residual carrying capacity. Foreach of the following increments ni, an updating iteration

    must be carried out taking redistribution of stresses and

    damage into account.

    Fatigue failure criterion. The following S N curve

    according to Ref. [6] has been applied to determine thenumber of load cycles to failure Nf:

    Smax 1 2 0:06621 2 0:556RlogNf 2 0:0294 logT; 4

    where Smax denotes the ratio of maximum stress to static

    strength,Rthe stress amplitude and Tthe period of one load

    cycle. Since the envelope stressstrain curve corresponding

    to monotonic loading can be assumed to cover the strain

    values of arbitrary load histories [7], fatigue failure isassociated with the moment, when a regular loading path

    intersects the envelope (point B,Fig. 1). The stiffness andstrain values at the moment of fatigue failure read for

    compression

    Ec;fail 1fail1 2 b

    Smaxfc

    b

    Ec

    21

    ;

    1fail 1c

    ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi21cgcfc

    11

    Smax

    ;

    s 5

    and tension

    Ect;fail Smaxfct1t;fail

    ; 1t;fail fct

    Ec2 gt ln Smax: 6

    Fig. 1. Material law and cyclic loading for uniaxial compression and tension.

    Y.S. Petryna et al. / Reliability Engineering and System Safety 77 (2002) 253261254

  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    3/9

    Fatigue damage evolution. Fatigue damage accumu-

    lation in concrete observed in experiments cannot be

    described by classical linear theories since it is a non-linear

    process. This accumulation is found to cohere with the

    evolution of total strains 1tot (Fig. 5), which proceeds

    parabolically in the initial and in the final stage of fatigue

    life, whereas in the intermediate phase a constant strain rate

    can be observed. A regression analysis of experimental data

    of several authors leads to the following evolution functions:

    1tot

    n

    1

    350d80 2 4d20n 2 5d80 35d20 1RS

    10max

    1fail 2 10 for n# 0:2; 7

    1tot

    n

    5

    3d80 2 d20 1RS

    10max 1fail 2 10

    for 0:2 , n # 0:8; 8

    1tot

    n

    50

    3 d20 2 4d80n 2 55d80 2 15d20

    1RS10max

    50n 2 40

    1fail 2 10 for n . 0:8; 9

    where ndenotesn=Nfwithnapplied load cycles, 10indicatesthe initial strain at the beginning of the fatigue process (point

    A,Fig. l),d20,d80represent the measured relative changes of

    total strains after 20 and 80% of the fatigue life, respectively.The strain evolution according to Eqs. (7)(9) is depicted in

    Fig. 5.

    Due to the fact that total strains in the first stage of life are

    governed by creep processes, which are not attributed to

    mechanical damage, a modified cyclic creep compliance

    function Jt; t0 according to Ref. [16] is introduced tocalculate the fatigue strains 1fat associated with fatigue

    damage:

    1fat 10 1tot 2 10nkfat

    Jn; t0 2 EcJ1; t0 2 Ec

    : 10

    Herein,kfat

    controls the fatigue strain rate in dependence of

    the period Tof cyclic loads acting since time t0. Since the

    calculated fatigue strains are imbedded into the constitutive

    equations by the fatigue failure criteria (5) and (6), the scalar

    parameter of fatigue damage for the uniaxial ease can be

    directly derived:

    Dc;fat 1fat 2 10

    1fail 2 101 2

    Ec;fail

    Ec

    : 11

    Its evolution is depicted inFig. 2.

    Damaged material parameters. The reduced elasticity

    modulus defining the slope of the reloading path for the

    current fatigue state reads:

    Ec;fat Ec1 2Dc;fat: 12

    The values of residual strengths fc,fat, fct,fat(Fig. 2) and the

    new peak strain 1c,fat of a damaged material can easily be

    determined by means of the original stress strain relations(1) and (2) presuming the knowledge of the intersection

    point of the current loading path with the initial envelope (at

    strain 1int). Due to difficulties in determining this point in

    closed-form, iteration algorithms have been applied. Then,

    one obtains for the modified localized crushing energy (in

    compression):

    Gcl;fat 1

    2fc;fat 1c;fat1 2 b b

    fc;fat

    Ec;fat

    !lim

    1c;fat1

    2 sc1d1: 13

    The first term corresponds to the area under the ascendingbranch of the damaged stress strain curve, whereas the

    integral represents the area under the descending branch.

