Abstract: Multiscale simulations are an essential computational tool in a range of research disciplines, and provide unprecedented levels of scientific insight at a tractable cost in terms of effort and compute re- sources. To provide this, we need such simulations to produce results that are both robust and actionable. The VECMA toolkit (VECMAtk), which is officially released in conjunction with the present paper, estab- lishes a platform to achieve this by exposing patterns for verification, validation and uncertainty quantification (VVUQ). These patterns can be combined to capture complex scenarios, applied to applications in dis- parate domains, and used to run multiscale simulations on any desktop, cluster or supercomputing platform.

Abstract: Modelling and simulation are increasingly well established techniques in a wide range of academic and industrial domains. As their use becomes increasingly important it is vital that we understand both their sensitivity to inputs and how much confidence we should have in their results. Nonetheless, few simulations are reported with rigorous validation (V) and verification (V), or even meaningful error bars (uncertainty quantification - UQ). EasyVVUQ is a Python library designed to allow the integration of non-intrusive VVUQ techniques into existing simulation workflows. Our aim is to provide the basis for tools which wrap around applications to allow the user to specify the scientifically interesting parameters of the model and the type of VVUQ algorithm they wish to implement and the details of the setup and analysis are abstracted from them. To this end we have designed JSON based input formats that provide a human readable and comprehensible interface to the code. The EasyVVUQ framework is based on the concept of a Campaign of simulations, the inputs of which are generated by a range of sampling algorithms. This Campaign is executed externally to the library but the results processed, aggregated and analyzed within it. EasyVVUQ provides simple templating features that facilitate mapping between scientific parameters and input options and files for a wide range of applications out of the box. Furthermore, our design allows simple user customization of both the input generation and the extraction of relevant data from simulation outputs by expert users and developers. We present it's use in three example multiscale applications from the VECMA project; protein-ligand binding affinity calculations, coupled molecular dynamics and finite element materials modelling and fusion.

Abstract: In order to predict the overall performance of a thermonuclear fusion device, an understanding on how microscale turbulence affects the global transport of the plasma is essential. A multiscale component based fusion simulation was designed by coupling together several single-scale physics models into a workflow comprising a transport code, an equilibrium code and a turbulence code. While previous simulations using such workflow showed promising results on propagating turbulent effects to the overall plasma transport [1], the profiles of densities and temperatures simulated by the transport model carry uncertainties that have yet to be quantified. The turbulence code provides the transport coefficients that are inherently noisy. These coefficients are then propagated through the transport code and produce an uncertainty interval in the calculated profiles, which would be used in the equilibrium and turbulence codes to calculate new uncertainty intervals. Our goal is therefore to study how these uncertainties propagate through the workflow, so that we can draw quantitative comparisons between numerical and experimental results. In this context, we are developing tools based on a non-intrusive polynomial chaos expansion [2] (PCE). For that, each sub-model is treated as a black box in which the PCE method is applied. Then, several statistical metrics are derived directly from the polynomial expansion, and finally we get the uncertainty quantification (UQ) and the parameter sensitivity of the multiscale model involved. References: [1] O.O. Luk, O. Hoenen, A. Bottino, B.D. Scott, D.P. Coster, ComPat framework for multiscale simulations applied to fusion plasmas, Computer Physics Communications (2019), https://doi.org/10.1016/j.cpc.2018.12.021. [2] R. Preuss, U. von Toussaint, Uncertainty quantification in ion–solid interaction simulations, Nuclear Instruments and Methods in Physics Research Section B (2017), https://doi.org/10.1016/j.nimb.2016.10.033.

Abstract: Uncertainty and sensitivity analysis provides insights on how uncertainty in the model inputs affects the model response [1, 2]. Usually, methods for such analysis are computationally expensive and may require high performance resources. In [3], we perform uncertainty quantification applying the quasi-Monte Carlo method for a two-dimensional version of an in-stent restenosis model (ISR2D) [4]. Additionally, in [5], we improve the efficiency of uncertainty estimation by applying the semi-intrusive multiscale method [6]. We observe approximately 30% uncertainty in the mean neointimal area as simulated by the ISR2D model. Depending on whether a fast initial endothelium recovery occurs, the proportion of the model variance due to natural variability ranges from 15% to 35%. The endothelium regeneration time is identified as the most influential model parameter. The model output contains a moderate quantity of uncertainty, and the model precision can be increased by obtaining a more certain value on the endothelium regeneration time. The results obtained by the semi-intrusive method show a good match to those obtained by a black-box quasi-Monte Carlo method (see Fig. 1). Moreover, we significantly reduce the computational cost of the uncertainty estimation. We conclude that the semi-intrusive metamodeling method is reliable and efficient, and can be applied to such complex models as the ISR2D model.

Abstract: We describe our efforts to create a multi-scale and multi-physics framework that can be retargeted across different disciplines. Currently we have implemented our approach in the astrophysical domain, for which we developed AMUSE, and generalized this to the oceanographic and climate sciences, which led to the development of OMUSE. The objective of this paper is to document the design choices that led to the successful implementation of these frameworks as well as the future challenges in applying this approach to other domains.