Double Weighted Least Square Method for P-S-N Curve Regression
建立P-S-N曲线的双加权最小二乘法

In order to take full advantage of fatigue data from both the group test and the staircase test and thus to get P-S-N curve with higher precision, a P-S-N curve regression approach was proposed based on double weighted least square method. In this approach, taking into account the effect of fatigue experiment method, experimental fatigue data scatteration and sample size on P-S-N curve, an allocation model for different error sources was founded. Firstly, the minimum sample size criterion was used to process the group experimental data, resulting in the length of the confidence interval of any percentile point for the data distribution, which was utilized to carry out the first weighted fitting. Secondly, the staircase test data were taken into account to implement the second weighted fitting, and finally the P-S-N curve at any reliability level was obtained. Seventeen groups of experimental data of metals and PMMA were employed to verify the ability, rationality and accuracy of prediction for present method. It is shown that the established P-S-N curves have higher accuracy and credibility.