PNNL

Abstract

This work introduces a methodology for assessment of data quality for creep properties of alloys. Data quality assessment is needed to ensure the reliability of data used in analytics to develop new materials and to predict the performance of established materials in new applications. As data quality metrics have not been standardized for material properties data, quality rating guidelines are developed here for the aspects of data completeness, precision, accessibility, and authority of source. The specific design requirements for creep-resistant alloy development were considered in creating each metric. Establishing the quality of a dataset in these areas will enable robust analysis. High quality data can be set aside to develop predictive models. Low quality data need not be discarded but can be used for verification and comparison of models. Determining the quality of a materials dataset will also provide additional metadata with the data resource and will promote data reusability. A sample high quality dataset is presented to highlight the areas where data gaps exist.