Illustrating the Future Prediction of Performance Based on Computer Code, Physical Experiments, and Critical Performance Parameter Samples

ABSTRACT

In this article, we present a generic example to illustrate various points about making future predictions of population performance using a biased performance computer code, physical performance data, and critical performance parameter data sampled from the population at various times. We show how the actual performance data help to correct the biased computer code and the impact of uncertainty, especially when the prediction is made far from where the available data are taken. We also demonstrate how a Bayesian approach allows both inferences about the unknown parameters and predictions to be made in a consistent framework.

KEYWORDS:

• Bayesian
• bias correction
• Markov chain Monte Carlo
• Monte Carlo

ACKNOWLEDGEMENT

We thank C. C. Essex for her support and encouragement of this work.