Bayesian Calibration of Computer Models with Informative Failures

TL;DR

This paper introduces a Bayesian calibration method that accounts for and excludes input regions causing computational failures, improving the robustness of model calibration in complex simulations.

Contribution

It proposes a Bayesian meta-model that integrates failure detection with calibration, enabling exclusion of unstable input regions in a unified probabilistic framework.

Findings

Successfully applied to carbon capture data with unstable CFD numerics

Effectively excludes failure-prone input regions from calibration

Enhances robustness of computer model calibration processes

Abstract

There are many practical difficulties in the calibration of computer models to experimental data. One such complication is the fact that certain combinations of the calibration inputs can cause the code to output data lacking fundamental properties, or even to produce no output at all. In many cases the researchers want or need to exclude the possibility of these "failures" within their analyses. We propose a Bayesian (meta-)model in which the posterior distribution for the calibration parameters naturally excludes regions of the input space corresponding to failed runs. That is, we define a statistical selection model to rigorously couple the disjoint problems of binary classification and computer model calibration. We demonstrate our methodology using data from a carbon capture experiment in which the numerics of the computational fluid dynamics are prone to instability.