Session: VVUQ for Fluid Dynamics and Heat Transfer

Paper Number: 153326

153326 - A Solution Verification Exercise for a Practical Application of a Coastal Hydrodynamics Model 

Abstract: 

Within renewable energy resources, tidal energy stands out due to its great reliability and predictability. Meanwhile, its extraction is highly site specific. An accurate resource assessment is thus crucial for the identification and development of a potential site. For this task, numerical models are a comparatively cheap option to obtain flow data over large areas of interest. In order to assess the credibility of such numerical models the estimation of numerical uncertainties is fundamental.

This article presents an engineering application of a tool to determine the time and space discretization uncertainty of steady and unsteady flow simulations. The most recent version of the tool enables error estimations based on one- or two-term power series expansion combined with free and fixed exponents. Four alternatives of weighting functions exist to tune the contribution of a numerical solution for a given spatial and temporal refinement level to the error estimation. Two of those weighting functions are based on robust regression. In addition to the discretization error, a safety factor calculated via a smooth transition function and the scatter in the numerical data are considered to finally obtain the numerical discretization uncertainty.

In this work, the above techniques find application to determine the discretization uncertainty of flow simulations conducted with MOHID Water (http://www.mohid.com/), a three-dimensional baroclinic free-surface model targeted at simulating coastal and estuarine hydrodynamics. To the best knowledge of the authors, it is the first time that a solution verification exercise is realized for this hydrodynamic model. As a test case, a complex real-world example is selected: the Tagus estuary, one of the largest river estuaries in Western Europe. The combination of river currents, tides and the narrowed river mouth accelerate the flow such that it makes the site a candidate for hydrokinetic energy extraction. The uncertainty analysis is carried out for time series of water level and current velocities extracted at potential locations for marine current turbines. For selected time points, the uncertainty is determined in the entire flow field and visualized by uncertainty heat maps.

It is observed that across the different levels of discretization the numerical results for the current velocities contain a considerable amount of noise. The robust regression weighting functions are able to partially or fully deactivate the contribution of potential outlier solutions and achieve a high quality of fit for the error estimation, meanwhile the scatter in the data still can lead to elevated uncertainties. Overall, more acceptable uncertainties are obtained only for finer spatial refinements than those currently used in common practice. The importance of High-Performance Computing resources is thus highlighted to make accurate numerical modelling feasible for engineering applications. For the future, this research exercise is the first step towards a full verification and validation study of the Tagus numerical model which will include also experimental on-site flow measurements.

Presenting Author: Manuel Rentschler IST / BlueOasis

Presenting Author Biography: PhD Student at IST

Authors:

Manuel Rentschler IST / BlueOasis
Luis Eca IST
Guilherme Vaz BlueOasis