Characterisation of long-term strength of structural materials is an important engineering task for prevention of potential catastrophic failures of critical equipment. However, studies of this type are usually very long-lasting, technically challenging and involve expensive experimental work. Thus, the main scope of this study is the formulation of a simple way to predict characteristics of the longterm material behaviour under creep and fatigue conditions using basic material properties. Based upon the extensive availability of experimental material data, a significant progress toward this challenge has been achieved so far and may be observed in the literature as indicated below.
One should start with the fundamental work of Bäumel and Seeger , who discussed a few methods for estimating fatigue behaviour of metals using strain-life approach on the basis of monotonic test results. They included the Method of Universal Slopes (MUS) proposed by Manson  and modification of MUS developed by Muralidharan & Manson , which are applicable to all metals. However, this approach contained a critical limitation regardless of a good accuracy, which is the requirement of reduction in area ψ availability.
The flaw of MUS formulations induced Bäumel and Seeger  to develop the Uniform Material Law (UML) using as its basis the extensive collection of fatigue data  with over 1500 experimental results. Unalloyed and low alloy steels were analysed separately from aluminium and titanium alloys resulting into two sets of equations both being based on only elasticity modulus E and ultimate tensile strength (UTS) σu, which can be easily correlated with Vickers hardness HV. Later, the applicability of UML concept was extended to high-strength steels by Korkmaz .
Comparative study by Kim et al.  evaluated seven basic methods for estimating uniaxial fatigue properties (including fatigue limit ) from tensile properties or hardness. This study was based upon the fatigue test data for eight ductile steels under axial and torsional loading. Three of the evaluated methods were able to predict over 93% of the test cases within a factor of 3 compared to the observed lives. The formulas for prediction included mechanical properties such as E, σu and true fracture ductility ɛf. Among the variety of empirical formulations for prediction with different combinations of aforementioned mechanical properties, the simplest are based on σu : = 1.9018 σu (MUS); = 1.5 σu (UML); and = σu + 345 MPa (Mitchell’s method), which showed an accuracy of R2 = 0.88. Another simple method in this comparison, proposed by Roessle & Fatemi , used a Brinnell hardness HB for prediction as = 4.25 HB + 225 MPa. This approach showed a reasonable accuracy of R2 = 0.86 for the experimental data fit.
Casagrande et al.  investigated a relationship between and Vickers hardness HV in steels and developed a method to predict . A good correlation was observed between HV and for four kinds of steels in different metallurgical states. However, the proposed empirical method is not straightforward and involves a number of parameters and equations to achieve a reasonable of accuracy of predictions. Recently, Bandara et al.  proposed a formula for predicting of steels in the Giga-Cycle Fatigue (GCF) regime. It uses a combination of σu and HV as material parameters and was verified using the experimental results for 45 steels.
An alternative approach was suggested by Li et al. , who estimated theoretically the cyclic yield strength and using the test data for 27 alloy steels. One formula expresses by two conventional mechanical performance parameters - σu and the reduction in area ψ. The other formula expresses the fatigue endurance limit through the cyclic yield strength with a reasonable accuracy of R2 = 0.883 as of = 1.13 . Despite the relative simplicity, the proposed relation can’t be considered as mathematically elegant, most probably because of the conventional assumption of 0.2% plastic strain offset for yield strength. Nevertheless, this formula by Li et al.  demonstrated the tendency that is not very different from .
A computational approach was developed by Tomasella et al. , who applied the Artificial Neural Networks to estimation of the cyclic material properties used in strain-life fatigue approach from a set of monotonic material properties. This approach was implemented into the software called Artificial Neural Strain Life Curves (ANSLC), and has been tested on a large database of steels [1, 4]. In comparison with the largely used UML , the results of the estimation with ANSLC program, even without including the support of real experimental tests in the regression, showed a considerably higher accuracy in the life-time estimation.
Recently, Pang et al.  did a comprehensive review of the relations between and other mechanical properties of metallic materials. They include the qualitative / quantitative relations between and hardness (HV or HB), strength (σu and σy) and toughness. Analysis of the numerous fatigue data resulted in the General Fatigue Formula (GFF): = σu (C − P · σu), where C and P are fitting parameters. GFF was found applicable to prediction with increasing σu in a wide range for many materials such as conventional metals and newly developed alloys. Pang et al.  suggested that GFF can provide a new clue to predict and select the appropriate materials within engineering design by adjusting parameters P and C adequately.
The concept of the fatigue limit has been comprehensively studied on microstructural scale by Bathias . This experimental and theoretical study was based upon a thorough and authoritative examination of the coupling between plasticity, crack initiation and heat dissipation for lifetimes that exceed the billion cycle. Moreover, the validity of GCF concept was proved, what questions the idea of infinite fatigue life, at least for practical applications.
