White Papers – SHAPE Projects

Overview of the White Papers from small and medium enterprises (SMEs) who took part in SHAPE Projects.

Title: Shiloh Industries Italia: Multiphasic Simulation in High-Pressure Die Casting Process

Authors: Bernardo Puddu*a, Raffaele Ponzinib, Davide Garettoa
a Shiloh Industries Italia, Verres, Italy
b CINECA, Milan, Italy

Abstract: Shiloh Industries Italia s.r.l. has been working in the market of casted components since 1992 updating manufacturing activities to novel standards imposed by continuous technological improvement. At the moment Shiloh is using standard software to develop new products and support production requirements. The adoption of CFD tools is a mandatory task today when trying to perform daily numerical modelling activities to support production needs with standard commercial ISV applications. Nevertheless, the adopted software workflow does not contain a multiphasic solver, thus it is difficult to analyse in a quantitative way the air entrapment volumes identifying bubble sizes. The present project’s main aim is to test a state-of-the-art HPC infrastructure and open-source technologies to evaluate possible benefits on cost, time-to-results for the single design, and number of achieved designs per day in a multiphasic simulation. A preliminary well-known reference case study will be evaluated to assess the possibility to implement relevant physical parameters into the computational model. Scalability of the CFD solver and automation of the overall identified workflow will also be tested in order to evaluate the portability to Shiloh’s present way of working.

Download paper: PDF

Title: Airinnova: Automation of High-Fidelity CFD Analysis for Aircraft Design and Optimization

Authors: Mengmeng Zhanga*, Jing Gongb, Lilit Axnerb, Michaela Barthb
a Airinnova AB, SE-182 48, Sweden
b PDC Center for High Performance Computing, KTH Royal Institute of Technology, SE-100 44, Sweden

Abstract: Airinnova is a start-up company with a key competency in the automation of high-fidelity computational fluid dynamics (CFD) analysis. Following on from our previous PRACE SHAPE project, we have continued collaborating with the PDC Center for High Performance Computing at the KTH Royal Institute of Technology (KTH-PDC), to investigate the performance analysis of the open source CFD code SU2 and further develop the automation process for the field of aerodynamic optimization and design.

Download paper: PDF

Title: High-fidelity Simulation of an industrial swirling Combustor

Authors: D. Miraa,*, O. Lehmkuhlа, A. Botha, R. Borrellb,I. Bejarb
a Barcelona Supercomputing Center, C/Jordi Girona 29, 08034-Barcelona, Spain
b E & M Combustion, Pol. Bildosola, Pab. F-3 y F-4, Bilbao, Spain

Abstract: The proposed project aims to demonstrate the potentials of high-fidelity simulations to predict pollutant emissions and energy efficiency in practical combustion systems. The study is based on the development of advanced numerical simulations to investigate the reacting flow field of an industrial combustor designed and manufactured by E&M Combustion. The focus is given to the assessment of performance measured in terms of thermal power, combustion efficiency, and global emissions, so the SME can make decisions to optimise the system in the future. In particular, the JBM 4.500 G burner has been simulated using large-eddy simulations with a flamelet combustion model. This project has been developed as collaboration between E&M Combustion and the Barcelona Supercomputing Center (BSC) within the SHAPE programme of PRACE.

Download paper: PDF

Title: AXESSIM – CINES Partnership: HPC for connected Objects

Authors: C. Girarda, B. Weberа, B. Ciroub,*, V. Cameo Ponzb
aAxesSim, 1 rue Jean Sapidus, 67400 Illkirch-Graffenstaden, France, www.axessim.fr, bruno.weber@axessim.fr
b Centre Informatique National de l’Enseignement Superieur (CINES), 950 rue St Priest, 34097 Montpellier, France

Abstract: AxesSim has developed an electromagnetic simulation tool named TETA, which is optimized for harnessing several GPUs in parallel. It is based on the Discontinuous Galerkin method in the Time Domain (DGTD). This method allows to handle meshes of complex geometries and can model designs with high precision such as a full human body with its organs, clothes and surrounding environment. The targeted architecture is a CRAY XC50 (PizDaint) which is a supercomputer with Intel Broadwell CPUs and NVidia P100 GPUs. Both CPU and GPU are addressed via the OpenCL library. MPI is used for communication between accelerators and hosts.

Download paper: PDF

Title: Numerical Modelling of 3D Printed Cranial Orthoses

Authors: P. Marsaleka, A. Grygarb, T. Karasekа,*, T. Brzobohatyа
a IT4Innovations, V SB – Technical University of Ostrava, Ostrava, Czech Republic
b Invent Medical Group s.r.o., Ostrava, Czech Republic

Abstract: The objective of this work is to replace laboratory testing of cranial orthosis design by virtual prototyping using numerical modelling and simulation technologies. In this paper, the focus is put on pre-processing stage of the numerical modelling process. In future Computational Structural Dynamics Simulations will be performed during development of new cranial orthosis design at Invent Medical Group (IMG) company. Stiffness of cranial orthosis will be evaluated by means of numerical modelling and simulation and will replace necessity of physical testing of each and every new design of cranial orthosis to ensure its proper behaviour. Objective of this project is to create a semi-automatic system of mesh generation from the input geometric model based on open source software Netgen Mesh Generator. Geometry of the structure as an input file for meshing is provided in STL format. The output of this project is a software tool which will take a geometric model as an input and produce the finite element mesh with all boundary conditions as an output. The mesh produced is then used for calculation of cranial orthosis stiffness using open source code ESPRESO. Simulation results are verified by comparing with results of the same numerical simulation performed using well-established software package ANSYS and with physical experiment as well.

