Sei sulla pagina 1di 3

IntJ Fatigue9 No 2 (1987) pp 119-121

A n e w definition of the rainflow cycle counting method


I. Rychlik

A new equivalent definition of the rainflow cycle counting method is presented, which expresses the rainflow cycle amplitudes in explicit analytical formulae. The method attaches to each maximum of the strain function the amplitude of a corresponding cycle or two half cycles, which are evaluated independently from each other. This definition is more convenient for the statistical analysis of rainflow cycle amplitudes for a random loading process. Key words: fatigue; rainflow counting; toplevel-up counting; cycle amplitude; random loading

The rainflow cycle (RFC) counting method, first proposed by Matsuishi and Endo, 1 is now generally regarded as the method leading to the best estimators of 'fatigue life', see Dowling. 2 Due to the great importance of the RFC counting method, many different algorithms have been proposed in the literature) -s However, all these procedures have a complicated 'sequential' structure, which makes them difficult to apply when their statistical properties are to be studied. It is well known that the rainflow counting method corresponds to the stable cyclic stress/strain behaviour of a metal in that all strain ranges counted as cycles will form closed stress/strain hysteresis loops, and those counted as half cycles will not. The range of a closed hysteresis loop is defined by its highest and lowest point, ie by the pair of a local maximum and minimum in the strain function (see also the section entitled Remarks below). The method presented here uses this correspondence explicitly by attaching to each maximum of the strain function (or stress function) the amplitude of a corresponding cycle (closed hysteresis loop). This way of defining the rainflow cycle counting method makes it possible to evaluate the long-run distribution of the RFC amplitude for a random loading process; see References 6 and 7.

_ S <t2

./"

,,
I
I
tt-1

I
t

I I
t+

tl

Fig. 1 Illustrations of the toplevel-up cycle, with amplitude H(t


= min(H-(t), H+(t)), defined by Definition 1

Rules
(a) If H+(t) >>. H-(t)and t- > - T , or H+(t) < H-(t) and t + < T, then a toplevel-up cycle is defined with an amplitude off

(b)

(c)

D e f i n i t i o n 1 : t o p l e v e l - u p cycle ( T U C ) counting method


Let y(s) be the stress function observed for - T ~< s ~< T. For each localy-maximum one cycle or two half cycles will be counted, according to the following procedure. Let the maximum occur at time t, and let t + be the time for the first up-crossing after t of the level y(t) (or t + = T if no such up-crossing exists for t < s < T), and let t- be the last down-crossing before t of the same level (or t - = - T if no such down-crossing exists for - T < s < t). Two ranges originating at (t,y(t)) are now defined, see Fig. 1;

H(t) = min (H-(t), H+(t)) (2) If the maximum at t is the first or the last y-extremum in the interval [ - T , T], o n e half cycle is defined with an amplitude of either H+(t) or H-(t), respectively. In all other cases two half cycles with amplitudes of H +(t) and H-(t) are defined.

For example, in Fig. 2 toplevel-up cycles are attached to the maxima at tl, t2 and t3 due to rule (a). Furthermore, one half cycle is attached to the maximum at to, and two half cycles to the maximum at t 4. We shall now prove that the toplevel-up cycle counting method is equivalent to the rainflow cycle counting method, and begin by defining the RFC counting method following Collins)

D e f i n i t i o n 2: r a i n f l o w cycle ( R F C ) counting method


I: The rainflow cycle counting method is illustrated in Fig. 3. The strain/time history is plotted so that the time axis

n - ( t ) = y(t) - min{y(s); t- < s < t} H+(t) = y(t) - min{y(s); t < s < t +}

(1)

0142-1123/87/02119-03 $3.00 1987 Butterworth & Co (Publishers) Ltd Int J Fatigue April 1987

119

H - (t4)

H+(t4)

3
1

~-

Time -T
I I

+T to tl
t

t2

t3

t4

Time

2:11

Fig. 4 Illustrations of the half cycle counting procedure by definition

Fig. 2 Example of the toplevel-up cycle counting method

H-(t 0)
Strain

H + ,to}

H-

(t I )

