Toward Quantifying the Predictive Maturity of Validated Numerical Simulations.

摘要

The topic of verification and validation (V&V) has received much attention during the past decade, and yet a central question within the scientific computing community remains unanswered: 'When is a numerical prediction good enough.' All too often, the answer to this question relies heavily on expert opinion: a qualitative and subjective approach on which to rely, given the intensely quantitative nature of V&V. Aiming to address this concern, a recent research effort at Los Alamos National Laboratory introduced a quantitative metric referred to as the Predictive Maturity Index (PMI). This metric takes into account three common concerns related to numerical models: their accuracy, their complexity, and the coverage of the design space to which they are applied, as compared to the available experimental data.