Download paper: PDF

Title: Airinnova: Automation of High Fidelity CFD Analysis in Aerodynamic Designs

Authors: Mengmeng Zhanga,*, Jing Gongb, Michaela Barthb, Lilit Axnerc
a Airinnova AB, SE-182 48, Sweden
b PDC Center for High Performance Computing, KTH Royal Institute of Technology, SE-100 44, Sweden

Abstract: Airinnova is a start-up company with a key competency in the automation of high fidelity computational fluid dynamics (CFD) analysis. The goal of this SHAPE project, a collaboration with the PDC Center for High Performance Computing at the KTH Royal Institute of Technology (KTH-PDC), was to develop automated procedures for carrying out CFD analysis in the field of aerodynamic optimization and design. The project is a significant step towards automation of the core processes that ordinarily would require specialist skills, such as creation of the simulation mesh, thus assisting a broader sphere of engineers to design aircraft in more efficient and simpler ways.

Download paper: PDF

Title: Milano Multiphysics: Evaluation of the Intel Xeon Phi performances for high fidelity nuclear applications

Authors: Carlo Fiorinaa,*, Giorgio Amatib, Vittorio Ruggierob, Ivan Spissoc
a Milano Multiphysics S.R.L.S, Via Giorgio Washington 96, 20146 Milan, Italy
b SuperCompunting Application and Innovation Department, Cineca, Via dei Tizii, 6 40133, Roma, Italy
c SuperCompunting Application and Innovation Department, Cineca, Via Magnanelli 6/3, 40133, Casalecchio di Reno, Bologna, Italy

Abstract: Milano Multiphysics is a start-up active in the field of advanced modelling of complex systems. The goal of this SHAPE project was to evaluate the pros and cons of the Intel’s Many Integrated Core technology in the highly demanding field of nuclear engineering. More specifically, the performances of the Intel Xeon Phi processors have been evaluated for two commonly used tools in the nuclear engineering community, namely: the OpenFOAM finite-volume library, and the Serpent Monte Carlo code. Some notable advantages compared to traditional Intel Xeon processors have been observed in terms of both speed-up and energy consumption. However, these advantages are limited to the case of OpenFOAM while Serpent seems to perform worse on Xeon Phi processors, which is partly due to the impossibility to vectorise traditional implementations of Monte Carlo algorithms. Additional margins of improvements for both codes may actually come from a better use of vectorization-friendly algorithms.

Download paper: PDF

Title: Scienomics Development of Chameleon Monte Carlo code for HPC: Toward Realistic Modelling of Complex Composite Systems

Authors: Orestis Alexiadisa, Xenophon Krokidisa, Isabelle Dupaysb, Sylvie Therondb, Thibaut Veryb,*
a Scienomics
b Institut du developpement et des ressources en informatique scientifique (IDRIS), CNRS, Universite Paris-Saclay, F-91403, Orsay, France

Abstract: Scienomics is developing Chameleon (Chain Altering Monte-Carlo) to perform Monte-Carlo (MC) simulations of several types of chemical systems such as polymers, nanomaterials or composite materials. Realistic modelling of materials requires that the systems are large enough to get reliable results for the properties computed. Chameleon uses an all-atoms model meaning that the number of interactions needed to compute the physical properties, hence the computational resources, increases fast as the number of atoms in the system increases. In this project the work of Scienomics and IDRIS engineers allowed Chameleon to treat larger systems faster thanks to a dramatic improvement of serial and parallel performances.

Download paper: PDF

Title: Renuda: Optimising 2D flow for faster, better steam turbine design

Authors: N. Tonelloaa, P. Asproulisaa, K. Stratfordb, N. Banglawalab, P.Grahamb,*
a Renuda, UK Ltd
b EPCC, University of Edinburgh, UK

Abstract: This project was a collaboration between Renuda UK Ltd and EPCC, the High Performance Computing centre at the University of Edinburgh. The goal was to improve the performance of Renuda’s industrial and power generation steam turbine modelling code, referenced as CodeX for this project. This report outlines the work that has taken place towards this goal, including the technical approach, code analysis and final results.
The code has been investigated, with initial benchmarks and analysis used to identify areas for improvement and to identify the most suitable optimisation approach to take. CodeX has been refactored and restructured to allow for the parallelisation of the code, but also to make improvements in the serial performance. The code has been parallelised using OpenMP directives, ensuring portability across platforms.
Benchmarks performed during the project indicate that for runs of typical operational models, the optimised serial code is over twice as fast as the original. For parallel runs this enhancement is reflected further, with a headline figure of over 27x faster than the original code on 16 cores – this equates to a reduction in runtime from over 1.5 days to less than 90 minutes. The end result of this process is an optimised, parallelised version of CodeX, which can be used to perform simulations in a significantly shorter timescale, thus enabling Renuda to offer enhanced services to their customers.

Download paper: PDF

Title: AmpliSIM DemocraSIM: Enable air quality Simulation as a Commodity

Authors: Olivier Oldrinia,*, Sylvie Perdriela, Sylvie Therondb, Isabelle Dupaysb
a AmpliSIM,Paris, France
b CNRS/IDRIS, Orsay, France

Abstract: AmpliSIM, a French SME expert in numerical simulation services and project partner at IDRIS, collaborated in the SME HPC Adoption Programme in Europe (SHAPE), organised within the PRACE-4IP project. The aim of the project, DemocraSIM, is to enable air quality simulation as a commodity. To do so, AmpliSIM provides air quality modellers with a web service which allows the users to configure / launch / explore air quality numerical simulations without any requirements for software installation or an in-house calculation cluster. In the framework of this project, calculations were performed on the IDRIS supercomputer which allowed fast calculation using parallelisation. Thanks to IDRIS expertise, 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 post processing. Moreover, the capabilities of the web portal to connect to an HPC infrastructure were substantially improved, including improved security of the portal in order to meet IDRIS standards. Finally, the capability of the web portal to support concurrent user access was strengthened to offer excellent scalability properties.

