# Introduction

he static essay, serving to determine the ultimate resistance to break, may be considered as a very particular case of a essay of fatigue, the number of cycles with break was equal at the most to 1. This ultimate resistance with break is the minimal value in which the break of the material occurs in a certain way.

The problem of dispersal of the trial results, as well in statics (for the determination of the ultimate resistance with break) that in fatigue (for the determination of the number of cycles with break or for the determination endurance limit) is the consequence of the structural heterogeneity of the defects since the elaboration of material [1]- [11]. The density of these defects vary from specimen to another, yet all taken from the same sample. In the case of fatigue, for a given stress level ? i , one can also observe a dispersion value of the number of cycles to failure N i [12]- [13]. The number of cycles to failure is related to the material structure and thus the defects that contain. That these defects are difficult to measure, it is advisable to use as a random variable [12], for a given stress level, the number of cycles to failure to explain the existence of defects by a probabilistic approach. We rely on the concept of the "weakest link" Weibull to formulate a new expression of the SN curve can take into account the actual volume of a flawless material considered. In the absence of a means of assessing defect in the structure, the results for two types of materials are comparable with those of Wöhler and Basquin. A simulation made by the variation of the tensile strength of a material that shows this new expression for describing the fatigue behavior of a hardened material.


# II.


# Concept of "Weakest Link" Weibull

According to this concept, for juxtaposed volumes, subjected to monotonic loading, the failure of one is independent of the other, if one of the volumes fails, the system is considered failed [14]. The survival probability of a system consisting of n elements (or volume) is given by: R n =1-e -? fdv (

Where f is the probability density of failure and dv is the elementary volume constituting the reference volume V.


# III.


# Specification of the Probability Density Function f

We will consider as a random variable the number of cycles to failure for a given stress level by the fact that the number of cycles determines the fatigue. Dependent on the dispersion, the number of cycles is a variable representing statistically the stress applied from the viewpoint fatigue (damage to a given level of stress). We will consider the failure probability density f as a function of number of cycles to failure; it is given in the following form:
f(N) = ?N m (2)
Where ? is a parameter to be determined, and the module m, within the meaning of Weibull [14].

The parameter ? can be determined when we know a characteristic value of the number of cycles N c to a level corresponding stress. At this number of cycles N c characteristic density of default probability take the value 1. It then pulls:
?=N c -m (3) or, f(N)= ? N N c ? m (4)

# T

In this case, the probability of survival given equation ( 1) becomes:
R n =1-e -?? N N c ? m .dv (5)
IV.


# Proposal for a New Expression of the S-N Curve

It is obvious that there is a relationship between the probability of failure P (function the number of cycles) and the level of responsible of this fatigue failure stress. We admit that there is a function g dependent on the characteristics of fatigue ? D and static R m linking the probability of failure P stress responsible for this failure.

Since ? D and R m are constants for a given material, then g is a constant function. A simple relation linking the stress at failure probability is that the stress is proportional to the probability of failure; with as the proportionality coefficient function g. This relationship is expressed by :
?=g?? D ,R m ?P(N) =g?? D ,R m ??1-R(N)? (6) ie, from (5): ?=g?? D ,R m ?e -?? N N c ? m
.dv (7) For a tension loading when the load is uniform, the expression (7) becomes :
?=g?? D ,R m ?e -V? N N c ? m (8)
Thus, two limiting cases are possible according to the value of the stress ?:

? For ? = R m , ultimate resistance to break, N = 0, the relation (8) becomes:
R m =g?? D ,R m ?(9)
? For ? = s D , endurance limit, N = N D , the relation ( 8) becomes :
? D =g?? D ,R m ?e -V? N D N c ? m (10) ie, g?? D ,R m ?=? D e V? N D N c ? m (11)
The equality of two expressions g, given by equations ( 9) and (11) gives R m depending on
? D R m =? D e V? N D N c ? m (12)
The expression of the S-N curve in tension given by equation ( 8) becomes :
? ? ? ? ? ?=R m e -V? N N c ? m ie ?=? D e V?? N D N c ? m -? N N c ? m ? (13) V.

