Project implementation and work plan

Figure: Work Packages of NextSim and main relationships.

Overall Work Structure

WP1: Technology Verification, Validation, and Error Quantification

WP1 considers four test cases of significant market impact covering different key business problematics for either complete aircraft configurations or aircraft components and that associated require HPC resources:

  • Long Range Aircraft in Transonic Flight Conditions

Here the proper modelling of compressible and viscous effects are the key drivers to achieve the reliable simulation of shocks and boundary layers and their interaction with possibly separations. 

  • Long Range Aircraft in Low Speed Flight Conditions with deployed High Lift

Here vortex dominated flows around complex geometries, their interaction with boundary layers and separated flows are challenging to model accurately.

  • Engine Jet

Here, the modelling of turbulence is critical for the accurate simulation of jet in the first place to properly predict interaction of jet plume with other aircraft components downstream the nozzle (e.g. wing, flap, horizontal tail plane) then for acoustics.

  • Generic Transport Aircraft

The accurate modelling of propellers and flows around complex geometries are at stake here.

Table 1: Proposed test cases

The test cases proposed including the challenges described, constitute the technical backbone of NextSim and will define the needs and tasks include in WP2, WP3 and WP4, focusing on increasing the technology readiness levels of new numerical methods and computing platforms, paving the way for aeronautical CFD simulations on future Exascale systems. As a final remark, WP1 also includes the development of the mini-apps concept, which is finally deployed in WP5. 

WP2: Algorithms improvement for efficient solution process

WP2 focuses on the development of novel fundamental algorithms that limit the efficiency and accuracy of CODA and aeronautical solvers, taking into consideration the new Exascale systems. 

Planned activities will be targeting algorithms with the strongest impact on convergence and accuracy of the numerical solutions. This WP will develop activities to achieve the first quantifiable objective of NextSim: to improve the convergence rate to a maximum of 0.95 in complex grids. 

The algorithms studied here will tackle temporal (implicit and explicit), and spatial integration schemes applied to (U)RANS models, 1/2-equation model (kω, kε, SA) and RSM (Reynolds Stress Model), hybrid RANS/LES (Large Eddy Simulation) including DES (Detached Eddy Simulation) and, if appropriate, full LES may also be considered for verifying the algorithms in terms of the “pros and cons” when a hierarchy of modelling approaches are used with increasing demanding to computational resources

WP3: Algorithms for data management, visualization and modelling

WP3 will investigate new approaches to manage the huge amount of data generated by CODA’s simulations: surrogate-based quality estimation and new techniques for data compressing and surrogate management. This two methodologies will be jointly considered. In this sense, WP3 is fed by the simulation data obtained in WP1 by the application of CODA to the test cases, and its outcomes are critical in order to get valuable information of the solution. 

Aditionally,  surrogate models will be directly constructed from the information contained in the modes obtained from POD/DMD analysis. These surrogate models will used to perform local or global searching algorithms in optimization loops. A large variety of learning models can be used as surrogates, e.g., Polynomial Regression (PR), Multivariate Adaptive Regression Splines (MARS), Gaussian Processes, Kriging (KG), Co-kriging, Artificial Neural Networks (ANN), Radial Basis Functions (RBF) and Support Vector Machines (SVM), among others.

WP4: Parallelization and Communication

WP4 will provide an efficient implementation on massively parallel processors of the algorithms developed in those WP2 and WP3.  

WP4 focuses on low-level and system-level parallelization. The different tasks include: Solvers and optimization methods with reduced global communication; algorithm reformulation to improve memory re-use; analysis of novel algorithms for data streaming; chip-level energy-aware algorithms; dynamic load balancing through task-based parallelism and chip-level fault tolerance.  This WP will also address optimal integration of the solvers, optimized for individual computing devices, for efficient execution on modern hybrid compute nodes.

WP5: Dissemination, training and exploitation.

WP5 will ensure an optimal dissemination of the project´s results and an optimal plan for exploitation of results. Dissemination will be made through the preparation of scientific publications, participation in international events and distribution of data in specialized repositories. The communication activities will address general public, gaining visibility for the EC funding and make the society more aware of the problems tackled at scientific level. Finally, WP5 will define and pursue an exploitation plan, based on the role that each partner plays in the supply chain to ensure that the results enter into the market as soon as possible.

In more detail, Task 5.1 will incorporate the software prototypes developed in WP1-WP4 into new open source “mini-apps” and kernels. The objective is to provide the broader Exascale community with a self-contained, fully documented numerical application encapsulating key technologies for each of our case studies. The philosophy behind it is that they can be used as easily customizable proxies for real applications when benchmarking emerging HPC core technologies and architectures; evaluating new tools and libraries and testing improvements to programming models and system software. 

We will release the source code of the mini-apps and kernels with a Berkeley BSD license, permitting the exploitation of the open source code in commercial software, thus accelerating industry uptake and economic impact. 

Contact Us


This project has received funding from the European High-Performance Computing Joint Undertaking Joint Undertaking (JU) under grant agreement No 956104. The JU receives support from the European Union’s Horizon 2020 research and innovation programme and Spain, France, Germany.