Less progress has been achieved in the methods for creep rupture strength evaluation, but recently an important observation was discovered by Kimura . The creep strength of ferritic and austenitic steels has been investigated in  through the correlation between creep rupture curve, presenting stress vs. creep rupture life, and 50% of 0.2% offset yield stress (half yield) at a wide range of temperatures. The inflection of the creep rupture curve at half yield was recognised for ferritic creep resistant steels with martensitic or bainitic microstructure, e.g. T91, T92 and T122. This was explained in terms of different mechanisms of microstructural evolution during creep at high- and low-stress regimes. The purpose of this study was to point out a significant risk of overestimation of long-term creep rupture strength by extrapolating the data for martensitic and bainitic steels (e.g. ASTM T91/P91) in high-stress regime to low-stress regime, which are separated by half yield.
A similar problem with particular application to ASTM P91 steel was investigated and discussed earlier by Gorash et al. [15-17] for the purpose of a creep constitutive model development. In these works, apart from inflection of the creep rupture curve, the simultaneous inflection of the minimum creep rate curve, presenting minimum creep rate vs. stress, was recognised. Alternation of the minimum creep rate slope was explained in terms of different creep deformation mechanism (linear creep for low stress and power-law for high stress), while alternation of the creep rupture life slope was explained in terms of different damage accumulation modes (brittle fracture for low stress and ductile for high stress). The inflection of both curves was characterised by the same value called transition stress, which had the meaning of material parameter in the developed “double power law” creep model. However, was identified in [15-17] using the minimum creep rate data, and no relation of to basic mechanical properties of ASTM P91 steel was recognised.
The principal aim of the present study is to investigate a link in characterisation of long-term strength of structural steel by finding a similar quantative feature in available experimental data. This work establishes relationships between characteristics of creep and fatigue behaviour on one hand and yield strength as a basic material property and characteristic of plasticity on other hand.
2 Concept of the safe structural design
2.1 Definition of the yield strength
Dowling  discusses several methods to characterise the yield strength σy. The first is the proportional limit , which is the stress where the first departure from linearity occurs. The second is the elastic limit , which is the highest stress that does not cause plastic deformation. The third is the offset yield strength , which is the stress in the point on stress-strain curve typically defined by the plastic strain offset of 0.2% from elastic line. This value is generally the most practical means of defining the yielding event for engineering metals. Therefore, is usually meant to define the yield strength σy in the literature. However, here the elastic limit , defined in the scope of unified Chaboche model [19, 20], is used as the yield strength σy.
This study proposes the cyclic yield strength as a key characteristic for the definition of safe design for engineering structures operating under fatigue and creep conditions, as illustrated in Fig. 1. It is conventionally defined in context of a cyclic stress-strain curve (SSC), which is obtained from results of cyclic tests for a number of different strain ranges. Each cyclic test produces a stabilised stress response, which is affected either by hardening or by softening depending on the type of steel. In case of steels with a cyclic softening effect, separates the low stress range of purely elastic behaviour from the moderate stress range of mixed elasto-plastic behaviour. Monotonic yield strength , which is conventionally defined in context of a monotonic SSC, separates the moderate stress range of mixed elasto-plastic behaviour from the high stress range of purely plastic behaviour. Both values of and are identified using a 2-steps fitting procedure of the experimental stress-strain curves. The first step applies the Ramberg-Osgood material model, which produces basic smoothing and extrapolation, to the both monotonic and cyclic SSCs separately. The second step of fitting involves a typical rate-independent form of the Chaboche material model with 3 kinematic backstresses. Fitting the Chaboche model with two separate sets of material constants sequentially to the both SSCs provides the values of and with minimum offset from the elastic line as elastic limits.
2.2 Stress-strain curves fitting procedure
As experimental SSCs usually demonstrate some level of scatter, the first step in data fitting for the material parameters identification is the basic curve smoothing. The conventional Ramberg-Osgood (R-O) equation  is optimal for such curve smoothing since it was formulated to describe the non-linear relationship between stress and strain in materials near their yield point. It is particularly useful for metals that harden or soften with plastic deformation showing a smooth elastic-plastic transition. The equations for monotonic and cyclic SSCs are: (1) where Δɛtot is the total strain range and Δσ is the total stress range (MPa) for each cyclic test respectively; B and β are the R-O material parameters; and Young’s modulus E in MPa. Using the value of E, the total strain ɛtot in the experimental curves is decomposed into elastic and plastic strain. Then the plastic component ɛp of strain is fitted using the the least squares method by the following power-law relations, which are derived from the Eq. (1): (2) The resultant R-O fits for monotonic and cyclic curves are then used to identify the parameters for the Chaboche material model. The range of applicability for the R-O fit is usually quite narrow not exceeding 1% of ɛtot depending on the grade of curvature grade for a SSC.