    For tension domain, the new localized fracture energy

    reads:

    Gf;fat leq gtfctexp 1

    gt

    fct

    Ec2 1int

    1

    2

    f2ct;fat

    Ec;fat

    " #: 14

    These modified parameters define a new, damaged material

    law for concrete (Fig. 3) used in the further structural

    simulations.

    Reinforcement. In contrast to concrete, the known

    Palmgren-Miner hypothesis provides acceptable estimates

    of damage accumulation in steel under elastic deformations,

    Fig. 2. Evolution of damage, degradation of compressive and tensile strength.

    Y.S. Petryna et al. / Reliability Engineering and System Safety 77 (2002) 253261 255

  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    4/9

    presuming the absence of inherent stresses or initial

    damage. The following modified rule has been applied

    below in order to allow for physical interpretation of

    damage variable:

    Ds;fat 21

    a 1 2 1 2 e2a

    1

    1 h

    Xmk1

    nkDsk

    NfDsk

    " #

    withXmk1

    nkDsk

    NfDsk# 1;

    15

    where Dsk; represents the stress amplitude, Nf the numberof load cycles to failure according to Ref. [13] and nk the

    number of load cycles within the corresponding stress range.

    The parameter Ds,fat allows for damage-oriented formu-

    lation of the material law for steel depicted in Fig. 3.

    4. Probabilistic model calibration

    Probabilistic aspects of fatigue. Experiments show that

    fatigue processes are generally subject of great variability

    both in load cycles to failure and in deformation history. The

    scatter ofNfhas been confirmed to fit the known Weibulldistribution [8]. Several procedures to estimate fatigue

    reliability have been discussed in Ref. [17]. However, they

    are mainly based on stochastic fracture mechanics and

    hardly applicable to real concrete structures exhibiting

    usually multiple cracks.

    Definition of random variables. The predictions offatigue damage presented above depend on two parameters

    of major influence: the fatigue life Nf and the relativechanges of total strains d20and d80. The evolution of these

    values and their scattering over time has been determined

    based on experimental data available in Refs. [8,18]. It

    seems to be rational to introduce these values as principal

    random variables into the model. Such an approach

    overcomes difficulties in finding adequate time-dependent

    distributions of conventional random variables such as

    strength and stiffness, since fatigue life Nf and strain

    evolution naturally include these values.

    Parameterization of random variables. Because of its

    physical assumptions, the Weibull distribution of fatigue lifeis proved to fit best the results of fatigue experiments in

    general [19] and for concrete in particular [4]. However,

    such parameters as the characteristic life Vs, the minimum

    life n0,s and the shape parameter ss of the distribution

    function

    FNn ; P{Nf # n} 1 2 exp 2n 2 n0;s

    Vs 2 n0;s

    !as16

    have seldom been specified with regard to concrete as in

    Ref. [8].A plenty of experimental data have been processed

    in this work in order to obtain reliable values of parameters

    of the Weibull distribution. The differences of these data

    due to unknown experimental conditions do not allow for an

    overall statistical analysis. Consequently, data sets of

    several authors have been separately studied by the methods

    described inRef. [20]. The results validated by statistical fit

    tests correspond to basic assumptions of the Weibulldistributions[21] and exhibit a strong dependence on the

    maximum stress level Smax(Fig. 4).

    Regarding the strain evolution, experimental data

    available in the literature [18] have been evaluated as

    before with respect to the distribution functions for the total

    strains 1tot (Fig. 5) and found to follow a two parameter

    lognorma1 distribution. The mean values can be calculatedaccording to Eqs. (7) (9), whereas the coefficient of

    variation exhibits dependence on the upper stress level

    Smax. The implementation of the distribution functions into

    the fatigue model causes the corresponding scatter of the

    material state predictions depicted for compression inFig. 5.

