It is common practice to perform accelerated creep testing (ACT) using time-temperature parameter (TTP) models. The TTP models are calibrated to creep-rupture data at high temperature and/or stress and extrapolate to lower temperature and/or stress where data is not available. The long-term creep rupture behavior (at low temperature and stress) is often not available due to the quantity, duration, and cost of testing. A limited scope of creep-rupture data is often analyzed using the TTP models. When conducting long-term extrapolation, statistical uncertainty becomes an issue. The ability of the TTP models to accurately predict creep-rupture at long life is often limited and the inherent material properties can dramatically influence creep-rupture life. Unfortunately, there is no consensus on the statistic for assessing the quality of TTP extrapolation. This study demonstrates methodology to assessing the uncertainty in creep rupture predictions for 316SS using the Larson Miller parameter. Over 2,000 creep-rupture data points are collected and digitized from the NIMS, ASM, MAPTIS, and ORNL databases; metadata such as the material’s form, thermomechanical processing, and chemical composition are recorded. Statistical uncertainty is measured using the “Z parameter”, which describes the deviation of creep-rupture data to a TTP model. The ability of the TTP models to extrapolate to long life is analyzed via exclusion of data. This is accomplished by: excluding 50% of the data, and by excluding the longest 10% of the data. It is shown that culling data in any way produces more conservative creep rupture predictions. The spread of the dataset will also affect the width of the reliability bands.

This content is only available via PDF.
You do not currently have access to this content.