The basic variant of the rate-independent Chaboche model [19, 20] is presented as a combination of nonlinear kinematic hardening and nonlinear isotropic hardening models. The model allows the superposition of several independent backstress tensors and can be combined with any of the available isotropic hardening models. Since in this study monotonic and cyclic SSCs are fitted separately only for the identification of σy, only the kinematic hardening component is considered: (3) where is the plastic strain rate, and ṗ is its magnitude. The total backstress X in Eq. (3) is given by the superposition of a number N of kinematic backstresses Xi with a corresponding evolution equation initially proposed by by Armstrong & Frederick  for Ẋi, where Ci and γi are kinematic material constants. Chaboche et al.  recommended N = 3 in order to provide a good fit of experimental SSCs, which include large strain areas. Therefore, three backstresses are considered in this study providing an excellent match of the R-O fit (1) for a whole range of strains.
The kinematic hardening constants (Ci, γi) and σy, which define the size of the yield surface, are identified as recommended in . The cyclic SSC is fitted by the following relation: (4) which is obtained in  by integrating Eq. (3) and considering ɛp ≈ const at the peak stresses for strain-controlled cyclic loading. Relation (4) is valid for the cyclic curve after stabilisation of the hardening or softening effects. Constants (Ci, γi and cyclic ) are identified by automatic fitting Eq. (4) to the R-O extrapolation (2) with “cyclic” values of constants B and β. The identification procedure is implemented in Microsoft Excel using an add-in Solver . The Solver searches for an optimal (minimum in this case) value for a formula in one cell - called the objective cell - subject to constraints, or limits, on the values of other formula cells on a worksheet. The Solver works with a group of cells, called decision variables or simply variable cells, that participate in computing the formulas in the objective and constraint cells. In this case, the Solver adjusts the values in the decision variable cells containing material constants (Ci, γi and ) in order to minimise the value in the objective cell. This cell contains an average value of the absolute difference between columns containing calculated by Eq. (2) and Eq. (4) correspondingly in a particular range of Δɛp. Applying this approach, an excellent match of Eqs (2) and (4) is achieved.
The monotonic SSC is fitted by the different relation in the following form : (5) which contains the monotonic and different values of kinematic hardening constants (Ci, γi). These constants are identified by fitting Eq. (5) to the R-O extrapolation (2) with “monotonic” values of the R-O parameters B and β. The identification procedure is implemented in Microsoft Excel using an add-in Solver in the same way as for cyclic SSC. An advanced step-by-step guideline for the estimation of the Chaboche viscoplasticity model parameters with their further optimisation was developed by Hyde et al. .
2.3 Application to three structural steels
The above described fitting procedure is applied to SSCs of three structural steels for the purpose of and identification. The first is ASTM A36 steel, with mechanical properties available in  and , which is a standard low carbon steel, without advanced alloying and is a principal carbon steel employed for bridges, buildings, and many other structural uses. The monotonic SSC for this steel at room temperature (RT) shown in Fig. 2a exhibits perfectly plastic behaviour when reaching the stress of 36 ksi = 248.211 MPa in average, which is considered as . The perfectly plastic yielding lasts for approximately of ɛp = 1(%) of strain plateau, which is followed by the strain hardening area, then gradually approaching failure at ɛtot = 30(%). The cyclic SSC for this steel shown in Fig. 2a from  is fitted by the 2-step procedure, and the obtained material parameters for the R-O (1) and Chaboche (3)-(5) models are listed in Tables 1 and 2 correspondingly.
The second material is AISI 4340 steel , a high-strength alloy steel, which has good machinability features and used for a wide range of applications including aircraft landing gears, shafts or axels for power transmission, gears, high pressure pump housings, etc. Both monotonic and cyclic SSCs shown in Fig. 2b and mechanical properties are taken from . Since it is available explicitly, the monotonic SSC is fitted by the Chaboche model (5) directly, and the material parameters are listed in Table 2. The cyclic SSC for this steel shown in Fig. 2b from  is available at ten times wider strain range than for the ASTM A36 steel. Therefore, the R-O model (1) is not able to provide an accurate fit of the cyclic SSC. In this case, the following modification of the R-O equation (1) proposed by Lemaitre & Chaboche  is used for fitting analysis: (6) Compared to Eq. (1), this notation contains an additional parameter of the yield strength σy in the meaning of , and can be applied for an accurate fitting of much wider strain range than Eq. (1). Thus, the cyclic SSC is fitted by the 2-step procedure. The obtained material parameters for the modified R-O (6) and Chaboche (3)-(5) models are listed in Tables 1 and 2 correspondingly.