Download paper: PDF

Title: Anemos SRL: SUNSTAR Simulation of UNSteady Turbulent flows for the AeRospace industry

Authors: D. Modestia, V. Ruggierob, S. Pirozzolic
a CNAM Paris
cAnemos SRL

Abstract: The SUNSTAR project aims at improving the robustness and testing the predictive accuracy of an in-house CFD software for the simulation and the analysis of turbulent high-speed flows relevant for the aerospace industry, exploiting the HPC premises provided by PRACE. The software kernel is a compressible parallel flow solver, incorporating state-of-the-art numerical methods and advanced features for the direct numerical simulation of supersonic flows. The engineering of the code for industrial applications prompts efficient exploitation of modern computer architectures, which in turn requires optimization of vector and parallel efficiency, while conserving numerical accuracy. Numerical tests show that these results have been achieved.

Download paper: PDF

Title: Creo Dynamics: Scale Resolving CFD and CAA processes for Ground Vehicles based on Open Source

Authors: Torbjorn Larssona*, Johan Hammara, Jing Gongb, Michaela Barthb, Lilit Axnerb
a Creo Dynamics AB, Westmansgatan 37A, SE-582 16 Linkoping, Sweden
b PDC Center for High Performance Computing, KTH Royal Institute of Technology, SE-100 44, Stockholm, Sweden

Abstract: Creo Dynamics [1] is a Swedish engineering company with core competence in fluid dynamics, acoustics and structural dynamics. Creo Dynamics has broad experience from participating in national and international research programmes focussing on the development of new emerging technologies for the automotive or aerospace industries. In the company, experienced engineers develop and deliver simulation tools and procedures – often based on open source software [2] – which are tailored towards specific needs and applications; this process always involves maintaining a conscious balance between turn-around time and accuracy.
Creo Dynamics has recently undertaken a PRACE SHAPE project together with application experts from the KTH Royal Institute of Technology. During the course of the project we developed templates and “recipes” for a range of tasks that are involved in the automotive industry – such as automated handling of computer-aided design (CAD), parallel meshing, solving and post-processing. These were all tailored towards a particular automotive application, namely the aerodynamics of a heavy-duty semi-trailer. We also focused on parallel implementations and executions for large-scale simulations. In addition to monitoring the efficiency of the processes (for meshing and solving run-time performance and scalability) and identifying critical bottlenecks, we gave significant attention to pinpointing performance deficits in the processes so as to give guidance for a further fine-tuning of the overall methodology.

Download paper: PDF

Title: ACOBIOM – CINES Partnership: MARS: Matrix of RNA-Seq

Authors: D. Piquemalb, R. Brunob, V. Cameoa, B. Ciroua1, L. Manchonb, F. Pierratb
a Centre Informatique National de l’Enseignement Superieur (CINES), 950 rue St Priest, 34097 Montpellier, France
b ACOBIOM, CS77394, 1682 rue de la Valsiere, 34184 Montpellier Cedex 4, France

Abstract: RNA-Seq approach is used in a wide variety of applications. These include identifying disease-related genes, analysing the effects of drugs on tissues, and providing insight into disease pathways. The RNA-Seq is widely used to characterise gene expression patterns associated with tumor formation. Since RNA-Seq provides absolute values and does not require any calibration with arbitrary standards, results can be compared at any time with other data, even raised by independent laboratories. Once collected, this data can be digitalised and then easily and reliably compared in silico with the growing library of RNA-Seq databases generated for normal and pathological situations in other laboratories around the world (Human: ~27000 libraries. Average size of a library: 1.7GB. Total size: 120TB).
The objectives of MaRS database is to provide and centralize a standardized data used in a wide variety of applications like identifying disease-related genes, analysing the effects of drugs on tissues or providing insight into disease pathways.

Download paper: PDF

Title: WB-Sails: Modelling Sail Boat Performance in HPC

Authors: Mikko Brummera*, Ruddy Brionnaudb, Jorge Ginesb, Jussi Heikonenc, Jyrki Hokkanenc, Esko Jarvinenc, and Jarmo Pirhonenc
a WB-Sails Ltd Oy (WBS)
b Next Limit Dynamics (NLD)
c CSC – IT Center for Science Ltd. (CSC)

Abstract: In this PRACE SHAPE project WB-Sails wished to implement a HPC based CFD analysis of sails and sailboats, using the XFlow dynamic simulation code. The code allows 6 degrees of freedom (DOF) for motion of the objects, for a realistic simulation of forces and moments around a boat moving in seaway.
Coupled with a FEA solver, full fluid-structure analysis including optimization is possible.

Download paper: PDF

Title: FDD Engitec S.L.: Pressure drop simulation for a compressed gas closed system

Authors: M. Vazqueza*, O. Lehmkuhlb, A. Pozob, J. Bragadob
a Barcelona Supercomputing Center – Centro Nacional de Supercomputacion, Spain
b FDD Engitec S.L., Spain

Abstract: The objective of the project was to calculate by simulation, the pressure drop of a firefighting pressure regulated discharge valve for inert gas agent. For doing that, we simulate the discharge of a pressurized closed canister through the valve under realistic conditions.
This project has been done in collaboration with the Barcelona Supercomputing Center (BSC). The code used to perform the numerical simulations is Alya, a multi-physics code developed at BSC. The code has been adapted to solve this application problem.

