SHAPE Success Stories

The following examples each give an overview of a sample SHAPE project. For more technical information on completed SHAPE projects please see the SHAPE White Papers section.

Engineering, construction

Meteo-ocean study

Naval engineering

Numerical simulation

Software

Wind Energy

Engineering, construction

Numerical simulation of accidental fires with a spillage of oil in large buildings

Project Title

Numerical simulation of accidental fires with a spillage of oil in large buildings.

Project Details

  • Company name: BACECG – BAC Engineering Consultancy Group
  • Country: Spain
  • PRACE partner: BSC, Barcelona, Spain

The Challenge

The objective of the project was to develop a fire engineering analysis (Performance-Based design) of a steel structure building that belongs to the ITER (International Thermonuclear Experimental Reactor) industrial complex in Cadarache (France), which is devoted to research in the field of Nuclear Fusion.

Numerical simulations of the accidental fires that can take place inside of the building have been carried out, some of them as complex as considering liquid combustible spillages or foam insulation combustion.

The Solution

The SME performed benchmarking and scalability tests of the Fire Dynamics Simulator code in BSC’s MareNostrum machine up to 2,048 cores and then the results were compared with its in-house cluster. The company verified that its infrastructure is not optimal with respect to node connectivity and system configuration that minimize the scalability of the code. The BSC team supported them to improve parts of the local installation helping to optimize their current infrastructure and detect the bottlenecks to tackle for future clusters and installations.

Business Impact

SHAPE has helped the Company to improve simulation times. It has allowed BAC to be much more competitive in the market, offering maximum quality service in a shorter time frame.

Benefits

  • The optimization of the cluster and a better knowledge of HPC technology (applied to BAC’s workstations infrastructure) have also made it possible to carry out two projects simultaneously, when previously one of the two projects would have had to be discarded due to the strict deadlines of the clients.
  • All this has been done while guaranteeing the quality of fluid dynamics and combustion models.

Images

Meteo-ocean study offer

Open ocean, Faster on-line statistics calculation

Project Title

Open ocean, Faster on-line statistics calculation.

Project Details

  • Company name: Open Ocean
  • Country: France
  • PRACE partner: Idris, GENCI/CNRS

The Challenge

The main goal of the project is to improve the post-processing step of the Open Ocean processing chain.

The Solution

The solution was to develop a parallel Python program.

Business Impact

Plan to change their actual proprietary scheduler for an open source one, cheaper and easier to maintain.

Help the SME working with other partners on different R&D projects.

Benefits

  • Identify the main bottleneck of its post-processing program.
  • Reconsider hardware choice.
  • Better understanding of the concept of HPC.

Images

Automatic Optimal Hull Design by Means of VPP

Project Title

Automatic Optimal Hull Design by Means of VPP.

Project Details

  • Company name: Hydros Innovation
  • Country: Italy
  • PRACE partner:

The Challenge

The main scope of this application was to evaluate the feasibility of automatic optimal hull design on HPC infrastructure and the impact of such a workflow on the day-by-day work of Hydros personnel with the final scope of reducing the cost and time required to obtain large CFD computational campaigns.

The Solution

The high-level solution is as follow: we coupled a CFD solver (OpenFOAM) and a parametric CAD design engine (CAESES®) on an existing HPC infrastructure. The python script drives the evolving of the computations in which every novel CAD design is built by CAESES and computationally analyzed by the CFD solver.

Business Impact

The main technical limitation of the direct optimization of complex geometries and flow as a hull shape with multi-fluids physics is the CPU time required and the associated cost to evaluate a big amount of designs. HPC solutions are of interest since we are able to launch a great amount of design evaluations simultaneously, thanks to a robust and automatized workflow. This would not have been possible with usual license schemes of commercial software.
HPC makes complex shape optimization affordable for industrial purposes thanks to an automatized, reliable and robust workflow using open-source software.

Benefits

Images

Numerical simulation services

DEMOCRatic Air quality SIMulation

Project Title

DEMOCRatic Air quality SIMulation.