The third material is ASTM P91 (modified 9Cr-1Mo) steel [23, 30], an advanced ferritic steel with martensitic microstructure, which has already been widely used over the last 2 decades as tubes/pipes for heat exchangers, plates for pressure vessels, and other forged, rolled and cast components for high temperature services. Both monotonic and cyclic SSCs shown in Figs 2c and 2d and mechanical properties at room temperature (RT), 500°C, 550°C, 600°C and 650°C are taken from . Firstly, the monotonic SSCs are presented in  by the material parameters for the R-O model (1) listed in Table 1. The cyclic SSCs are presented in  by raw data, which is fitted by the R-O model (1) with material parameters listed in Table 1. Secondly, both monotonic and cyclic R-O extrapolations are fitted by the Chaboche model (3)-(5) with material parameters listed in Table 2.
3 Relation in mechanical characteristics
The next step is a search for possible correlations between the experimentally obtained yield strength values ( and ) for ASTM A36, AISI 4340 and ASTM P91 steels and their fatigue and creep behaviour characteristics. This identifies a clear similarity for characteristic transition stresses in S-N fatigue, minimum creep strain rate and creep rupture curves, as explained below.
3.1 Fatigue behaviour at normal temperature
Engineering structures operating under cyclic loading conditions at normal temperature are usually designed against the fatigue failure using the conventional stress-life approach. This approach involves experimental fatigue S-N curves with a number of cycles to failure N* vs. stress. A typical S-N curve is a straight line in double logarithmic coordinates, which conventionally described by the Basquin model . Referring to [18, 27], a distinctive minimum stress lower bound, which is called a fatigue endurance limit , is observed on S-N curves for a number of structural steels with polished surface of a specimen in benign (non-corrosive and RT) environment. Referring to , fatigue limit, endurance limit and fatigue strength are all expressions used to describe a property of materials under cyclic loading: the amplitude (or range) of cyclic stress that can be applied to the material without causing fatigue failure. In these cases, a number of cycles (usually 107) are chosen to represent the fatigue life of the material.
Comparison of defined as a material constant and experimentally observed reveals that they are close. This assumption is confirmed by the high-cycle fatigue (HCF) experimental data for ASTM A36  and AISI 4340 [32-34] steels shown in Fig. 3. Comparison of with summarised in Table 4 for ASTM A36 steel gives 27.6% accuracy and 5.5% accuracy for AISI 4340 steel. These observations indicate that the safe fatigue design is guaranteed in the purely elastic domain defined by .
In general, each S-N curve exhibits two limits: one, when stress tends towards the static fracture σu (fracture in a quarter of the cycle), and the other, when stress tends towards the fatigue limit σlim. The most known concepts able to describe a reverse sigmoidal shape of a generic S-N curve are presented by two models. The conventional one is the Bastenaire model [37,38]: (7) where σa is an alternating stress. Three material parameters A, B, C and fatigue limit σlim are derived from fitting the experimental raw data.
A more advanced formulation was developed by Lemaitre & Chaboche  using the damage mechanics approach: (8) where σ̄ is the mean stress of the cycle; σmax is the maximum stress in the cycle; and the other variables are material parameters defined in the material property set: a - non-linear damage sensitivity, b - mean stress correction factor, c - the Chaboche equation coefficient, and α - the Chaboche equation exponent. Because the Chaboche concept incorporates its own mean stress σ̄ correction  resembling the Goodman method, the Eq. (8) is fitted to a family of S-N curves with different stress ratios R.
For the purpose of this study, the following equation for a S-N curve proposed in  was used as a basis: (9) where N* is a number of cycles to fatigue failure, σa is an alternating stress, and f1, f2 and f3 are fitting parameters. This equation provides a smooth transition of the S-N curve into the horizontal plateau of . Hence, it requires only introduction of the UTS σu, which is implemented as (10)
it produces a fully adjustable reverse sigmoidal shape with a mathematical minimum of fitting parameters;
it contains an upper and lower bounds as σu and ;
it is fully reversible.