    5. Structural simulation under fatigue

    Fatigue load modeling. In contrast to linear approaches,

    damage in general, and fatigue damage in particular,

    immediately make the problem strongly non-linear and

    path-dependent, so that the useful superposition rules for

    load combinations[22]become inapplicable. The aspect of

    representative load histories applicable to non-linear

    systems remains still generally unsolved[23].

    Although both static and dynamic structural responses

    may be considered, this contribution concentrates on slow

    (quasi-static) cyclic loading processes. The fatigue load isassumed to have a constant period and amplitude within a

    Fig. 3. Change of stressstrain relationships for concrete (compression and tension) and steel reinforcement due to fatigue damage.

    Y.S. Petryna et al. / Reliability Engineering and System Safety 77 (2002) 253261256

  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    5/9

    given lifetime interval Ti. These two parameters can,

    however, change between two successive time intervals.

    Then, the duration of time intervals DTi; the amplitudePmax;iand the frequencyaiof cyclic loading constitute three

    potential random variables of the fatigue load history

    consisting ofn intervals:

    Pfat [ni1

    PfatDTi; Pmax;i;ai: 17

    Non-linear finite element analysis. Structural damage

    processes in a rigor treatment are always attached to non-

    linear responses. Non-linear response of an arbitrarystructure discretized by finite elements under static loading

    can be incrementally traced on the basis of the known

    tangent stiffness equation[24]:

    KTV; dDV DlPFIV; d; 18

    whereV, DVdenote the vector of nodal displacements and

    its increment, Dl the increment of the load factor, P the

    vector of nodal forces, KTV; d the non-linear tangentstiffness matrix depending on the current damage extent d.

    FIV; ddenotes the vector-functional of internal forces.For numerical simulation of quasi-static structural

    response processes described by Eq. (18), a large variety

    of NewtonRaphson, BFGS or path-following methods[25,26] offers approved techniques. The structural simulations

    of the present study have been carried out within the

    FEMAS software[27]already used in numerous structural

    damage simulations[28].

    Lifetime simulation. Life assessment of structures either

    implicitly or explicitly deal with two principal time scales.

    The micro-time scale tdescribes usually the processes of

    structural dynamics measured in seconds to hours or simply

    the time-independent structural response. Just in this time

    scale, the limit states may be properly detected. On the other

    hand, fatigue processes last over the macro-scale T

    (lifetime) usually measured in years. Since both scales are

    relevant, but not directly compatible, structural simulations

    currently proceed in turns. The entire life history is

    subdivided by discrete time instants Ti; i 1; ; n intointervals of generally non-equal duration DTi Ti 2 Ti21;

    i 1; ; n 2 1:The fatigue damage Dc,fat is predicted for lifetime

    intervals DTi according to the model presented in Section

    4. After a correction of material properties due to damage at

    the end of each time interval, structural simulations are

    carried out anew:

    KTV; dTiDV DlPFIV; dTi: 19

    Such a simulation strategy provides complete informationnecessary for reliability examinations including stability,

    Fig. 4. Estimation of Weibull parameters and S Ncurve with Weibull distribution.

    Fig. 5. Scatter of strain evolution and of stressstrain relationship.

    Y.S. Petryna et al. / Reliability Engineering and System Safety 77 (2002) 253261 257

  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    6/9

    deflections, strains, stresses, crack widths, etc. during the

    entire service life of the structure under consideration.

    6. Time-variant reliability assessment

    Reliability assessment of structures exposed to fatigue

    loading needs an approach combining methods of non-

    linear structural analysis, damage mechanics and probabil-

    istic reliability theory. A majority of publications on

    reliability degradation due to damage still concerns simple

    eases, when local damage in a single cross-section may be

    directly associated with a global limit state. A typical

    example is fatigue crack growth[17].However, the fatigue

    limit state in reinforced concrete structures is difficult to find

    as a single cause of global failure. Fatigue interacts withother failure mechanisms and can rather influence typical

    limit states of carrying capacity or serviceability. Due to

    enormous computational difficulties of such complex

    problems, a general and unified approach is still missing.

