Variations in fatigue damage during the measurement and reconstruction of service load histories

In this investigation, a number of problems related to the general problem of obtaining a measure of the accuracy with which reconstructed histories represent service conditions have been addressed. The difference in fatigue damage between histories reconstructed from a given Markov matrix is determined by examining the variation of average damage per cycle. This variation is shown to decrease at a rate proportional to the square root of the number of cycles in the histories. A model was constructed based on the assumption that the distribution of fatigue damage for any cycle number is independent of the cycle number and identically distributed for all cycle numbers. This model was shown to predict the inverse square root proportionality of the simulation results, but was conservative by a factor of 1.7 for the Markov matrix. The predicted variation in the average damage per cycle, when the measured service history was stored directly into the path-reversal transition matrix, was slightly less than the predicted variation for the Markov matrix. The path-reversal transition matrix is an extension of the Markov matrix that includes material response and, when used as a cycle counting method, correctly stores and reproduces the strain range and mean stress for every closed loop.;The difference between the average damage per cycle for a measured service history and the expected average damage per cycle of reconstructed histories is a result of errors introduced by discretization and by cycle counting methods. The possible error in calculated fatigue damage as a result of discretization is determined by modelling the distribution of peaks and valleys within a strain interval. The number of discrete strain levels that can be analysed using the path-reversal transition matrix is limited due to the size of the matrix. It was found for this matrix that twelve discrete strain levels provides a good trade-off between computational effort and accuracy in fatigue life prediction. (Abstract shortened by UMI.);The variation in average damage per cycle of repeated measurements of the strain in a suspension component of a vehicle being driven over cobble stones, which was assumed to give strains that comprised a stationary random process, was determined. The variation in average damage for a group of these recordings showed the same power law relationship with number of cycles as the reconstructed histories did.