Some high-throughput experiments aim to test many samples simultaneously under nominally the same conditions. However, whether a particular high-throughput experiment does so must be certified. We previously described a high-throughput experiment to study the statistics of rupture stretch of materials. In such an experiment, a large set of samples were tested simultaneously. We noticed a systematic bias that samples in different subsets give different statistical distributions of rupture stretch. Here we describe an approach to detect and reduce systematic bias in the high-throughput experiment. We divide the whole set of the data of the experiment into subsets, obtain the statistical distribution of rupture stretches for each subset, and compare the “closeness” of the distributions among the pairs of subsets by using the Anderson-Darling (A-D) test. We then try to reduce the systematic bias by improving the experimental design, such as increasing the rigidity of the frame and the firmness of the grips. With the improved experimental design, the newly obtained rupture data passes the A-D test. We use the new rupture data to fit the Weibull distribution. The improved high-throughput experiment greatly increases the accuracy of fitting and prediction of rare-event rupture stretch.

一些高通量实验旨在在名义上相同的条件下同时测试许多样本。但是，特定的高通量实验是否这样做必须经过认证。我们之前描述了一个高通量实验来研究材料断裂拉伸的统计数据。在这样的实验中，同时测试了大量样本。我们注意到系统性偏差，即不同子集中的样本给出不同的破裂拉伸统计分布。在这里，我们描述了一种在高通量实验中检测和减少系统偏差的方法。我们将整个实验数据集划分为子集，得到每个子集的破裂拉伸统计分布，并使用 Anderson-Darling (AD) 检验比较子集对之间分布的“接近度”。然后，我们尝试通过改进实验设计来减少系统偏差，例如增加框架的刚度和握把的坚固性。通过改进实验设计，新获得的破裂数据通过了AD测试。我们使用新的破裂数据来拟合 Weibull 分布。改进的高通量实验大大提高了罕见事件破裂拉伸的拟合和预测精度。我们使用新的破裂数据来拟合 Weibull 分布。改进的高通量实验大大提高了罕见事件破裂拉伸的拟合和预测精度。我们使用新的破裂数据来拟合 Weibull 分布。改进的高通量实验大大提高了罕见事件破裂拉伸的拟合和预测精度。