Download paper: PDF

Title: Pharmacelera S.L.: HPC Methodologies for PharmS

Authors: Aleksey Kondratyeva*, Thomas Ponweisera*, Enric Gibertb, Enric Herrerob
a RISC Software GmbH, Softwarepark 35, 4232 Hagenberg, Austria
b Pharmacelera S.L., Pl. Pau Vila, 1, Edifici Palau de Mar, 08039 Barcelona, Spain

Abstract: Pharmacelera is a company that develops hardware and software solutions for drug discovery. Among other products, Pharmacelera offers PharmScreen, a revolutionary software tool for ligand-based drug design. PharmScreen scans compound databases consisting of hundreds of thousands of molecules and identifies potential hits by comparing molecules using a full 3D representation of its interaction fields. In the course of the SHAPE project “HPC methodologies for PharmScreen”, Pharmacelera has been exploring the potential of heterogeneous HPC platforms for PharmScreen and, in cooperation with RISC Software, has been working on (1) performance profiling and analysis, (2) porting computation kernels to OpenCL, enabling the usage GPU, Xeon Phi and other accelerator platforms, (3) porting the application to MareNostrum, (4) static code analysis and refactoring, and (5) improving the testing methodology.

Download paper: PDF

Title: BAC Engineering Consultancy Group: Numerical Simulation of Accidental Fires with a Spillage of Oil in Large Buildings

Authors: C. Munoza*, D. Vicenteb*
a BAC Engineering Consultancy Group, Spain
b Barcelona Supercomputing Center – Centro Nacional de Supercomputacion, Spain

Abstract: BAC’s (BAC Engineering Consultancy Group) participation in PRACE programme has made possible to increase its knowledge in the field of fire engineering and the application of advanced computing strategies. These fields may be applied directly to a fire engineering analysis (Performance-Based design) of the steel structure buildings that belong to the International Thermonuclear Experimental Reactor industrial complex in France, which is devoted to research in the field of Nuclear Fusion. In particular, the project could be focused on the building that hosts the cryogenic system of the Tokamak. This building is characterized by its large size. The building hosts a large number of helium and nitrogen compressors, as well as the charcoal filter system, that is necessary for the cryogenic system.
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 the combination of a liquid combustible spillage added to an oil cloud due to the breakage of a high pressure pipe.

Download paper: PDF

Title: Algo’Tech: A Successful Switch to HPC

Authors: A. Pere-Lapernea, P. Rametb
a Algo’Tech Informatique, Bidart, France1
b Inria, HiePACS team, Bordeaux, France2

Abstract: Manufacturers increase the use of electrical and electronic components in all kinds of products to provide new functions, improve performance and reduce the operational and development costs. The counterpart is the introduction of new constraints such as the electromagnetic vulnerability where the equipment functioning can be disturbed by electromagnetic effect. The perturbation can happen in various conditions: (1) Inside an electrical harness between power supply wire and command wire; (2) Between electrical harnesses; (3) Or because of external sources (electrical equipment, engines, Wi-Fi etc.) or natural ones like lightening. This concerns machine design, robotic assembly lines, devices for commercial buildings and public facilities, embedded systems (aircraft, trains, boats, tramways, drones), and in general, all automatic devices and electrical systems. To reduce the time to market, manufacturers need tools able to: (1) Detect bad electromagnetic effects as soon as possible: before prototyping, during the design phase; (2) Avoid cable shielding when it is not necessary: lighter equipment, less manufacture operations (welding). An electromagnetic simulator, integrated into an electrical CAD software suite, used during the design phase, on a simple PC connected to HPC, would allow the manufacturers to adapt the design to electromagnetic constraints before building the first prototype. As of today, they generally identify the problems during testing phase, if not on the customer site during the installation with all consequences behind that.

Download paper: PDF

Title: Airinnova: Automation of High Fidelity CFD Analysis in Aerodynamic Design

Authors: Mengmeng Zhanga*, Jing Gongb, Michaela Barthb, Lilit Axnerb
a Airinnova AB, SE-182 48, Sweden
b PDC Center for High Performance Computing, KTH Royal Institute of Technology, SE-100 44, Sweden

Abstract: Airinnova is a start-up company with a key competency in the automation of high fidelity computational fluid dynamics (CFD) analysis. The goal of this SHAPE project, a collaboration with the PDC Center for High Performance Computing at the KTH Royal Institute of Technology (KTH-PDC), was to develop automated procedures for carrying out CFD analysis in the field of aerodynamic optimization and design. The project is a significant step towards automation of the core processes that ordinarily would require specialist skills, such as creation of the simulation mesh, thus assisting a broader sphere of engineers to design aircraft in more efficient and simpler ways.

Download paper: PDF

Title: Cybeletech: Parallelization and optimisation for plant selection with Cybeletech

Authors: B. Ciroua*, D. Fernandezb, G. Hautreuxb, D. Woutersb
a Centre Informatique National de l’Enseignament Superieur (CINES), 950 rue Stt Priest, 34097 Montpellier, France
b Cybeletech, 2 rue de la Piquetterie, 91680 Bruye`res le Cha^tel, France

Abstract: Breeding a new variety is a long process that requires a decade and thousands of experimental trials in fields so as to select the most robust and efficient traits. Some steps in the process of plant selection could be conducted in-silico to reduce the duration and development cost of a new variety.
The plant growth model used in numerical simulations must be calibrated with plant phenotypes data. To define the optimal experimental protocol to be followed for calibrating the model, the model is run with a genetic algorithm.
Optimisation of the plant growth model enabled to reduce its computing time by a factor five. Performances of the whole application were highly improved by implementing a master-slave approach to the optimisation of the evolutionary algorithm. These performances are function of the number of protocols and realisations considered.