Project Details

  • Company name: AmpliSIM
  • Country: France
  • PRACE partner: Idris, GENCI/CNRS

The Challenge

The DemocraSIM project aim was to bring Air Quality simulation to the general public. When the Fukushima hazard occurred, it was possible to display crowd-sourced data of radioactivity measurements on a single web map. If people can do their own measurement of radioactivity, and share it, then why can they not simulate an event and get an idea of their exposure?

The Solution

Integration of the AmpliSIM server to IDRIS computational resources, in compliance with HPC security policies. Interconnection with software and submission on the IDRIS computer.

Business Impact

The AmpliSIM toolset, including numerical models and post processing tools, was improved to meet the standards of a supercomputing parallelised architecture. This led to large time reductions in simulation, and the post processing capability of the web portal to support concurrent user access was strengthened to offer excellent scalability.

Benefits

    • Offered new services to customers
  • SME now use HPC in-house
  • rapidly visualise large amounts of data
  • developed their proprietary post-processing suite
  • can now visualise very large computational domains
  • enhanced security of the web services

Images

Software

RAPHI (Rarefield Flow Simulations on Xeon Phi Architectures). Optimising KOPPA code in order to exploit the potential of the Intel Xeon Phi architecture

Project Title

RAPHI (Rarefield Flow Simulations on Xeon Phi Architectures). Optimising KOPPA code in order to exploit the potential of the Intel Xeon Phi architecture.

Project Details

  • Company name: Optimad
  • Country: Italy
  • PRACE partner: CINECA, Bologna, Italy

The Challenge

The drastic increasing of computational resources during the recent decades opens new possibilities in computational simulations which start to be widely used in industries and SMEs. In this context, new codes have been implemented, targeting applications as for example in aerospace industry. In particular, equations for rarefield gas dynamics are really challenging because of their complexity. Without particular care in the implementation, the execution time of such a code becomes prohibitive. For this reason , an efficient code is needed in order to dramatically reduce the overall computation time.

New architectures for High Performance Computing (HPC), like Multi-Integrated-Cores architectures, allow the use of a very high number of computational units (cores) to execute parallel codes, obtained by distributing the computational load among the cores. However, the implementation of such codes requires expertise in HPC, and even then it remains challenging to efficiently exploit these architectures for numerical simulations. Indeed, such expertise is rarely present within an SME.

Therefore PRACE partners are in a good position to support SMEs to improve numerical codes applications. They give the opportunity to rethink the way codes are implemented in order to obtain decreased computational time on classical architectures., but also to exploit newly developed heterogeneous architectures such a MIC. Hence, the goal is to obtain a significant gain in performance to allow simulations that were previously unaffordable.

The Solution

KOPPA code is used for rarefield gas simulations and is computationally expensive compared to other CFD or CAE applications. It solves a model of the Boltzmann equation named ES-BGK.

The model requires a discretization of both of the physical and microscopic velocity space. The discretization of the latter is unusual In fluid dynamics and makes this problem computationally costly. Both discretizations are based on Cartesian grid but space discretization can also be done on a hierarchical grid with dynamic refinement. This procedure is implemented using the PABLO library. The code and the library are parallelized using message passing paradigm. The equation is time dependent and the solution is approximated using a discrete sequence of time steps. The main goal is to improve the computational time requirement of a single time step.

Running some test cases the analysis of the code performance was done in two steps, the first concerned the vectorization, the second the parallelization.
Vectorization is crucial feature on Xeon Phi platforms in order to exploit the potential of the platform. That is why the first step of the work focused on this aspect. In particular some parts of code are changed and directives to compiler are inserted to favorite the vectorization. The results show that the vectorization has a significant effect of the execution time of time step; in fact the gain due to the vectorization is about 10% against almost 0 in the initial version of the code.

