On this page you will find details of the projects awarded under the SHAPE 4th Call.
Project Title: Adoption of high performance computing in Neural Designer
SME: Artelnics, Spain
The predictive analytics market is undergoing an impressive growth. Indeed, organizations that incorporate that technique into their daily operations not only better manage the present, but also increase the probability of future success.
Artelnics develops the professional predictive analytics solution called Neural Designer. It makes intelligent use of data by discovering complex relationships, recognizing unknown patterns, predicting actual trends or finding associations. Neural Designer out-stands in terms of usability, functionality and performance.
Current technology lacks from advanced model selection techniques, and usually requires many computational resources. The main challenge for Neural Designer is to include a framework capable of untangling complex interactions in big data sets. In order to do that, the software must achieve high performance by means of parallel processing.
The users of the solution are professional data scientists, which work at analytics departments of innovative companies, consulting firms specialized in analytics or research centres. Neural Designer will be capable of analysing bigger data sets in less time, providing our customers with results in a way previously unachievable.
Project Title: Use of CFD on new HPC accelerators for an accurate prediction of erosion and corrosion induced by flowing liquid metals
SME: Milano Multiphysics s.r.l.s, Italy
The project fits in the frame of a collaboration between Ansaldo Nucleare S.p.A and Milano Multiphyiscs s.r.l.s that aims at modelling via an optimized version of OpenFOAM the combined effects of corrosion and erosion of structural materials in flowing liquid metals, with a focus on nuclear applications. The performance of the structural materials envisioned to be used in GEN IV Heavy Liquid Metal-cooled fast nuclear reactors is limited a) by the degradation of physical and mechanical properties by long term exposures to neutron fluxes and b) by the chemical interactions with the flowing fluid. The computational power required for a high fidelity simulation is so demanding that it is really difficult for an SME to get access to the necessary resources. The new generation of Intel Phi processors, KNL, appears a very promising way to bridge this gap. Not having the necessary expertise available, PRACE contribution will be extremely valuable to support the company in the porting, optimization and benchmarking process.
Project Title: Optimising 2D flow for faster, better steam turbine design
SME: Renuda UK Limited, United Kingdom
CodeX models and simulates the performance of industrial steam turbines in 2D using first-principles and sophisticated physics. As a design and optimisation assist tool used in the power generation industry, CodeX PRACE AISBL – 98, Rue du Trône B -1050 Bruxelles Belgium www.Prace-ri.eu 16 must be robust, support quick decision making, deliver accurate and verifiable results fast, whilst remaining intuitive to use by non-specialists. Although faster than 3D software, CodeX’s time-to-solution takes several days, running serially on desktops machines. For CodeX to be widely adopted by industry, a runtime of several hours is required. Validation tests against 3D simulations and experimental data show that for better accuracy, simulations must run on larger computational meshes than currently possible. This project aims to develop a parallelised version of CodeX for HPC platforms, reducing its runtime and improving its computational mesh capabilities and solution algorithms. HPC adoption will mean a step change in CodeX’s capabilities, enabling further automation and improvement of the turbine design process.
Project Title: Development of Chameleon Monte Carlo code for HPC: Toward Realistic Modelling of Complex Composite Systems
SME: Scienomics, France
In order to perform de novo materials development and/or optimization scientists need to work with reliable models at the atomistic level. Increasing complexity of modern materials make these models become excessively complex and large and therefore exhibit characteristic that cannot be studied by standard techniques. In order to overcome this limitation Scienomics developed a software, called Chameleon, which combines Monte Carlo approach with modern algorithms for relaxing such materials, such as chain altering moves. Chameleon is now in its second phase of development and is currently able to simulate many different systems, further code optimization and validations is however still needed. The goal of this project is to improve Chameleon capabilities in several areas and test them on crucial industrial problems.
The access to a supercomputing center will allow Scienomics to improve and/or validate three different aspects of the Chameleon code:
– Improve Chameleon performance for single processor simulations
– Improve Chameleon parallelization:
a. increase the part of the code parallelized
b. improve the parallelization scaling for large number of processor
c. test hybrid implementation
– Scienomics will leverage the large supercomputer capabilities to run both long simulations and a statistical number of simulations in parallel to validate the ergodic theorem
These developments and validations will then be applied to optimize the polymer formulation for the mechanical properties of graphene/polymer composites. Such systems are crucial in many industrial sectors.
These simulations will involve the generation of very large systems. Scienomics will then exploit powerful machine memory capabilities to test the limitations of the MAPS platform capabilities to generate and visualize these very large systems (over 1 million atoms). Based on these tests, improvement of MAPS capabilities, such as porting MAPS to GPU/CPU architecture or improving MAPS code to speed visualization of large system for example, will then be considered.