Using a previously suggested assumption, a S-N curve lower bound in Eqs (7, 8,10) can be replaced with . This would let to identify two material parameters (σu and ) from tensile and cyclic experiments correspondingly rather than from fatigue tests. The efficiency of Eq. (10) for description of S-N curves is demonstrated in Fig. 3 in application to ASTM A36 steel with σu from  and AISI 4340 steel with σu from . For both steels is taken from Table 2. Identification of the fatigue fitting parameter (f1, f2 and f3) is implemented in Microsoft Excel using an add-in Solver in the same way as described previously for fitting of cyclic and monotonic SSCs by the Chaboche model. All parameters for the S-N curves are reported in Table 3. Finally, the whole range of mathematical formulations for S-N curves is discussed by Castillo & Fernández-Canteli .
3.2 Creep behaviour at elevated temperature
Engineering structures operating under constant loading conditions at high temperature are usually designed against the creep failure using the conventional time-to-failure approach. This approach involves experimental creep rupture curves with stress vs. time to failure t*. A typical creep rupture curve is a trilinear smoothed curve in double logarithmic coordinates, with two inflections corresponding to and . These inflections separate three sections on the creep rupture curve, which are characterised by three different creep damage accumulation modes - brittle, ductile and mixed. Three sections with different creep deformations mechanisms can be typically observed on the minimum creep rate curve, presenting minimum creep strain rate vs. stress, which is also a trilinear smoothed curve in double logarithmic coordinates. The deformations mechanism (linear creep, power-law creep and power-law breakdown) are separated by the same two inflections. This assumption is confirmed by the experiments for ASTM P91 steel at elevated temperatures and corresponding theoretical developments.
Previously, the creep modelling using different creep exponent values in different stress ranges was studied by Gorash et al. [15-17]. These studies were devoted to the formulation of constitutive creep model, called double power law and applied to ASTM P91 steel at 600°C: (11) where = 100 MPa is a material parameter called “transition stress”, which characterises a transition from linear creep to power creep. Other material parameters were identified as n = 12 and C = 2.5 • 10-7 MPa-1 /h.
To implement two transitions into the constitutive model, Eq. (11) is modified by adding the third “power-law breakdown” component to become the triple power law: (12) where is a transition stress, C and n are secondary creep parameters, which have the temperature dependence expressed by Arrhenius-type functions as follows : (13) with σ* = 0.0587 [MPa] and Qσ = 54100 [J/mole], (14) with σC = 1.9916 [1/h] and QC = 8757.1 [J/mole], (15) with n0 = 0.2479 and Qn = 28648.4 [J/mole].
In notations (13)-(15): T is a temperature in K; Qσ, QC and Qn are creep activation energies and R = 8.314 [J • mol-1 • K-1] is the universal gas constant. The transition stress (T) and creep parameters C(T) and n(T) were obtained by fitting the data for minimum creep strain rate from studies by Sklenička et al. , Kloc & Fiala  and Kimura  and shown in Fig. 4. The inflections of corresponding curves are observed at 550, 600 and 650°C and explained in terms of transition between different creep deformation mechanisms. Comparison of defined by Eq. (13) and reported in Table 4 with from Table 2 reveals that they are close. As summarised in Table 4, the accuracies are 26.7, 6.4 and 19.2% corresponding to 550, 600 and 600°C.
Another important observation was done in [15-17] for creep rupture behaviour of this steel. The creep-rupture curve, which describes the dependence of time to rupture on stress, exhibits the distinctive inflection too. This inflection is explained as a transition from ductile character of rupture to brittle. This transition was observed at the approximately same stress as the inflection of min. creep rate − = 100 MPa at 600°C. Recently, Kimura  observed inflections of the experimental curves at 600 and 650° C as shown in Fig. 5 and explained them in terms of half monotonic yield (/2). In contrast to Kimura’s idea , in this study, inflections are defined by using Eq. (13). The creep constitutive model (12) is inserting into the Monkman-Grant relationship as in previous work  to produce the time to rupture t*: (16) where k1 = 0.23 [1/h]k2-1 and k2 = 0.83 are the tertiary creep constants for the ASTM P91 steel, which are identified by fitting equation (16) to the creep-rupture experimental data  for the temperature range from 550° C to 650°C.
It should be noted that Dimmler et al.  associated these inflections with the microstructurally determined threshold stresses (back-stress concept). The applicability of this concept was shown using the experimental minimum creep rate and creep rupture curves for several 9-12%Cr heat resistant steels (P91, GX12, NF616, X20 and B2). Dimmler et al.  emphasised that the knowledge of these threshold stresses limits the range of experimentally based predictions, thus preventing from overestimation of the long-term creep rate and creep strength from extrapolated short-term creep data.
Since the inflections are captured reasonably well on both types of creep data in Figs 4 and 5, the transition stresses on min. creep rate curves and creep rupture curves proposed by Gorash et al. [15-17] can be explained by relating them to . Therefore, these observations prove that the most safe creep design is guaranteed in linear creep domain with brittle failure mode, which is also defined by the .