# Experimental Study of the New Expression

To validate the sensitivity of the proposed fatigue behavior of a material model for V = 1, we have experimental data from two types of materials: steel martensite P220 and 100C6. A comparative study is made, from these data, with models of the most utilized S-N curves [15]- [16] such that Basquin, the Wöhler and the Stromeyer :
? ? ? ? ? log(N) =-a B log(?)+b B : Basquin log(N) =-a W ? +b W : Wöhler log(N) =-a S log(?-? D )+b S : Stromeyer log(N)=a N log ?log ? R m ? ?? +b N : proposed(14)
Where, for the proposed model
? a N = 1 m b N =log(N C )
The curves parameters are determined for each fatigue testing each of the two steels result, by the method of least square.


# a) Test Data i. P220 Steel

The data are taken from the steel work performed LaMI (Laboratory of Mechanics and Engineering) in order to investigate the influence of cutting processes on the fatigue behavior of parts. Two procedures were used, laser and shears, it, with several execution qualities (good, standard and bad) corresponding to different sets of machines settings. The tableaux.2 and 3 below give, respectively, the chemical composition and the fatigue test result of the 100C6 steel [11]. The value of the ultimate resistance to break R m of the steel P220 is between 600 and 800 MPa, it will take three values in this range for the simulation thus allowing to highlight the sensitivity of the proposed model.


# b) Comparison

The comparison will be made by tracing curves in the plane (log (?), log (N))). Parameter values, determined by the method of least squares, corresponding to different models, given by Equations ( 14), for the two steels, are given in the following Table     


# Conclusion

The sensitivity of the proposed model to the values of R m , is reflected in an angular shift in direction of rotation hands of a watch while R m increases (figure.2). The limit of D fatigue increases substantially, but the value of the number of cycles N c dimunie characteristic significantly well as the modulus value m while R m increases, the appreciable increase in the endurance limit with that of the breaking strength reflecting the inflection the S-N curve traced in the plane ( ; N), thus modifying the scope of limited field of fatigue.

The proposed model introduces a new parameter, the number of cycles characteristic N c for which the failure probability density function takes the value 1and therefore the smallest number of cycles which the rupture occurs with certainty. It takes also into account the volume of the material. All the calculations are done by taking V=1. To take into account the defects can be considered as a fraction V corresponds to useful volume of the material constituting the piece to explain the variations of cycles number to failure for a given level of stress. The model allows, given its sensitivity to the change R m , to track the fatigue behavior of work hardened materials, a phenomenon for which the value of its breaking strength R m varies notably.

VIII.
![.](image-2.png "")
44![Value of parameters of the models used for the two steels P220 steel Model Parameters Basquin a B =23.66 ; b B =148.2 Wöhler a W =0.078 ; b W =36.65 Stromeyer a S =1.52 ; b S =17.73 Proposed Model R m =600MPa; ? D =271.24MPa; m=0.062; N c =214037564; Year 2014 ume XIV Issue III Version I A New Expression of the Curve S-N in Fatigue based on the Concept of the "Weakest Link" of Weibull a N =16.13; b N =19.18 R m =700MPa; ? D =271.57MPa; m=0.051; N c =15622128; a N =19.61; b N =16.56 four models and the linear trend of the experiment; b) Models Basquin of Wölher those proposed and the linear trend of the experience.](image-3.png "4 :Table 4 :")
1![Figure 1 : SN curves in the plane (ln (?) ln (N)) of the models studied for P220 steel](image-4.png "Figure 1 :")
2![Figure 2 : Sensitivity of proposed model with the values of R m Here we can see clearly that the SN curve model Stromeyer is not linear and strongly deviates from the experimental curve in the oligocyclic domain (Figure.1.a). Basquin, Wöhler curves and that proposed for different values of R m approximate linear trend given by experience. We find a correlation between the last three curves (Figure.1.b), the relative error on the stress values provided for by these curves do not generally exceed 5% compared to the values of the experiment. ? 100C6 steel](image-5.png "Figure 2 :")
3![Figure 3 : SN curves in the plane (ln (?) ln (N)) of the models studied, for steel 100C6](image-6.png "Figure 3 :")
.
1the standard laser process resultMaterial: P220Type of sollicitation: tension-compression at constant amplitudeSolicitation frequency: 20HzDimensions and shape of the specimen?(MPa)344321,8319,9319,8311,5311,1310309,5290280275270235N10596,529682828124832046737116230550320423830919152008966239739099475335707(*)7522596(*)(*) no ruptureii. Martensite Steel 100C6
2Martensite steel 100C6SiMnPSCrCuNiMoO0,2420,3390,0120,0081,461-0,1470,032-
3Matereal: martensite steel 100C6 (E=210GPa, R m =2300MPa, ? D =850MPa et N D =6.0265x10 9 cycles).Solicitation frequency: 20KHzFatigue test results?(MPa)975960950940930900890880870860850Nmoy.x10 60,11830,10190,16399,806167466664,4542,1990,751853,36026,5
			© 2014 Global Journals Inc. (US)
		