Regarding the second step, a scalability study of the initial version of the code has been done on a single Xeon Phi card. The scalability degraded rapidly for more than 16 processes for our test cases. A profiling of the code with Scalasca has been performed to identify the crucial bottlenecks. It turned that 50% of the time was spent in MPI functions calls. This was due to a bad load balance among the processes. To solve this problem a new feature was introduced in PABLO library, allowing to specify different weights for each cell for better load balance.
Another investigation was done analyzing the large time spent at different calls to MPI_Barrier in the code. The problem was solved reducing the number of barriers and collapsing all the communication at the same moment in the code.
In this modified version of the code the time spent in MPI procedures was reduced, in particular which was reduced from 50% to 10% running on 16 processes.

Business Impact

The project permitted Optimad to put hands on MIC architecture and, through the support of the computing center, to gain insight on the code and its suitability for this architecture. The entire technical staff has a profound knowledge in MPI based parallel computing and during this project this knowledge has been extended to different paradigms suitable for the Intel architectures.

This type of information enables Optimad to program in a more efficient and rational way the transition to heterogeneous architectures, which is considered as a strategic development goal within company, in fact the company is aware that the employed processor (Knights Corner) was only a transition point within Intel’s roadmap for next generation HPC.

The increased experience will help the company to find competitive solutions in terms of reducing computational time and therefore also costs.

Benefits

  • Saved costs per simulations
  • SME now write more efficient codes

Images

Wind Energy

SHAPE Project Vortex Bladeless: Parallel multi-code coupling for Fluid-Structure Interaction in Wind Energy Generation

Project Title

SHAPE Project Vortex Bladeless: Parallel multi-code coupling for Fluid-Structure Interaction in Wind Energy Generation.

Project Details

  • Company name: VortexBladeless – Vortex Bladeless S.L.
  • Country: Spain
  • PRACE partner: BSC, Barcelona, Spain

The Challenge

Fluid-structure interaction problems are multi-physics problems where a fluid interacts with a deformable solid body exerting forces on it. These forces can deform the solid body, which modifies the fluid flow as it moves. This kind of interaction is highly complex and different interesting physical phenomena arise. One of them is the vortex induced vibrations (VIV), which are vibrations in the solid body resulting from the synchronization of the vortex shedding frequency and the natural frequencies of the structure through the so called lock-in effect. This interesting phenomenon has been investigated along years and has been recently used to design new wind energy generators, like the proposed by Vortex-Bladeless.

The numerical simulation of such problems is challenging and has received a lot of attention in the past few years. Different approaches and techniques exist to solve the FSI problems numerically and can be useful to simulate the VIV phenomena coming from the Vortex-Bladeless device. One of these approaches is the staggered multi-code one, where a code is in charge of the fluid flow simulation and another code is in charge of the solid dynamics simulation while the interaction among them comes through the exchange of information of boundary conditions and surface forces. In the present project the Alya code, developed at the Barcelona Supercomputing Center, is adapted and used to perform the Fluid-Structure Interaction (FSI) problem simulation for a scaled experimental vortex-bladeless device, and a comparison between the numerical and experimental results is done.

The Solution

The Alya code was adapted in order to simulate the physical response of a scaled device of the Vortex-Bladeless wind energy generator. The simulation was done using a multi-code coupling approach with a loose coupling algorithm. The comparison between the numerical results and the experimental data provided by the SME is very good, with maximum relative errors of less than 10%. However, the resources needed to perform the simulation are very high at the present, and the code can be improved in order to maintain the accuracy while accelerating the simulation.

The developed tool provided guidance to the SME in the design and understanding of the physical response of the wind energy generator. It is expected to extend the tool in order to simulate the full-scale wind energy generator and give energy production prediction for this device. This SHAPE project has undoubtedly approached the SME to the HPC and provided an excellent framework to achieve the obtained results.

Business Impact

Vortex Bladeless is still developing the technology. However, the SHAPE project has had a very important impact for the business since it helped the SME to improve the development of the device. Both computational fluid dynamics and finite element analysis have allowed for a better understanding of the phenomena under study, implementing the conclusions made and, at the end, confirming the sustainability of the business of Vortex Bladeless in the future.

Benefits

The main benefit has been to be able to use BSC resources (its computing cycles, in-house software and expertise) in order to have a better understanding of the VIV phenomenon underlying the Vortex technology.

Images

Contact SHAPE Team



Comments are closed.