蜂窝夹层板撞击实验数据中的野值判别

基于超高速撞击物理实验数据对蜂窝夹层板撞击极限方程进行修正是获得高可信度新方程的一种常用方法,为了提高物理实验数据的可靠性,以国外131个碳纤维复合材料(CFRP)面板的蜂窝夹层板实验数据为对象,进行野值判别方法研究,发现该批数据中存在1个野值。将该野值剔除并基于剩余的130个数据重新进行方程修正后,新方程的总体预测率和安全预测率分别达到82.3%和93.1%,其绝对误差平方和、相对误差平方和分别为0.010、0.506,相对于剔除野值前的修正方程有所改善,表明实验数据野值判别方法可行、有效。为考核方法的适用性,对铝合金面板的蜂窝夹层板的实验数据也进行野值判别分析,结果显示该方法可合理识别野值。

Outlier identification from impact experimental data of honeycomb sandwich panel

To correct the ballistic limit equations of honeycomb sandwich panel based on hypervelocity impact physical experiment data is a common method to obtain high credibility equations. In order to improve the reliability of physical experimental data, a method to distinguish the outliers is studied. Among 131 experimental data of honeycomb sandwich panel with carbon fiber reinforced plastic (CFRP) faceplates, 1 outlier is found using the method. Eliminating the outlier, new equations are obtained by re-correcting the equations based on the rest 130 data. The new equations'' totality predicted rate and safety predicted rate reach 82.3% and 93.1%, respectively. And the sum of squared absolute errors and sum of squared relative errors are 0.010 and 0.506, respectively. Compared with the correction equations without eliminating the outlier, the new equations are improved. The results show that the experimental outlier data identification method is feasible and effective. To test the feasibility of the method, the experimental data of honeycomb sandwich panel with aluminum alloy faceplates are also analyzed and the results show that this method can be reasonably used to identify outliers.