Calibration and improved prediction of computer models by universal\n Kriging

Abstract

This paper addresses the use of experimental data for calibrating a computer\nmodel and improving its predictions of the underlying physical system. A global\nstatistical approach is proposed in which the bias between the computer model\nand the physical system is modeled as a realization of a Gaussian process. The\napplication of classical statistical inference to this statistical model yields\na rigorous method for calibrating the computer model and for adding to its\npredictions a statistical correction based on experimental data. This\nstatistical correction can substantially improve the calibrated computer model\nfor predicting the physical system on new experimental conditions. Furthermore,\na quantification of the uncertainty of this prediction is provided. Physical\nexpertise on the calibration parameters can also be taken into account in a\nBayesian framework. Finally, the method is applied to the thermal-hydraulic\ncode FLICA 4, in a single phase friction model framework. It allows to improve\nthe predictions of the thermal-hydraulic code FLICA 4 significantly.\n