    The present work proposes an assessment strategy appli-

    cable to such complex reliability problems, depicted in

    Fig. 6and described below.

    Since the number of random variables may be crucial for

    efficiency of stochastic reliability assessment, they are

    defined on the basis of preliminary sensitivity analysis and

    describe generally statistical uncertainties of load, material

    and structural origin. Additionally, statistical distributions

    and their eventual changes during the lifetime shall bedefined for each of the selected random variable.

    Among various approaches to reliability of reinforced

    concrete structures, the response surface method (RSM)

    [29] is of advantage, if closed-form models either are not

    available or may be based on too rough structural

    idealizations. The method combines the full-scale non-

    linear structural analysis with statistical simulations by

    means of a uniform interface in the form of limit state points

    calculated deterministically. The latter are simply critical

    combinations of governing parameters X1; ;Xm; whichsatisfy the limit state condition

    gX RX2

    SX 0; 20where R and S denote the conjugate resistance and load

    variables. For example, the critical load Pmax at collapse

    (Fig. 6) and concrete strength fc form a limit state pointX Pmax;fcof two components.

    According to the RSM[29],an analytical approximation

    of the implicit limit state function (20) is searched for,

    currently in the form of the second order polynomial:

    gX a0 Xki1

    biXiXki1

    Xkj1

    cijXiXj: 21

    Having the analytical function (21) at disposal, one can

    further calculate the instantaneous failure probabilities,practically the failure rates hTi; by means of the most

    efficient statistical simulation tools available. The adaptive/

    importance sampling procedures within the COSSAN

    software[30]are currently used.

    The failure rates hTi obtained for instantaneous values

    of the resistanceRTiand the load effect STican then be

    numerically integrated over the macro-timeT[3]providing

    finally the cumulative time-dependent failure probability:

    pfT 1 2 2expT

    0hTdT

    1 2 2expXni0

    hTiDTi

    !" #: 22

    7. Example: concrete arch

    In order to illustrate the entire assessment procedure on a

    possibly simple example, we consider solely the concrete

    response to cyclic compression observed in the arch under

    uniformly distributed normal pressure (Fig. 7). The arch is

    made of concrete with compressive strength fc 35 MPa

    and elasticity modulusEc 33; 200 MPa:It was discretizedfor structural analysis by 20 finite elements.

    The non-linear load displacement diagram computed

    for the mid-span point are shown inFig. 8. Structural failure

    occurs each time due to the overloading of the upper

    concrete layer at mid-span following by an immediate lossof global equilibrium and collapse.

    For efficiency purposes, only two random variables are

    considered in the present example: compression strength of

    concrete fc and external pressure qls. Both variables are

    assumed to be normally distributed with the parameters in

    Table 1.

    The concrete strength is assumed to be uniformlydistributed along the arch in the initial state, it changes

    then according to the fatigue damage model in each point

    individually. The reference pressure q lsis taken stationary,

    i.e. its amplitude does not depend on time.

    The high-cycle fatigue loading is modeled as cyclic

    pressure of constant frequency and the magnitude varyingfrom zero to qfat 2:0 MPa:It amounts to about 0.53qmax,where qmax denotes the ultimate load capacity of the archwithout damage. The stresses in the arch vary at that from

    zero to Smax 0:70fc: The load history is assumed to beregular high-cycle loading over the entire life-span. The

    number of load cycles is taken according to a typical

    European traffic situation measured at Auxerre [31] and

    corresponding to 150 heavy trucks per day and approxi-

    matelyN 32,000 cycles per year.

    In order to estimate residual capacity and reliability, the

    arch has been statically loaded at each of time instants Tiup

    to collapse (Fig. 8). The corresponding reduction of qmaxdue to accumulation of fatigue damage is depicted also inFig. 9.

    Y.S. Petryna et al. / Reliability Engineering and System Safety 77 (2002) 253261258

  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    7/9

    Fig. 6. Concept of structural reliability assessment.

