EMR-SSM: Synchronous surrogate modeling-based enhanced moving regression method for multi-response prediction and reliability evaluation

To achieve multi-response prediction and reliability evaluation of complex structural system, a high efficient and precision strategy, namely synchronous surrogate modeling-based enhanced moving regression (EMR-SSM, short for) method, is proposed based on synchronous surrogate modeling (SSM) approach and enhanced moving regression (EMR) framework. In this method, the SSM approach is developed by integrating matrix analytical theory and surrogate modeling method, which is used to establish these related cell arrays and to synchronously build multi-response performance functions of complex structural system; the EMR framework is evolved by intelligent optimization algorithm and moving least squares technique, which is employed to search effective modeling samples with optimal radius and to determine unknown coefficients. Besides, we explore four SSM approaches with the responses surface method (RSM), Kriging model, support vector machine (SVM) and artificial neural network (ANN) by the EMR framework, which include synchronous RSM-based EMR (EMR-SRSM), synchronous Kriging model-based EMR (EMR-SKM), synchronous SVM-based EMR (EMR-SSVM) and synchronous ANN-based EMR (EMR-SANN) methods, and then develop hybrid SSM-based EMR (EMR-HSSM) method. Furthermore, three examples, including multi-objective prediction and probabilistic analysis of complex nonlinear function, left and right flap deflection angle prediction and reliability evaluation, and multi-failure responses prediction and reliability evaluation of high-pressure turbine blisk, are applied to illustrate the effectiveness of the proposed methods in modeling features and simulation characteristics.