3.3 Fatigue behaviour at elevated temperature
The fatigue performance of ASTM P91 steel is analysed using the HCF experimental data by Matsumori et al.  at three different temperatures (RT, 400 and 550° C) illustrated in Fig. 6. From these data, it can be concluded that at elevated temperatures the heat-resistant steels don’t exhibit on S-N fatigue curves, which is usually observed at normal temperature. The reason for this is the elimination of purely elastic behaviour at high temperature, since there is always some amount of inelastic strain, which is caused by creep. Therefore, there is always a permanent accumulation of creep damage, even at low stress levels and high-strain rate, which leads to inevitable failure. This fact is confirmed by the experimental observations , which demonstrated the extinction of at 550° C for over 108 loading cycles. However, a good match of with with accuracy of 2.8% is observed at RT for this steel as shown in Table 4, which makes advanced martensitic steels different from simple ferritic steels for prediction. This effect can be explained by the assumption of Terent’ev , who recognised two types of the fatigue endurance limit - standard in HCF range (N = 102-107 cycles) and ultrahigh in giga-cycle fatigue (GCF) range (N = 107-1011 cycles). The existence of ultrahigh was proved by the experimental data for high-strength steels (50CrV4,54SiCrV6 and 54SiCr6), which demonstrated two inflections of the fatigue curves followed by horizontal plateaus - first in HCF area (N ≈ 105-106), second in GCF area (N ≈ 108-109). The correspondence of with ultrahigh for ASTM P91 steel is expected to be found at N > 108 cycles, but no experimental data is available for this range.
Following these assumptions, experimental S-N curves for ASTM P91 steel by Matsumori et al.  are described by the Eq. (10), where is replaced by as shown in Fig. 6. The experimental values of σu for all temperatures are taken from . The value of for RT is taken from Table 2, while from  is taken instead of for 400° C, and = 0 is assumed for 550° C because of creep. All parameters for the S-N curves are reported in Table 3.
Finally, the values of and for ASTM P91 steel defined by Chaboche model as shown in Figs 2c and 2d are listed in Table 4. They are plotted versus temperature in Kelvins (K) in Fig. 7. The temperature dependence of a yield strength defined as is extrapolated by a simple elliptic equation, which can be assumed as an extension of the von Mises yield criterion by temperature consideration, in the following form: (17) where Teut = 1000 K is a typical eutectic temperature for steel alloys; = 210 [MPa] and = 2 · = 420 [MPa] are theoretical yield strengthes at absolute zero temperature for monotonic and cyclic responses correspondingly. Since it is not possible to measure the values of and experimentally, they can be estimated using the results of only one experimental measurement (for instance, at RT) in Eq. (17). However, the value of Teut = 1000 K is proved experimentally to be an eutectic temperature in the iron-carbon system, which characterises the coexistence of solid and liquid phases on iron-carbon phase diagram.
This study explains the existence of the fatigue limit and creep transition stress by the cyclic yield strength using the fatigue and creep experimental data for a few structural steels at normal and elevated temperatures. The comparison of yield strengths (, ), and for ASTM A36, AISI 4340 and ASTM P91 steels is summarised in Table 4. Monotonic and cyclic yield strengths (, ) are defined as elastic limit specified in the scope of the Chaboche model [19, 20]. Fatigue limit of ASTM A36 and AISI 4340 steels complies with . Equality of and at RT for ASTM P91 steel can be explained by the GCF concept  introducing two fatigue limits.
Creep transition stress of ASTM P91 steel complies with . Moreover, Kimura’s assumption  of half monotonic yield (/2) agrees very well with the outcomes of the current study. According to Table 4, the relation ≈ /2 is valid for all temperatures except the highest 650° C regarding ASTM P91 steel. This assumption is not relevant to AISI 4340 steel, which exhibits ≈ .
An important finding is that the temperature dependence of yield strengths (, ) resembles the von Mises yield criterion, which is elliptic in terms of the principle stresses. In the proposed formulation in form of Eq. (17), the yield surface is also presented by ellipse in coordinates of yield strength and temperature as shown in Fig. 7.
The principal advantage of the application to the characterisation of fatigue and creep strength is the relatively fast experimental identification. The total duration of all cyclic tests, which are required to reach the stabilised stress response for the construction of cyclic SSC is much less than the typical durations of fatigue and creep rupture tests at stress levels around .