--T

t3

t2

t~

Y
to

III I I I I t31 =t2_1 ~ =t1_1 t I =t_l

to

T I_ r Time

Fig. 5 Illustration of the strain history used to prove the equivalence of the TUC and RFC counting methods

I II ii
4 IL I I AI

= H+(to)

H-(t4)

-1 II
I.-

i ! = H(ta) H (t2) H(t3)


H+(t4)

Fig. 3 Example of the rainflow cycle counting method applied to the strain history of Fig. 2

of the history. This in Fig. 4 half cycles are also counted between peaks 6 and 5, 5 and 4, and so on. After the most negative minimum in the history, half cycles are counted which terminate at the most positive maximum occurring subsequently in the history, the most negative minimum occurring after this maximum, and so on to the end of the history. Again referring to Fig. 4, half cycles are counted between peaks 8 and 9, 9 and 10, and so on. The strain ranges counted as half cycles therefore increase in magnitude to the maximum and then decrease.

is vertically downward, and the lines connecting the strain peaks are imagined to be a series of pagoda roofs. Several rules are imposed on rain dripping down these roofs so that half cycles are defined. Rainflow begins successively at the inside of each strain peak. The rainflow initiating at each peak is allowed to drip down and continue except that, if it initiates at a minimum, it must stop when it comes opposite a minimum more negative than the minimum from which it initiated. Similarly, if the rainflow initiates at a maximum, it must stop when it comes opposite a maximum more positive than the maximum from which it initiated. A rainflow must also stop if it meets rain from a roof above. II: When this procedure is applied to a strain history, a half cycle is counted between the most positive maximum and the most negative minimum. In Fig. 4, for example, a half cycle is counted between the peaks 7 and 8. Assume that o f these two the most positive maximum occurs first. Half cycles are also counted between the most positive maximum and the most negative minimum that occurs before it in the history, between this minimum and the most positive maximum occurring prior to it, and so on to the beginning

Proof of the equivalence of T U C and RFC counting methods


Assume that the most positive maximum of the j-stress function in the interval [ - T, 7] occurs at t. Let {t i}, _ T ~ < . . . < t 2 < t 1 < t o be the times of maxima which satisfy the condition (b) or (c), and let t~ and t_il be the time points defined on Fig. 1. Then, for each i > 0, H + ( t i) /> H - ( t i) and (ti) - = - T, ie t i is the most positive maximum in the interval [ - T , t~] and til is the most negative minimum in the interval [ - T, t i], see Fig. 5. By Definition 2:II, the half cycles t ~ tl = tll ~ t 1 ~ t_11 = t~ --* t 2 --* t_21 = t~ -* . . . . with ranges H - ( t ) , H+(I1), H - ( I 1 ) , H + (t 2), H - (t 2). . . . , respectively, are counted. Consequently, to the maximum at t i one can attach two half cycles, H+(t i) and H - ( t i ) . The set of half cycles which occur after t o can be treated in a similar way, and hence the same half cycles are counted by both procedures. Since for any maximum at time t, say, which satisfies condition (a) a full cycle must be counted, we need only prove that the amplitude of the cycle is H(t) = min(H-(t),

H+(t)).
Int J F a t i g u e A p r i l 1 9 8 7

120

y (s), strain history

)_ s, Time Fig. 6 Illustration of a strain function in which a minimum is the lowest point of two cycles

