Low-Cycle Fatigue Life Prediction and Reliability Evaluation of Turbine Blades with Distributed Collaborative LSSVR

Abstract

Low-cycle fatigue (LCF) of turbine blades involves several disciplines with multi-uncertain factors and is a high-nonlinear complex problem. To improve prediction accuracy and computational efficiency, this paper develops DC-LSSVR approach, integrating least squares support vector regression (LSSVR) into the distributed collaborative (DC) strategy, for the LCF life prediction and reliability evaluation of turbine blades. Considering the influence of the uncertain factors, i.e., design sizes, applied loads and material properties, the reliability assessment framework is constructed. Through the integration of the DC-LSSVR reliability method with the theoretical models, including the Smith-Watson-Topper (SWT) mean stress correction and linear cumulative damage (LCD) rule, the LCF life is predicted and reliability evaluation is completed. Finally, the DC-LSSVR is proved to be a promising approach for the reliability assessment of complex structures.