Abstract: Monte Carlo modification is carried out in one of imperfect cases, the negative prediction of standard deviation of the logarithm of fatigue lives in high stress amplitude range, for historic probabilistic fatigue S-N curves. An adjustable random numerical generator which utilizes mixed multiplicative congruential method is applied in the random simulation process. Sample size for the simulation is controlled in a range of 5 to 15 for each group of specimens on the same stress level to address real possible fatigue tests. Reconstructions are made successively on the curves in mid-long life regime and the fatigue limits. Modification is then performed by an improved maximum likelihood method to estimate the curves. And the curves are extended into super long life regime by a so-called concurrent probability approach. Reasonability and feasibility of the present method are verified by a modification of Q235 steel historic curves. Results show that, for the case of negative prediction of standard deviation, the historic curves have cross points in high stress amplitude range and tend to give a non-conservative prediction. After modification, the above deficiencies are overcome and the curves are reasonably extended into super long life regime.

Key words: Fatigue, Modification, Monte Carlo simulation, Probabilistic analysis, S-N curve, Angle Measurement, Autocollimation, Automatic Measurement, Posture Matrix, Satellite Assembly