The critical point in the work presented here is an application of the advanced material model [19,20] to the estimation of a single value of elastic limit , which may seem to be complicated. However, this approach is effective in typical cases when experimental SSCs are unavailable in explicit form, but available in the form of R-O  fittings using Eq. (1). In other cases, when all necessary experimental SSCs are available in form of raw data, the modified form (6) of the R-O model proposed by Lemaitre & Chaboche  may reduce the fitting procedure just to one step. Since Eq. (6) contains σy as a material parameter, the application of Chaboche model equations (3)-(5) may no longer be needed.
Bäumel A., Seeger T., Materials data for cyclic loading: Supplement 1, no. 61 in Materials Science Monographs, Elsevier Science Publishers, Amsterdam, Netherlands, 1990 Google Scholar
Boller C., Seeger T., Materials Data for Cyclic Loading: Volumes 1-5, no. 42a-42e in Materials Science Monographs, Elsevier Science Publishers, Amsterdam, Netherlands, 1987 Google Scholar
Korkmaz S., Uniform Material Law: Extension to high-strength steels. A methodology to predict fatigue life of high-strength steels, VDM Verlag, Saarbrücken, Germany, 2010 Google Scholar
Kim K.S., Chen X., Han C., Lee H.W., Estimation methods for fatigue properties of steels under axial and torsional loading, Int. J. Fatigue, 2002, 24(7), 783-793 Google Scholar
Roessle M.L., Fatemi A., Strain-controlled fatigue properties of steels and some simple approximations, Int. J. Fatigue, 2000, 22(6), 495-511 Google Scholar
Bandara C.S., Siriwardane S.C., Dissanayake U.I., Dissanayake R., Fatigue strength prediction formulae for steels and alloys in the gigacycle regime, Int. J. Mat. Mech. Man., 2013,1(3), 256- 260 Google Scholar
Li J., Sun Q., Zhang Z.-P., Li C.-W., Qiao Y.-J., Theoretical estimation to the cyclic yield strength and fatigue limit for alloy steels, Mech. Res. Commun., 2009, 36(3), 316-321 CrossrefWeb of ScienceGoogle Scholar
Tomasella A., el Dsokia C., Hanselka H., Kaufmann H., A computational estimation of cyclic material properties using Artificial Neural Networks, Procedia Engineer., 2011,10, 439-445 CrossrefGoogle Scholar
Pang J.C., Li S.X., Wang Z.G., Zhang Z.F., Relations between fatigue strength and other mechanical properties of metallic materials, Fatigue Fract. Eng. Mater. Struct., 2014, 37(9), 958-976 CrossrefWeb of ScienceGoogle Scholar
Bathias C., Fatigue limit in metals, Wiley-ISTE, London, UK, 2014 Google Scholar
Kimura K., Creep rupture strength evaluation with region splitting by half yield, in: Proc. ASME 2013 PVP Conf., no. 97819 in PVP2013, ASME, Paris, France, 2013, 1-8 Google Scholar
Gorash Y., Development of a creep-damage model for nonisothermal long-term strength analysis of high-temperature components operating in a wide stress range, PhD thesis, Center of Engineering Sciences, Martin-Luther-University Halle-Wittenberg, Halle (Saale), Germany, 2008 Google Scholar
Altenbach H., Gorash Y., Naumenko K., Steady-state creep of a pressurized thick cylinder in both the linear and the power law ranges, Acta Mechanica, 2008,195(1-4), 263-274CrossrefWeb of ScienceGoogle Scholar
Gorash Y., Development of a creep-damage model for a wide stress range, SVH-Verlag, Saarbrücken, Germany, 2015 Google Scholar
Dowling N.E., Mechanical Behavior of Materials: Engineering Methods for Deformation, Fracture, and Fatigue, 4th ed., Pearson Education Limited, Harlow, UK, 2013 Google Scholar
Chaboche J.-L., Dang Van K., Cordier G., Modelization of the strain memory effect on the cyclic hardening of 316 stainless steel, in: Trans. 5th Int. Conf. on Structural Mechanics in Reactor Technology, no. L11/3 in SMiRT5,1-10, IASMiRT, Berlin, Germany, 1979 Google Scholar
Higashida Y., Lawrence F.V., Strain controlled fatigue behavior of weld metal and heat-affected base metal in A36 and A514 steel welds, FCP Report No. 22, University of Illinois, Urbana, Illinois, USA, 1976 Google Scholar
Smith R.W., Hirschberg M.H., Manson S.S., Fatigue behavior of materials under strain cycling in low and intermediate life range, Technical Note No. D-1574, NASA, Cleveland, Ohio, USA, 1963 Google Scholar
