基于核密度应力谱外推的转向架构架疲劳寿命评估

提出了一种基于多样本核密度应力谱外推的疲劳寿命评估方法，研究了核密度估计中的最优带宽与核函数的确定问题，采用灰色关联度分析方法对应力谱的外推优度进行量化评价与检验，并讨论了疲劳寿命评估的相对误差与外推倍数之间的关系；为了验证方法的正确性与可行性，以某转向架构架转臂定位安装座焊缝附近某测点为研究对象，选取了该测点在车轮分别处于镟轮初期、中期和末期时的3组动应力测试数据，进行多样本核密度应力谱外推与疲劳评估。研究结果表明:基于最小渐近均方积分误差的概率密度函数拟合优度良好，所研究的4种核函数类型中，基于Epanechekov核函数外推的相关性最好，相关系数为0.99，较其他3种核函数提高了0.01%~0.12%，基于Circular核函数外推的一致性最好，灰色关联度为0.592 0，较其他3种核函数提高了0.17%~0.32%；基于多样本核密度应力谱外推10倍后的疲劳评估寿命相比基于线性外推的评估寿命减少了1.15%；当应力谱外推至全寿命周期时，基于核密度外推所评估的安全运营里程减少了6.45%。可见，基于核密度应力谱外推的疲劳寿命评估更偏于安全，能保证车辆结构的安全服役。 

Fatigue life evaluation of bogie frame based on kernel density extrapolation for stress spectrum

A fatigue life evaluation method based on the multi-sample kernel density stress spectrum extrapolation was proposed. The determination of optimal bandwidth and kernel function in the kernel density estimation was studied. The grey correlation analysis method was proposed to quantitatively evaluate and verify the extrapolation optimization of stress spectrum. The relationship between the relative error of fatigue life evaluation and the extrapolation multiple was discussed. To verify the correctness and feasibility of the method, taking a measurement point near the weld of the positioning mounting seat of a bogie frame at the research object, three sets of dynamic stress test data were selected to conduct the multi-sample knernel density stress spectrum extrapolation and fatigue evaluation when the wheel was in the initial, middle, and final stages, respectively. Research results show that the probability density function based on the minimum asymptotic integral mean square error has a goodness of fit. Among the four types of studied kernel functions, the correlation based on the Epanechekov kernel function is the best, and the correlation coefficient is 0.99 and 0.01%-0.12% higher than the coefficients of the other three kernel functions. The consistency based on the Circular kernel function is the best, and the grey correlation degree is 0.592 0 and 0.17%-0.32% higher than the degrees of the other three kernel functions. The assessment fatigue life based on 10-time multi-sample kernel density stress spectrum extrapolation reduces by 1.15% compared with that based on the linear extrapolation. When the stress spectrum is extrapolated to the whole life cycle, the safe operation mileage evaluated based on the kernel density extrapolation reduces by 6.45%. Therefore, the fatigue life evaluation based on the extrapolation of kernel density stress spectrum is safer, and can ensure the safe service of the vehicle structure. 4 tabs, 12 figs, 31 refs.