工程p-S-N曲线小样本拟合方法比较

机械结构件疲劳寿命预估的基础是其材料的p-S-N曲线,而p-S-N曲线的拟合需要大量的试验数据,对人力物力的成本需求较高。为研究小样本数据拟合p-S-N曲线的方法,分别运用Bootstrap法和样本信息聚集原理,得到了泵头体材料的p-S-N曲线。结果表明,Bootstrap法拟合的p-S-N曲线误差较均匀;样本信息聚集原理在基准应力为716 MPa比756 MPa时预测10~4～10~6次范围内的疲劳寿命误差更小,且应力级别相差较大时的疲劳寿命的预估比Bootstrap法更为精确;结合经济效益及试验误差,最佳应力级别数为5组,每组应力级别下进行5根试样。

Comparison of Small Sample Fitting Methods for Engineering p-S-N Curve

The basis of fatigue life prediction of mechanical components is the p-S-N curve of materials, and the fitting of p-S-N curve needs a lot of test data, which requires a high cost of human and material resources. In order to study the method of fitting p-S-N curve with small sample data, the p-S-N curve of pump head material was obtained by Bootstrap method and the principle of sample information aggregation. The results show that the error of p-S-N curve fitted by Bootstrap method is more uniform. The fatigue life prediction error of sample information aggregation principle is smaller in the range of 104-106 times when the base stress is 716 MPa than 756 MPa, and the fatigue life prediction is more accurate when the stress level is different than Bootstrap method. According to the economic benefit and test error, the optimum stress level is five groups, and five specimens are tested under each stress level.