Download paper: PDF

Title: Design Methods:Coupled Sail and Appendage Design Method for Multihull Based on Numerical Optimisation

Authors: Ubaldo Cellaa*, Francesco Salvadoreb, Raffaele Ponzinib
a Design Methods(www.designmethods.aero)

Abstract: An optimisation procedure for catamarans sail plan and appendages is descripted. The method integrates a parametric CAD model, an automatic computational domain generator and a Velocity Prediction Program (VPP) based on a combination of sail RANS computations and analytical models. The sailing speed and course angle are obtained, with an iterative process, solving the forces and moment equilibrium system of equations. Hull forces analytical formulations were developed and tuned against a matrix of CFD solutions. The appendages aerodynamic polars are estimated applying preliminary design criteria from aerospace literature. The procedure permits to find the combination of appendages configuration, rudders setting, sail planform, shape and trim that maximise the VMG (Velocity Made Good). Within the SHAPE programme, the possibility to implement the procedure using Open-Source software is investigated.

Download paper: PDF

Title: Ingenieurburo Tobias Loose: HPCWelding: Parallelized Welding Analysis with LS-DYNA

Authors: T. Loosea*, M. Bernreutherb, B. Gro?e-Wohrmannb, J. Hertzerb, U. Gohner c
a aIngenieurburo Tobias Loose, Herdweg 13, D-75045 Wossingen, www.tl-ing.eu
b Universitat Stuttgart, Hochstleistungsrechenzentrum, Nobelstra?e 19, D-70569 Stuttgart, www.hlrs.de
c DYNAmore Gesellschaft fur FEM Ingenieurdienstleistungen mbh, Industriestra?e 2, D-70565 Stuttgart, www.dynamore.de

Abstract: In this paper the results of the PRACE SHAPE project “HPC Welding” are presented. During this project, a welding structure analysis with the parallel solvers of the LS-DYNA code was performed by Ingenieurburo Tobias Loose on the Cray XC40 “Hazel Hen” at the High Performance Computing Center (HLRS) in Stuttgart. A variety of test cases relevant for industrial applications have been set up with DynaWeld, a welding and heat treatment pre-processor for LS-DYNA, and run on different numbers of compute cores. The results show that the implicit thermal and mechanical solver scales up to 48 cores depending on the particular test case due to unbalanced workload. The explicit mechanical solver was tested up to 4080 cores with significant scaling. As we know, it was the first time that a welding simulation with the LS-DYNA explicit solver was performed on 4080 cores.

Download paper: PDF

Title: Open Ocean: High Performance Processing Chain – faster on-line statistics calculation

Authors: L. Fallettia*, S. Theronda, Y. Moisanb, Y. Kervellab
a IDRIS – Orsay, France
b Open Ocean, France

Abstract: Open Ocean is a French SME company which develops innovative on-line solutions to help plan and manage offshore developments. They conceived an oceanographic data study tool which computes and formats data (Pre-Processing and Processing) and which provides relevant oceanographic information to industrial marine companies (Post-Processing) through a web interface. But the “time-to-solution” of this post-processing step is too long and hence not compatible with industrial use. Therefore, the goal of this SHAPE project is to improve post-processing by optimising a parallelized Python program of Open Ocean which processes and computes statistics (e.g. wind speed) on big datasets. To carry this out, engineers of Open Ocean and IDRIS worked together to optimise this program by using resources available in a national supercomputing centre: high performance parallel machine and parallel file system (GPFS, 100 GB/S bandwidth). This paper describes the parallelisation process implemented by Open Ocean and its porting on the Ada machine (IBM cluster of Intel E5-4650 processors, 332 compute nodes) at IDRIS. It also covers performance testing and identification of the bottleneck in the execution.

Download paper: PDF

Title: Optimad Engineering srl.: RAPHI: Rarefied Flow Simulations on Xeon Phi Architecture

Authors: Haysam Teliba1, Marco Cisterninoa, Vittorio Ruggierob, Florian Bernardc
a Optimad Engineering srl.,
b Cineca
c Inria

Abstract: KOPPA (Kinetic Octree Parallel PolyAtomic) is a parallel numerical code for the simulation of rarefied gas dynamics. It is based on a library named PABLO (PArallel Balanced Linear Octree) used to manage octree grids in parallel. The main issue with such numerical codes is the very high execution time which can become prohibitive for some industrial applications. Thanks to the SHAPE project, important improvements have been achieved with respect to execution time and scalability. In particular, some parts of the code have been reimplemented to suit better a MIC (Multi Integrated Cores) architecture. So far, the computational time requirements have been decreased by a factor of almost 8 and a good scalability has been obtained up to 64 processors against 16 initially.

Download paper: PDF

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

Authors: J.C. Cajasa*, G. Houzeauxa, D.J. Yanezb, M. Mier-Torrecillaa
a Barcelona Supercomputing Center – Centro Nacional de Supercomputacion, Spain,
b Vortex Bladeless S.L., Spain

Abstract: Vortex-Bladeless is a Spanish SME whose objective is to develop a new concept of wind turbine without blades called Vortex or vorticity wind turbine. This design represents a new paradigm in wind energy and aims to eliminate or reduce many of the existing problems in conventional generators. Due to the significant difference in the project concept, its scope is different from conventional wind turbines. It is particularly suitable for offshore configuration and it could be exploited in wind farms and in environments usually closed to existing ones due to the presence of high intensity winds. The device is composed of a single structural component, and given its morphological simplicity, its manufacturing, transport, storage and installation has clear advantages. The new wind turbine design has no bearings, gears, etcetera, so the maintenance requirements could be drastically reduced and their lifespan is expected to be higher than traditional wind turbines. It is clear that the proposed device is of prime interest, and that scientific investigation of the response of this wind energy generator under different operation scenarios is highly desirable. Thus, the objective of this SHAPE project is to develop the needed tools to simulate Fluid-Structure Interaction (FSI) problems and to reproduce the experimental results for scaled models of the Vortex-Bladeless device. In order to do so the Alya code, developed at the Barcelona Supercomputing Center, is adapted to perform the Fluid-Structure Interaction (FSI) problem simulation. The obtained numerical results match satisfactorily with the experimental results reported.

