Towards an Intelligent Non-stationaryPerformance Prediction of Engineering Systems

摘要

The analysis of complex engineering systems can often be expensive thereby necessitating the use of surrogate models within any design optimization. However, the time variant response of quantities of interest can be non-stationary in nature and therefore difficult to rep-resent effectively with traditional surrogate modelling techniques. The following paper presents the application of partial non-stationary krig-ing to the prediction of time variant responses where the definition of the non-linear mapping scheme is based upon prior knowledge of either the inputs to, or the nature of, the engineering system considered.