Validation in fusion research: Towards guidelines and best practices

Abstract

Because experiment/model comparisons in magnetic confinement fusion have not yet satisfied the requirements for validation as understood broadly, a set of approaches to validating mathematical models and numerical algorithms are recommended as good practices. Previously identified procedures, such as verification, qualification, and analysis of error and uncertainty, remain important. However, particular challenges intrinsic to fusion plasmas and physical measurement therein lead to identification of new or less familiar concepts that are also critical in validation. These include the primacy hierarchy, which tracks the integration of measurable quantities, and sensitivity analysis, which assesses how model output is apportioned to different sources of variation. The use of validation metrics for individual measurements is extended to multiple measurements, with provisions for the primacy hierarchy and sensitivity. This composite validation metric is essential for quantitatively evaluating comparisons with experiments.To mount successful and credible validation in magnetic fusion, a new culture of validation is envisaged.

Plasma turbulence, suprathermal ion dynamics and code validation on the basic plasma physics device TORPEX

“…Validating numerical codes is of upmost importance to assess the maturity of the understanding of plasma dynamics and the predictive capabilities of simulations (Terry et al 2008;Greenwaldet al 2010). A validation project can be thought as a four-step procedure: (i) the model needs to be qualified by establishing the applicability of the model hypotheses for the simulated physical phenomenon; (ii) verification of the code is necessary to prove that the code solves correctly the model equations; (iii) simulations and experiments have to be compared and analyzed using the same techniques, considering a number of observables, i.e.…”

Langmuir probe-based observables for plasma-turbulence code validation and application to the TORPEX basic plasma physics experiment

“…A validation project is a four step procedure [1]. (i) The simulation model needs to be qualified, i.e.…”

Simulation of Diffusive Lithium Evaporation Onto the NSTX Vessel Walls

“…The error bars in Fig. 4 are actually 2/π of this value, as is suggested by the validation metric in [13].…”

“…The lower ends are 7/24 of these rates, representing the complete hydration of deposited lithium into LiOH. Note that the 90% confidence intervals described in [13] are much smaller than than the uncertainty introduced by the undetermined chemical composition of the QMB deposits and are not shown in Fig.4.…”