SEQUENTIAL SAMPLING METHOD BASED KRIGING MODEL FOR STRUCTURE CONDITION ASSESSMENT

摘要

The present invention relates to a sequential sampling method based on a Kriging model for structure condition assessment for improving the computational efficiency of an analysis model repeatedly performed for structure reliability assessment, finite element model-based optimum design, probabilistic analysis, and inverse analysis problems (finite element model correction). The sequential sampling method of the present invention comprises: an initial sample construction step of configuring verification samples for independently determining the accuracy of an initial sample, which is to be used for constructing an approximation model, and the constructed approximation model; an FE analysis step of performing FE analysis based on the configured samples; a Kriging model construction step of constructing a Kriging model by using an output vale and an input value obtained from the FE analysis; an application step of calculating an estimated value for the verification samples by using the constructed Kriging model and assessing the accuracy of the approximation model; a stop determination step of whether the constructed Kriging model corresponds to a stop condition by determining whether the Kriging model has sufficient accuracy; a prediction improvement checking step of searching for an additional sample for improving the prediction accuracy of the approximation model based on a prediction improvement indicator when the Kriging model does not correspond to the stop condition in the stop determination step; and a sampling addition step of adding sampling based on the sequential sampling from the prediction improvement indicator. The method of the present invention returns to the FE analysis step after the sampling addition step.