We begin with a maximum such that H+(t) >1 H-(t), and t - > - T . Since t- > - T , there exists a maximum at time t*, say - T ~< t* < t - , such that the rain path started at t* drops down from the minimum at time t-1. Since y(t*) > ),(t) and j(t_l) I> .Y(6), the rain path started at the maximum at t is stopped by the rain path started at t*, and a half cycle with range H-(t) is counted. Observe that all rain paths started at maxima in the interval (t-, t) are stopped before or at time t, and the rain path started at t* cannot be stopped before the time t +. Next, the rain path started at a minimum at t-1 is stopped in the interval (6 tO, giving a half cycle with range H-(t), and the RFC procedure counts a cycle with amplitude H(t) = min(H-(t), H+ (t)). We turn now to a maximum such that H-(t) > H+(t) and t < T. The rain path started at t is stopped at t +, giving a half cycle with range H + (t). Furthermore, the rain path started at a minimum at time tl will be stopped by the rainflow dropping down from the maximum at t, giving a half cycle with the range H+(t), and a full cycle with amplitude H(t) = min(H-(t), H+(t)) is counted.

a cycle has been determined, the maximum and minimum associated with it can be eliminated from further analysis. Consequently, a cycle defines uniquely a pair of maximum and minimum values which are its highest and lowest points, k what we have extensively used in the definition of the TUC method. Since the TUC procedure, which is equivalent to RFC method, can be started at any maximum, let the RFC algorithm start at a local maximum at time, say t. By Definition 2:1, the rainflow initiated at t has a range H+(t). Next, we change the time direction and start a (new) rainfiow from the same maximum. The range of this path is H-(t). Now, if the maximum is a highest point of a cycle (closed hysteresis loop), the amplitude of this cycle is min(H-(t), H+(t)) by the TUC method. So the TUC procedure can be seen as a stationary (independent of the starting point) and time-invariant RFC procedure. Finally, note that the TUC procedure, which chooses max(H-(t), H+(t)) as the amplitude of a cycle, cannot be equivalent to the RFC procedure since it is possible to have a strain function with a minimum which is the lowest point of two cycles see Fig. 6.

Acknowledgements
The author is indebted to Professor Georg Lindgren for discussion and valuable comments.

References
1. Matsuishi, M. end Endo, T. 'Fatigue of metals subjected to varying stress' Paper presented to Japan Soc Mech Engrs (Jukvoka, Japan, 1968) Dowling, N. E. 'Fatigue predictions for complicated stressstrain histories" J Mater1 (1972) pp 71-87 Downing, S. D. and Socie, D. F. 'Simple rainflow counting algorithms" IntJ Fatigue4 No 1 (January 1982) pp 31--40 Okamura, H., Sakai, S. and Susuki, I. 'Cumulative fatigue damage u nder random loads" Fatigue Engng Mater and Struct 1 (1979) pp 409-419 Socia, D. F. "Fatigue-life prediction using local stress/strain concepts' Exptl Mech 17 No 2 (1977) pp 50-56 Lindgren, G. and Rychlik, I. 'Rain flow cycle distributions for fatigue life prediction under Gaussian load processes' Star Res Rep (University of Lund, Sweden, 1986) pp 1-17 to appear in Fatigue Engng Mater and Struct Rychlik, I. 'Rain flow cycle distribution for a stationary Gaussian load process" Stat Res Rep 1986:4 (University of Lund, Sweden, 1986) pp 1-36 Collins, J. A. Failure of Materials in Mechanical Design (Wiley Interscience, New York, NY, USA, 1981 ) pp 282-285

2. 3. 4.

Remarks
We finish with some remarks about the relationship between the RFC and TUC methods. The RFC procedure has a complicated 'sequential' structure, and the amplitude of the cycles and half cycles can depend on the order in which one initiates the rainflows. Usually, rainflows begin successively from the first extremum of the strain function. However, they can also be initiated from the last extremum, changing the time direction, and the same cycles and half cycles will be obtained. Thus the RFC method is time invariant. The sequential structure of the RFC method is caused by the fact that it must keep track of those extremes which have not yet formed a cycle (closed hysteresis loop). Once

5. 6.

7.

8.

Author
Dr Rychlik is with the Department of Mathematical Statistics, University of Lund, Box 118, S-221 00 Lund, Sweden.

Int J F a t i g u e April 1 9 8 7

121

Potrebbero piacerti anche