Download paper: PDF

Title: Hydros Innovation: Automatic Optimal Hull Design by Means of VPPApplications on HPC Platforms

Authors: S. Dyena*, A. Lukowskia, R. Ponzinib
a aHydros Innovation, Switzerland
b CINECA, Italy

Abstract: Hydros is an Engineering & Research Swiss company founded in 2007 with several patented designs in the field of marine and sailing yachting. In recent years Hydros is exploring new market segments such as yachts and super-yachts hull design. The usage of HPC resources and open-source softwares can be a valuable tool in the massive shape design optimizations usually performed by Hydros. The main scope of the proposed SHAPE pilot was therefore 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. To accomplish this task the project was subdivided into a set of steps including a preliminary validation of a 2DoF CFD analysis of an industrial hull design using open-source code, the scalability test of commercial and open-source CFD code for hull 2DoF modelling, the coupling of the CFD result into an existing CAD modification and optimization loop, and finally the running of a complete optimization loop for an industrial hull design using open-source code on the HPC platform and usability evaluation of the solution provided.

Download paper: PDF

Title: Ergolines – ITU Partnership: HPC-based Design of a Novel Electromagnetic Stirrer for Steel Casting

Authors: Isabella Mazzaa, Ahmet Duranb, Yakup Hundurc, Cristiano Persia, Andrea Santoroa, Mehmet Tuncelb
a Ergolines s.r.l., Area Science Park, Padriciano 99, 34016, Trieste, Italy
b Istanbul Technical University (ITU), Mathematical Engineering, 34469 Sariyer, Istanbul, Turkey.
c Istanbul Technical University (ITU), Physical Engineering, 34469 Sariyer, Istanbul,.

Abstract: Ergolines, an Italian SME expert in the design of electromagnetic stirrers for metal casting, and the project partners at ITU collaborated in a European research project under the EU’s Horizon 2020 Research and Innovation Programme, in particular under the SME HPC Adoption Programme in Europe (SHAPE), organized within the PRACE Research Infrastructure. Custom codes were developed for HPC-based magnetohydrodynamics (MHD) simulations, enabling the design of a dedicated electromagnetic stirrer (EMS) for the electric arc furnaces (EAF). We performed parallel simulations using an OpenFOAM solver and other related programs on IBM-FERMI (a PRACE Tier-0 system) at CINECA, Italy. The fluid-dynamics of liquid steel within the EAF under the effect of electromagnetic stirring has been studied under different simulation parameters.

Download paper: PDF

Title: AMET s.r.l.: ROSPA: Robustness in Safety Performances Analysis

Authors: M. Rabitob, C. Arlandinia, I. Spissoa, A. Vannia, P. Cavalloa
a CINECA – Centro di Supercalcolo – Consorzio Interuniversitario
b AMET s.r.l.

Abstract:Fuel consumption issues are currently forcing car manufacturers to increase the use of lightweight
materials in the development of new vehicles. Among those materials, composites are the most
interesting solution. Unfortunately, they show a spread in performance characteristics due both to the
manufacturing process and to the material providers, thus limiting the use of this material to niche
productions. This issue could be solved once a more robust design is in place, taking into account the
scattering of material characteristics when evaluating global performances. This document describes
the work done in the framework of the PRACE SHAPE pilot experiments, aiming to develop with the
help of a HPC approach a new robust methodology (where a robust approach means a statistical one,
compared to the old deterministic approach) to assess the behavior of composite materials with respect
to their resistance to impact load cases, which usually drive vehicle design.

Download paper: PDF

Title: SHAPE Pilot Lapcos: Virtual Test Bench for Centrifugal Pumps

Title: OPTIMA pharma GmbH: Enhanced Airflow Simulations around Filling Machines in Clean Rooms

Authors: R. Eisenschmida*, B. Groe-Wohrmannb
aOPTIMA pharma GmbH, Otto-Hahn-Strasse 1, 74523 Schwabisch Hall, Germany
bHigh Performance Computing Center (HLRS), Nobelstr. 19, 70569 Stuttgart, Germany

Abstract: In this white paper, the main results of the SHAPE pilot project Enhanced airflow simulations around filling machines in clean rooms are presented. During this project on the Tier-0 system Cray XE6 Hermit at HLRS, Germany, the open source CFD software package OpenFOAM v2.2.2 was utilized to run simulations meeting the requirements of industrial production. Besides testing different turbulence models, emphasis was placed on parallel mesh generation with snappyHexMesh and the decomposition and reconstruction of large meshes with more than 10 million cells.
Furthermore, the cooperation between PRACE and the SME OPTIMA pharma GmbH is described, and a report on the benefits for the SME, the lessons learned and the future activities is added.

Download paper: PDF

Title: NSilico Lifescience Ltd: Novel HPC Technologies for Rapid Analysis in Bioinformatics

Authors: Tristan Cabela, Gabriel Hautreuxa, Eric Boyera, Simon Wongb, Nicolas Mignereyc, Xiangwu Lud, Paul Walshd
aCentre Informatique National de l’Enseignement Superieur, 950, rue de Saint Priest, 34097 Montpellier Cedex 5, France.
bIrish Centre for High-End Computing, 7/F Tower Blg., Trinity Technology & Enterprise Campus, Grand Canal Quay, Dublin 2, Ireland.
cGrand Equipement National de Calcul Intensif, 12, rue de l’Eglise, 75015 Paris, France.
dNSilico Lifescience Ltd., Melbourne Building, CIT Campus, Bishopstown, Cork City, Ireland.