NRIM Fatigue Data Sheet No. 78, Data sheets on elevated- temperature, time-dependent low-cycle fatigue properties of ASTM A387 Grade 91 (9Cr-1Mo) steel plate for pressure vessels, Tech. rep., National Research Institute for Metals, Tokyo, Japan, 1993 Google Scholar
Ramberg W., Osgood W.R., Description of stress-strain curves by three parameters, Technical Note No. 902, NASA, Washington DC, USA, 1943 Google Scholar
Armstrong P.J., Frederick C.O., A mathematical representation of the multiaxial Bauschinger effect, Report no. RD/B/N731, CEGB, Berkeley, UK, 1966 Google Scholar
Microsoft® Office Professional Plus, Excel Help System // Analyzing data // What-if analysis // Define and solve a problem by using Solver, Microsoft Corp., Release 2010 ed., 2009 Google Scholar
Hyde T., Sun W., Hyde C., Applied Creep Mechanics, McGraw-Hill Education, New York, USA, 2004 Google Scholar
ASTM Standard, Standard Specification for Carbon Structural Steel, no. A36/A36M - 08 in ASTM, ASTM International, West Conshohocken, USA, 2008 Google Scholar
Lemaitre J., Chaboche J.-L., Mechanics of Solid Materials, Cambridge University Press, Cambridge, UK, 1994 Google Scholar
Kimura K., Kushima H., Sawada K., Long-term creep deformation property of modified 9Cr-1Mo steel, Mater. Sci. Eng. A, 2009, 510-511, 58-63 Google Scholar
Wang X.G., Crupi V., Guo X.L., Zhao Y.G., Quantitative thermographic methodology for fatigue assessment and stress measurement, Int. J. of Fatigue, 2010, 32(12), 1970-1976 CrossrefWeb of ScienceGoogle Scholar
Boyer H.E., Atlas of Fatigue Curves, ASM International, Materials Park, Ohio, USA, 1986 Google Scholar
Dowling N.E., Mean stress effects in Stress-Life and Strain-Life fatigue, SAE Technical Paper, 2004, 2004-01-2227, 1-14 Google Scholar
Matsumori Y., Nemoto J., Ichikawa Y., Nonaka I., Miura H., High cycle fatigue properties of modified 9Cr-1Mo steel at elevated temperatures, in: Proc. ASME 2012 Int. Mechanical Engineering Congress & Exposition, no. 87329 in IMECE2012, ASME, Houston, Texas, USA, 2012, 85-89 Google Scholar
Basquin O., The exponential law of endurance tests, Proc. ASTM, 1910,10(II), 625-630 Google Scholar
Bastenaire F., New method for the statistical evaluation of constant stress amplitude fatigue-test results, in: R.A. Heller (Ed.), Probabilistic Aspects of Fatigue, no. STP 511 in ASTM special technical publication, chapter 1, 3-28, ASTM, Philadelphia, USA, 1972 Google Scholar
Newbold P., ANSYS® nCode DesignLife Theory Guide, HBM nCode, Catcliffe, Rotherham, UK, Release 15.0 ed., 2013 Google Scholar
Newbold P., ANSYS® nCode DesignLife Worked Examples, HBM nCode, Catcliffe, Rotherham, UK, Release 15.0 ed., 2013 Google Scholar
Castillo E., Fernández-Canteli A., A unified statistical methodology for modeling fatigue damage, Springer, Berlin, Germany, 2009 Google Scholar
Kloc L., Sklenička V., Transition from power-law to viscous creep behaviour of P-91 type heat-resistant steel, Mater. Sci. Eng. A, 1997, A234-A236, 962 - 965 Google Scholar
Sklenička V., Kuchařová K., DlouhýA., Krejčí J., Creep behaviour and microstructure of a 9cr steel, in: D. Coutsouradis et al. (Ed.), Proc. Conf. Materials for Advanced Power Engineering 1994: Part I, Kluwer Academic Publishers, Liege, Belgium, 1994, 435-444 Google Scholar
Gorash Y., Altenbach H., Lvov G., Modelling of high-temperature inelastic behaviour of the austenitic steel aisi type 316 using a continuum damage mechanics approach, J. of Strain Analysis for Eng. Design, 2012, 47(4), 229-243 Web of ScienceCrossrefGoogle Scholar
Dimmler G., Weinert P., Cerjak H., Extrapolation of short-term creep rupture data - The potential risk of over-estimation, Int. J. Pres. Ves. Pip., 2008, 85, 55-62 Web of ScienceCrossrefGoogle Scholar
About the article
Published Online: 2017-06-23
Citation Information: Open Engineering, Volume 7, Issue 1, Pages 126–140, ISSN (Online) 2391-5439, DOI: https://doi.org/10.1515/eng-2017-0019.
© 2017 Yevgen Gorash and Donald MacKenzie. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. BY-NC-ND 3.0