Abstract
Solution verification is a part of the larger framework for assessing the implementational and practical accuracy of computer models, often referred to as verification, validation and uncertainty quantification (VVUQ). Through the years different methods have been developed to estimate the error from using a discretised version of a mathematical model, implemented on a finite precision computer. Given the variety of mathematical approaches to solution verification, these methods can bring great benefits to the computational wind engineering (CWE) community. Some methods like grid convergence and residual tracking have gained popularity for estimating discretisation and iterative uncertainty and, as a result, are being applied as universal tools for solution verification and are often falsely believed to ensure comprehensive verification. This is compounded by the fact that solution verification is a computationally expensive process and is therefore often neglected or performed without consideration for the fine detail of individual simulations.
In this paper, we use a case study from the Architectural Institute of Japan’s (AIJ) catalogue, as an example for CWE practitioners, which we use to demonstrate one of the most recognised metrics for spatial discretisation verification, the grid convergence index, and to study a variety of measures for iterative solution verification. We examine key conditions that must be satisfied for these methods to be applied and discuss advantages and potential issues with each of them. We conclude our study by exploring the place of solution verification in the grand scheme of VVUQ and elucidate some priority areas for future effort.