Abstract: NSilico is an Irish based SME that develops software for the life sciences sector, providing bioinformatics and medical informatics systems to a range of clients. One of the major challenges that their users face is the exponential growth of high-throughput genomic sequence data and the associated computational demands to process such data in a fast and efficient manner. Genomic sequences contain gigabytes of nucleotide data that require detailed comparison with similar sequences in order to determine the nature of functional, structural and evolutionary relationships. In this regard NSilico has been working with computational experts from CINES (France) and ICHEC (Ireland) under the PRACE SHAPE programme to address a key problem that is the rapid alignment of short DNA sequences to reference genomes by deploying the Smith-Waterman algorithm on an
emerging many-core technology, the Intel Xeon Phi co-processor. This white paper will discuss some of the parallelisation and optimisation strategies adopted to achieve performance improvements of the algorithm keeping in mind both existing and future versions of the hardware. The outcome has been an extremely successful collaboration between PRACE and NSilico, resulting in an implementation of the algorithm that can be readily deployed to realise significant performance gains from the next generation of many-core hardware.

Download paper: PDF

Title: Monotricat SRL: Hull resistance simulations for an innovative hull using OpenFOAM

Authors: Lilit Axnera,b, Jing Gonga,b, Alessandro Chiarinic, Luigi Mascellarod
aPDC-HPC, KTH Royal Institute of Technology, Stockholm, SE-10044, Sweden
bThe Swedish e-Science Center (SeRC), KTH Royal Institute of Technology, Stockholm, SE-10044, Sweden
cSuper Computing Solutions, CINECA, Bologna, 40033, Italy
dMonotricat, Via Fosso della Castelluccia 146/22, Roma, 00134, Italy

Abstract: In the project we conduct a whole hull simulation of a ship, (including mesh generation of complex geometries, efficient solvers with various turbulent modelling and optimized parameters as well as visualization), along with a performance analysis for the simulation. This project makes it possible for Monotricat to take advantage of HPC-enabled simulation tools and potentially replace the traditional methods.

Download paper: PDF

Title: Juan Yacht Design: Testing LES turbulence models in race boat sails

Authors: Herbert Owena*, Patricia Izaguirreb, Gonzalo Kouyoumdjianb and Mariano Vazqueza
aBarcelona Supercomputing Center
bJuan Yacht Design

Abstract: The objective of this project is to implement LES turbulence models outside the academic world to simulate flow around sailsIto replace RANS modelsIthatIare the standard in the industry. The implementation and testing in the finite code ALYA is done by the Barcelona Supercomputing Center so that Juan Yatch Design SL (JYD) can appreciate the advantages of using a LES formulation for their problem. The obtained results show that LES can provide significant advantages over RANS simulation] at least for the cases we have studied. There is a significant difference in the predicted forces that can be related to the fact that RANS can not accurately capture two vortices created at the top and bottom of the Genoa sail.

Download paper: PDF

Title: ENTARES Engineering: Electromagnetic simulation for large model using HPC

Authors: Jose Maria Tamayoaa, Pascal de Resseguiera
aENTARES Engineering – Nexio Group, 48 rue Rene Sentenac, 31300 TOULOUSE – FRANCE

Abstract: We have started the extension of the electromagnetic software CAPITOLE developed at ENTARES Engineering to HPC machines. This has been possible with the aid of the SHAPE pilot project, which has provided us with some expertise and computation hours at Marenostrum III. Two numerical methods have been addressed to solve the resulting dense MoM linear system, MSCBD and MLACA. A new implementation based on asynchronous tasks has been performed for the direct method MSCBD. Dependencies between tasks need to be well defined before moving to a runtime scheduling such as STARPU. As for the iterative method MLACA, a hybrid MPI-OpenMP parallelization has been done with excellent results. So far, electromagnetic models up to 6 Million unknowns have been properly solved.

Download paper: PDF

Title: Albatern: Numerical Simulation of Extremely Large Interconnected Wavenet Arrays

Authors: William Edwardsa, David Findlaya, David Scottb, Paul Grahamb
aAlbatern Ltd.

Abstract:Albatern develop novel interconnected offshore marine renewable energy devices. The goal of this SHAPE project was to begin the development of a physics code capable of simulating a large scale Wavenet array (100 or more devices) using HPC technology to extensively parallelise the solution. The simulation capability is intended to build on Albatern’s in house modelling expertise, allowing the prototyping of arrays that are currently not possible to simulate due to their scale. Computer visualisation and power prediction of large scale arrays are also vital to the success of Albatern’s efforts to continue investment in the technology. To facilitate development, the project was split into an algorithm development task and a HPC implementation task. The completed HPC simulation method will capture fundamental dynamic phenomena associated with arrays of wave energy devices while providing a time and cost effective development technology.

Download paper: PDF

Title: AUDIONAMIX: A benchmark of linear algebra libraries for HPC

Autjors: Sebastien Eggenspieler, Guillaume Vincke, Pierre Leveau

Abstract: The source separation technology developed by Audionamix requires a significant amount of computation. To increase the speed of the technology, several High Performance Computing libraries for linear algebra have been benchmarked with two applications: a simple matrix multiplication and a Non-Negative Matrix Factorization.
The benchmarked libraries were Eigen, Armadillo, MKL and the GPU library Magma. The benchmark returned that for both applications the GPU-based solution performed better than the other solutions. But among CPUbased solutions, Eigen performed better on a complex task than MKL. Thus, making the Magma library as an Eigen back-end seems to have a good potential for HPC-based linear algebra, since this solution would benefit at the same time from the GPU-accelerated computation of Magma and from Eigen’s optimized data transfer.

