Issue 14

B. Lo it. Lee [19] t at least 15 s pproximate statistical p res. ually with a on and Moo ize should b e grossly es e samples fro f the less fre

bato da Silva et recommends pecimens.

alii, Frattura ed a value 5%

Integrità Struttu less than fa

rale, 14 (2010) tigue limit i

36-44; DOI: 10 nitially estim

.3221/IGF-ESIS.

14.04

dev rec Sta Th lim the Th equ dis nor Bro De 9 a

iation of th ommends ru tistical analys e DM metho it. It requires survivals or e stress levels ations propo tributed; ( ii ) mal distribu wnlee [21] a noting by n i t nd 10, respec

e fatigue lim nning the tes is d provides a that the two only the failu S spaced eq sed by Dix the sample s tion should b ssures that th he number o tively.

ated. Collin

[20]

formulas to roperties be chosen increm d [15] respe e big, aroun timated prev m 5 to 10 sp quent event

calculate the determined b ent d are nu ct three assu d 40 to 50 s iously in ord ecimens are at the stress l

mean, DM  , a y using the d mbered i wh mptions: ( i ) pecimens or er to specify reliable to de evel i, two qu

nd standard ata of the le ere i=0 for t the fatigue more and ( the step of termine the m antities can

deviation,  ss frequent e he lowest str strength sho iii ) the stand stress increm ean fatigue be calculated

DM ,of the fat vent, either o ess level S 0 . uld be norm ard deviation ents. Howe limit. : A and B, E (9) (10) t is survival

igue nly The ally of ver, qns.

i A   i B  

in

i

2 i n ws the estim sed minus si i

and

Th oth

e Eq. 11 sho erwise, it is u

ate of the gn. The Eqns

the plus sig how the estim

n is used if ate of stand

the more fr ard deviation

equent even .

mean, where . 12 and 13 s

   

  

A

1 2

(11)

S

d

DM  





0

n

i

 B 

B n 

  

  

2

 n A n  2 i

2

A 

if

(12)

d

0.

3



i

1

.62

0.029

DM  











2

n

i

i

or

 B n A n     2 i i

2

(13)

if

.53 d

0

DM  

0.3 

to quantify th sing these m s near the m on and de M n and propo es. y Svensson-L his correctio

Th to fati acc ord be Th dev ten

e Staircase m predict estim gue [22]. Thi urate standar er to evaluat an improvem e Eq. 14 sho iation by Di d increase th

ethods are n ate accurate s method con d deviation. e and to imp ent in all ma ws the estim xon-Mood a e deviation e  ection was d n. The form dent on the n 3 DM N N      

otably accura of fatigue l centrates th Braam and v rove the relia ximum-likelih ate of stand nd N is the t stimate by Di

te and efficie imit standard e most exper an der Zwaag bility of stan ood evaluati ard deviation otal specime xon-Mood.

nt in terms deviation u imental point [23], Svenss dard deviatio on procedur corrected b n number. T

e mean fatig ethods with ean therefore aré [24], Lin sed a linear c óren [26],  n is a strictly

ue strength small samp is more diffi [17] and Rab orrection fac  function of SL , where D

cult ycle an d in d to ard and

but very diffi les at high c cult to obtain b [25] worke tor and foun is the stand sample size M

(14)







SL

A m Ló con

odified corr ren correctio stants depen

eveloped wh of the propo umber of sa

ich attempted sed standard mples, see T

to allow a g deviation es ab. 5.

reater range timate, PC  , i

of unbiased e s shown in E

stimation th q. 15, where

on- are

an the Svenss A , B , and m

m

N

  

    

 

DM B d 

(15)

A

PC  



3  largest devia DM N 

 tion will be u

SL  or 

In

this work the

sed,

.

PC

39