Using Historical Fatigue Testing Data to Determine Design Allowable Properties for Additively Manufactured Materials

The lack of accurate mechanical property and performance measurements for additively manufactured (AM) materials is a major issue in the eventual implementation of AM fabricated components in critical applications. Since the processing routes for AM materials differ significantly from traditional wrought and cast products, existing property data bases and design allowables are not applicable. Current approaches for developing material property data bases require the testing and characterization of large quantities of pedigreed data, which can become cost and time prohibitive. Fatigue properties present a significant challenge when considering both the inherent variability or scatter due to differences in processing and testing and the sensitivity of AM materials to the formation of defects. Historical fatigue testing data provides a potential pathway for compiling large fatigue data sets in a cost-effective manner and to identify gaps where new pedigreed data could be added. However, the historical data were not collected in a consistent manner and were impacted by differences in the testing procedures, material properties, surface preparation, and other important fatigue contributors. A procedure for acquisition, standardization, and categorization of the available data sets was developed for different materials systems. Using this standardized data, statistical analysis tools and emerging machine learning techniques were used to develop categorical and predictive models, extract trends in the fatigue data, and interrogate the data to identify the underlying mechanisms governing the observed fatigue properties.