Sei sulla pagina 1di 1

Prediction of Fatigue Life A3

Another equally important facet of the life prediction process is treatment of the
remaining or residual strength of the damaged structure (as distinguished from
residual stress). Such considerations are intimately related to Fracture Mechanics
and to fail-safe design; further discussion will be found in those Sections.
Particularly have the considerations of crack propagation by Fracture Mechanics
now acquired the power and validity conferred by successful numerical results,
[275], [312], [382] et al are excellent examples of its application.

The form of the input load function has been the chief topic above, and it has
been tacitly assumed that the life prediction methods utilized the conventional,
monotonic properties of the materials. However, prediction methods are evolving
which use the "fatigue" or cyclic properties for much added realism (see Sec. 4.1.2).
In the following brief summary of the types of life prediction methods, the depen­
dence of these types on both the load function and on the different properties is
quite evident. All the methods are followed by the application of a damage rule,
usually linear.

1. Use of an appropriate fatigue curve, S-N or Goodman, for the applied stresses
that are assumed known or calculable. Limited by the available data to sine
wave input (constant amplitude).

2. Random process input, usually constant amplitude in test, but variable in


service. Response analyzed by Fourier method and converted to equivalent
sine (g) level, or by the methods of Section 3.2.1.

3. The load spectrum is divided into a small number of blocks within which the
sine-wave amplitude is constant, but is different for each block. Blocks are
then programmed for the required, variable number of cycles, repetitively or
randomly. Limitations are similar to (1). Analysis by method of Section 3.2.1.

4. The Local Strain Method requires the cyclic or "fatigue" properties; accomo-
dates the sequence of loading, pre-strain and mean stress effects, but is
limited to relatively simple states of stress (as are most of the currently
used fatigue damage procedures). It also easily accepts the results of the
more sophisticated events-counting schemes, as Rainflow.

5. The Nominal Stress Method is a relatively approximate scheme, adaptable to


low cycle fatigue. It assumes that the smooth σ-N curve and the notched S-N
curves are straight lines on log-log paper; at 10 cycles S = σ/Kt, and at 1
cycle S = σ Kt. It accomodates plastic flow but neglects mean stress effects,
as is appropriate in LCF.

Ten excellent papers on the last two methods are given in [435], together with a
discussion by Prof. Fuchs of Stanford University, which ends with several cogent
suggestions for an engineering approach to Cumulative Damage:

a. Look at the load histories and decide whether or not one should expect sequence
effects. This decision requires judgment, but no more than the choice of the
most highly stressed areas in a proposed structure, or the assumption of the
loads on a member that failed in the field.

b. If sequence effects can be neglected, use the simple nominal stress range
method to estimate life to crack initiation. Condensed load histories are
recommended.

c. If sequence effects must be expected, use a local strain method to estimate


life to crack initiation. Condensed load histories are recommended.

Potrebbero piacerti anche