This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).
The current research in Bayesian model prediction and validation of computer models mainly focuses on computer experiments with single output and fixed input variables. The proposed research focuses on Bayesian approach for calibration, validation, prediction, and experimental design of computer models with functional output. Bayesian predictive models for calibrating computer models based on functional computer outputs and physical observations will be constructed. Methods and metrics for calibrating and validating computer models will be developed. Experimental design and optimization strategies for data collection will be established. Theoretical properties of the developed methodologies will be investigated and assessed.
If successful, the research results will bridge the gap between statistical researchers and engineering practitioners, and stimulate additional research that improve effective and efficient utilization of expensive computer models developed by scientists and engineers. There is an increasing demand for accurate predictive models and metrics for calibrating and validating computer models from model analysts and engineering designers. The proposed research will allow scientists and engineers to effectively assess and evaluate expensive computer models for various scientific and engineering applications, including IC packaging and fabrication, chemical and nuclear energy equipment development, cellular material design and manufacturing, nano material design and manufacturing, etc. The major impact of the proposed research is to improve the effectiveness of computer model developers (model analysts) and users (scientists and engineering designers) in scientific understanding as well as in design and manufacturing in various important scientific and engineering applications.