		

			

## Acknowledgement

We thank the Committee CONCOFORT CNAR (Supports National Center for Research in Chad) who funded our trip to France for the finalization of this article. Faculty of Exact Sciences at the University of N'Djamena (Chad) and the Polytechnic University of Mongo to be accepted to release us this work. All of our gratitude to the responsible of the Laboratory MMS Pascal Montlucon Institute for their technical support (France).

			

			

				


* 
	
		Strength of materials and the Weibull distribution
		
			EricSLindquist
		
		
	




* 
	
		Calibration of the Weibull stress scale parameter
		
			JasonPPetit
		
		
			RobertHDoddsJr
		
	
	
		Engineering Fracture Machanics
		
			72
			
			2005
		
	
	? u using the Master curve




* 
	
		Unsing Markov chain Monte Carlo methods to solve full Bayesian modeling of PWR vessel flow distributions. Reability Engineering and System Safety
		
			GCeleux
		
		
			MPorsz
		
		
			JNWandji
		
		
			FPerrot
		
		
			1999
			66
			
		
	




* 
	
		Equation and a fast algorithm for determining the probability of failure intiated by flows
		
			MTTodinov
		
	
	
		International Journal of Solid and Structures
		
			2002
		
	




* 
	
		Limiting of probability of failure for composants containing flows
		
			MTTodinov
		
	
	
		Computational Materials Sciences
		
			
			2005
		
	




* 
	
		Failure modeling in finite elements analysis: Random material imperfections
		
			MUnosson
		
		
			LOlovsson
		
		
			KSimonsson
		
	
	
		Mechanics of Material
		
			37
			
			2005
		
	




* 
	
		Some notes on the correlation between fracture and defect statistic: Are Weibull statistic valid for very small specimens
		
			RobertDanzer
		
	
	
		Journal of the European Ceramic Scociety
		
			27
			
			2006
		
	




* 
	
		Multiaxialité, hétérogénéité intrinsèque et structural des essais d'autoéchauffement et de fatigue à grnad nombre de cycles
		
			MartinPoncelet
		
		
			2007-2008
			6
		
		
			ENS Cachan/CNRS/Université de Paris
		
	
	Thèse




* 
	
		Prediction of self-hearting measurements under propositional and non-proportinal multiaxial cyclic loadings
		
			MartinPoncelet
		
		
			CédricDoudard
		
		
			SylvainCalloch
		
		
			FrancoisHild
		
		
			BastienWeber
		
		
			AndréGaltier
		
	
	
		C. R. Mecanique
		
			335
			
			2007
		
	




* 
	
		Determination of an HCF criterion by thermal measurements under biaxial cycle loading
		
			MCédric Doudard
		
		
			SPoncelet
		
		
			CCalloch
		
		
			FBoue
		
		
			AHild
		
		
			Galtier
		
	
	
		International Journal of Fatigue
		
			29
			
			2007
		
	




* 
	
		Prediction of very high cycle fatigue failure for high strength steels, based on the inclusion geometrical properties
		
			GonzaloMDominguez Almaraz
		
	
	
		Mechanics of Material
		
			40
			
			2008
		
	




* 
	
		Intégration d'une démarche fiabiliste dans l'exploitation des critères de fatigue multiaxiaux
		
			BSoh Fotsing
		
		
			EMaronne
		
		
			NNadjitonon
		
		
			J.-LRobert
		
		
			2004
		
	
	23 ème Journée de Printemps de la Commission de Fatigue. Du 25 au 26 mai




* 
	
		Fatigue multiaxiale des structures industrielles sous chargement quelconque. Thèse INSA Lyon, N) 99ISAL0056
		
			WeberBastien
		
		
			1999
			248
		
	




* 
	
		A statical theory of the strength of materials. Ingeniors Vetenskaps Akademiens
		
			W
		
	
	
		Hamdlingar NR
		
			151
			1939
		
	




* 
	
		Nouvelle loi non linéaire d'endommagement par fatigue basée sur la courbe de Bastenaire
		
			BTikri
		
		
			NNadjitonon
		
		
			JLRobert
		
		
		
			IUT de Montluçon-Département Génie Mécanique et Productique-