针对基于$k$系数的裕量和不确定性量化方法只适用于正态分布的实验数据的局限性, 文章分析了系统性能特征分位数与裕量和不确定性量化评估所关注的问题之间的联系, 通过分位数回归求解指定的性能特征分位数及其置信限, 实现了对实验数据的裕量和不确定性量化评估. 新的评估方法对数据的分布形式及方差一致性没有要求, 具有较强的数据适应性. 示例分析对此进行了说明.

The paper reviews the limitations of Quantification of Margins and Uncertainties (QMU) based on $k$-factor and analyses the relation between the percentile of the system performance characteristic and the attention of the QMU. A new QMU method with the performance characteristic percentile and its confidence bound estimated by quantile regression is presented. The proposed QMU method does not need any assumption about the distribution or homoskedasticity of the test data, and it can be applied to wide variety of datasets. Examples are furnished to demonstrate the methodology.