    Y.S. Petryna et al. / Reliability Engineering and System Safety 77 (2002) 253261 259

  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    8/9

    Finally, the instantaneous probability of collapse (failure

    rate)

    hTi pHCFf Ti pqmaxfc; qfat; dTi # qlsTi 23

    is calculated by means of the RSM and depicted inFig. 9.

    The results show a strongly non-linear character of fatigue

    damage evolution in concrete even under regular cyclic

    loading. This fact makes the application of linear accumu-

    lation rules [32] to concrete structures on system level

    problematic.

    8. Conclusions

    Reliability assessment of structures under fatigue

    conditions is a highly complicated problem, which implies

    interaction of different scientific fields such as damage and

    continuum mechanics, non-linear structural analysis andprobabilistic reliability theory. Due to the evident complexity

    of such a general formulation, a majority of publications

    still concerns simple eases of local damage in a single cross-

    section and associate them with a global limit state, such as

    in typical fatigue crack growth models.

    Fatigue evolution in reinforced concrete structures

    exhibit rather global character with multiple cracks and

    large material volumes affected by damage, such that local

    assessment becomes unacceptable. Only a global approach

    taking stress redistribution within the structure and

    interaction of various degradation mechanisms into accountcan provide correct predictions of structural response and

    reliability. The present contribution proposes main elements

    of such an approach.

    A fatigue damage model of reinforced concrete capable

    to simulate arbitrary damage states under cyclic loading has

    been developed. The model is embedded into the elasto-

    plastic material law for concrete with damage component.

    The damage evolution function depending on fatigue strains

    has been fitted to a plenty of fatigue test data available in the

    literature.

    This model has been further incorporated into a new

    structural simulation strategy based on interval assessmentof damage accumulation processes in the lifetime scale. The

    Fig. 7. Geometry and loading of concrete arch.

    Fig. 8. Loaddisplacement diagrams of arch under fatigue.

    Table 1

    Variable Mean value Standard deviation Coefficient of variation

    fc(MPa) 35.0 5.25 0.15

    qls(MPa) 2.1 0.315 0.15

    Fig. 9. Degradation of carrying capacity and increase of failure rate over lifetime.

    Y.S. Petryna et al. / Reliability Engineering and System Safety 77 (2002) 253261260

  • 8/13/2019 [Elsevier] Reliability of Reinforced Concrete Structures Under Fatigue

    9/9

    interaction of instantaneous and long-time damage mech-

    anisms can then be traced over the entire lifetime

    contributing to a prognosis of structural response.

    The time-variant reliability assessment is proposed to

    carry out by means of the RSM combined with time

    integration of the calculated instantaneous failure rates.

    Such an approach advantageously combines deterministic

    damage-oriented structural analysis with probabilistic

    assessment procedures.

    The results obtained in a test problem show that linear

    damage accumulation rules widely used on local level

    deliver incorrect predictions of global structural behavior.

    Acknowledgements

    This work has been carried out within the Projects Cl,C2

    of the Collaborative Research Center 398 (SFB 398) at the

    Ruhr-University Bochum. Financial support of the German

    Science Foundation (DFG) is gratefully acknowledged.

    References

    [1] Schueller GI. Design for durability including deterioration and

    maintenance procedures. Working document. Joint Committee on

    Structural Safety. Zurich: IABSE-AIPC-IVBH; 1990.

    [2] Enright MP, Frangopol DM. Service-life prediction of deteriorating

    concrete bridges. ASCE J Struct Engng 1997;124(3):30917.

    [3] Melchers RE. Structural reliability analysis and prediction. Chiche-ster: Wiley; 1999.

    [4] Clifton JR, et al. Service life predictionstate-of-the-art report.

    Report ACI 365.1R-00. Farmington Hills: American Concrete

    Institute; 2000.

    [5] Saria A, Vesikari E, editors. Durability design of concrete structures.

    Report of RILEM Technical Committee 130-CSL. London: E & FN

    Spon; 1996.

    [6] Hsu TTC. Fatigue of plain concrete. ACI J 1981;78:292 305.

    [7] Park YJ. Fatigue of concrete under random loadings. J Struct Engng

    1990;116(11):322835.

