Prediction of creep characteristics of superalloy bolts based on itransformer and θ projection method

Creep life (CL) of bolts in one batch will have noticeably differences in the same working conditions sometimes due to the machining error and other facts. The traditional creep life prediction method requires a full creep test in a specimen with target specification to collect data in modelling and predicting. Creep tests are timeconsuming, so it is important to develop a faster and cheaper method for bolt creep life prediction. In this study, we proposed a method which uses a deep learning model consisting by conventional neural networks (CNN) and itransformer to analyze the fluctuation of force transducer reading (FTR) while loading to predict constant in theta projection method (PM) equation and creep fracture time to plot predicted creep curve. The proposed method can utilize existing data in modelling and takes FTR of the target specimen to predict. The measurement of FTR in this study can be done in a few minutes, creep curve can offer more information when determining bolt life. The result shows that the predicted GH2132 (ASTM Incoloy A-286) threaded specimen curve fit the true creep curve well in 480 MPa uniaxial load and 650 degrees C. This method indicates that the creep curve information can be obtained from fast measured FTR in constant stress creep (CSC) when using deep learning models such as itransformer to analyze FTR properly. The findings of this study provide the basis for developing a practical creep life (CL) prediction method in further steps.