Assessing the Accuracy of Creep Life Prediction Models

摘要

This paper assesses the relative accuracy of the Monkman-Grant relation and a low strain modification. These models were estimated from short-term data on 2.25Cr-1Mo steel. This data was obtained by merging together different batches of steel based on tests for data homogeneity. These models were then be used to predict the known longer-term creep lives published in creep data sheet No. 3B of the National Research Institute for Metals. Predictions were assessed using a one step prediction error and the mean square error which was further decomposed into systematic and random components. Two different batches of short term test results were successfully merged together and the Monkman-Grant and its low strain modification were both capable of accurately predicted lives at close to actual operating conditions. The prediction errors made were predominantly random in nature. Neither model was capable of predicting the lives associated with very high temperatures.