    [8] Oh BH. Fatigue-life distributions of concrete for various stress levels.

    ACI Mater J 1991;88(2):1228.

    [9] Hofstetter G, Mang HA. Computational mechanics of reinforced

    concrete structures. Braunschweig: Vieweg & Sohn; 1995.

    [10] Polling R. Close-to-practice damage-oriented material description ofreinforced concrete for structural analysis. PhD Thesis. Bochum:

    Ruhr-University; 2000 (in German).

    [11] Mazars J, Pijaudier-Cabot G. Continuum damage theoryapplication

    to concrete. J Engng Mech 1989;115(2):34565.

    [12] Lemaitre J. A course on damage mechanics. Berlin: Springer; 1992.

    [13] CEB-FIP Model Code 1990. London: Thomas Telford Services Ltd.[14] Dafalias YF, Popov EP. A model of nonlinearly hardening materials

    for complex loadings. Acta Mech 1975;1(3):17392.

    [15] Bazant ZP. Mathematical modeling of creep and shrinkage of

    concrete. Chichester: Wiley; 1988.

    [16] Bazant ZP, Kim JK. Improved prediction model for time-dependent

    deformations of concrete. Mater Struct 1992;25:1639.

    [17] Yao JTP, Kozin F, Wen YK, Yang JN, Schueller GI, Ditlevsen O.

    Stochastic fatigue, fracture and damage analysis. Struct Safety 1986;

    3:23167.

    [18] Holmen JO. Fatigue of concrete by constant and variable amplitude

    loading. NTH-Trondheim: Division of Concrete Structures; 1979.

    [19] Ellingwood BR. Fatigue reliability. J Struct Div 1982;108(ST1):

    323.

    [20] Wirsching PH, Yao JTP. Statistical methods in structural fatigue.

    J Struct Div 1970;96(ST6):120119.

    [21] Gumbel EJ. Parameters in the distribution of fatigue life. Engng Mech

    Div 1963;89(EM5):563.

    [22] Wen YK. Structural load modeling and combination for performance

    and safety evaluation. Amsterdam: Elsevier; 1990.

    [23] Petryna Y, Pfanner D, Stangenberg F. Adapted probabilistic

    assessment of deteriorating RC structures. Proceedings of the 2nd

    European Conference on Computational Mechanics, ECCM 2001,

    Cracow, Poland; June 2629, 2001. Contr. No. 7.

    [24] Kratzig WB, Petryna YS. Assessment of structural damage and

    failure. Arch Appl Mech 2001;71:115.

    [25] Bathe KJ. The finite element procedures. New Jersey: Prentice-Hall;

    1996.

    [26] Zienkiewicz OC, Taylor RL. The finite element method. Solid

    mechanics, vol. 2. Oxford: Butterworth-Heinemann; 2000.[27] Beem H, Konke C, Montag U, Zahlten W. FEMAS-2000 users

    manual. Release 3.0. Bochum, Germany: Institute for Statics and

    Dynamics, Ruhr-University; 1996.

    [28] Harte R, Kratzig WB, Noh SY, Petryna YS. On progressive damage

    phenomena of structures. Comput Mech 2000;25:40412.

    [29] Schueller GI, Pradlwarter HJ, Bucher CG. Efficient computational

    procedures for reliability estimates of MDOF systems. Int J Non-

    Linear Mech 1991;26(6):96174.

    [30] Schueller GI. COSSAN users manual, stand-alone toolbox.

    Innsbruck, Austria: Institute of Engineering Mechanics; 1996.

    [31] Merzenich G, Sedlacek G. Hintergrundbericht zum Eurocode 1Teil

    3.2: Verkehrslasten auf Straenbrticken. Forschung Straenbau und

    Strassentechnik. Heft 711. Bundesministerium fur Verkehr: Bonn-

    Bad Godesberg; 1995.

    [32] Hashin Z. A reinterpretation of the Palmgren-Miner rule for fatigue

    life prediction. J Appl Mech 1980;47:3248.

    Y.S. Petryna et al. / Reliability Engineering and System Safety 77 (2002) 253261 261