Download paper: PDF

Title: Thesan srl: Design improvement of a rotary turbine supply chamber through CFD analysis

Authors: R. Ponzinia, A. Penzaa, R. Vadorib, B. Puddub
aCINECA, SCAI Department – (ITALY)
bThesan srl – (ITALY)

Abstract: This work conducted in the field of the PRACE’ SHAPE pilots deals with the optimization of a volumetric machine. The machine is under active development, and a prototype is already working and fully monitored in an experimental mock-loop setup. This prototype operates under controlled conditions on a workbench, giving as an output the efficiency of the machine itself. The main goal is to obtain an increased efficiency through the design and realization of the moving chambers in which fluid flows. In order to obtain such a task, an extensive CFD modelling and simulation is required to perform virtual tests on different design solutions to measure the physical quantities assessing the performance of a given geometry. The final goal is to design a better geometry of the different components, mainly the supply and exhaust chambers, cutting down time and resources needed to realize a physical prototype and to limit the physical realization only on a single geometry of choice. The CFD modelling should allow then, through scientific visualization paradigms and quantifications, to perform a detailed characterization of the fluid dynamics patterns present within the prototype and to identify the main geometrical parameters able to drive the optimal configuration. High Performance Computing facilities and Open-Source tools, such as OpenFOAM, are therefore of capital interest to handle the complex physical model under consideration and to perform a sufficient amount of design configuration analysis.

Download paper: PDF

Title: Axyon – A scalable HPC Platform for AI Algorithms in Finance

Authors: Fabio Franzosoaa, Riccardo Folloniaa, Jacopo Crediaa, Eric Pascolobb
aAxyon AI S.R.L, Italy
bCINECA Consorzio Interuniversitario, Italy

Abstract: With the advent of Big Data, the increasing use of AI algorithms in finance and the consequently fast-growing need for computational power, it has become necessary for FinTech companies to use large scalable HPC systems able to handle and explore a wide array of Machine Learning models (e.g. deep neural networks) and to remain up-to-date with state-of-the-art methods.

With this work, we wanted to maximize the efficiency of accessing different types of remote computational resources potentially available to our proprietary Machine Learning platform, without losing the flexibility provided by in-house compute power. This is a mandatory requirement for a FinTech company oftentimes working with proprietary data that cannot be uploaded to cloud systems.

In this whitepaper, we outline the development of a scalable and flexible DB-centric Master-Slave system architecture able to exploit any connected internal or external computational resource (including an HPC cluster) in a flawless and secure way.

Download paper: PDF

Title: Vision-e: Deep Tile Inspection


  • Rudy Melli, aVision-e S.r.l.
  • Eric Pascolob, CINECA

Abstract: Vision-e is an Italian company focused on machine vision application. One of the main market is the ceramic tile where inspection system is used to automatically check the quality of the product. The fourth industry revolution tends to create factories with few human controllers and need automatic systems without configuration or supervision. This project goes in that direction trying to increase the performance of actual machine vision systems we created, in particular the clustering phase that is crucial for a good inspection. We investigated Deep Learning technologies applied to clustering tiles using the DeepCluster approach and the good results we get, show us that this is the right way to reach the goal and to continue, starting from a study on the choice of input data.

Download paper: PDF

Title: HPC Simulation of Particles Deposition in Human Airways Authors
  • B. Eguzkitza, Barcelona Supercomputing Center, Barcelona, Spain
  • C. Butakoff, Elem Bio-Tech, Barcelona, Spain
  • G. Houzeaux, Barcelona Supercomputing Center, Barcelona, Spain
  • M. Vázquez, Barcelona Supercomputing Center, Barcelona, Spain
  • B. Mignot, Fluidda, Kontich, Belgium
Abstract: The goal of this project was to optimise the Alya HPC simulation code for the problem of airflow and particle deposition in human airways, validate its results and establish a pipeline for the use of HPC in the design of treatments for respiratory diseases. High-resolution human airway model was constructed using computed tomography (CT) images of a patient. Using the patient data, it is possible to assess how inhaled drugs (represented by particles) are transported and identify the areas of deposition, which can be compared to the deposition data acquired in vivo. The focus is given to the assessment of performance measured in terms of efficiency of the code in HPC context and validation. This project has been developed as collaboration between Fluidda and the Barcelona Supercomputing Center (BSC) within the SME HPC Adoption Programme in Europe (SHAPE) programme of PRACE. Download paper: PDF

Title: Improvement of cardiovascular stent design through advanced numerical simulation


  • E. Casonia, Barcelona Supercomputing Center, Barcelona, Spain
  • M. Vázquez, Barcelona Supercomputing Center, Barcelona, Spain and Elem Bio-Tech, Barcelona, Spain
  • A. Vidal, LVD Biotech, Barcelona, Spain

Abstract: The project aims to demonstrate the capability of advanced numerical simulation based on Finite Element modelling to investigate transcatheter aortic valve implantation (TAVI) designs, deployment geometries and the stent’s capacity to be loaded and unloaded. Numerical simulation appears as a cheap option to study the devices’ behaviour and the different factors that may influence it. In this study, it is proposed as a support and guidance for the design process. The focus is given to the assessment of the loading capacity of the stent into the device, deployment of the stent and interactions with the deployment device. These studies require large computational models which are also highly non-linear. In that sense, high performance computing can help to reduce computational time and increase the number of cases to study without increasing expensive experimental setup.

The objective of this study is to provide a tool to simulate TAVI behaviour and help the SME to optimise the design in the future. In particular, this project aims at investigating how different modelling strategies in terms of material model, geometric properties and discretisation procedures can impact analysis results. In order to do so the Alya code, developed at Barcelona Supercomputing Center, is adapted to perform the simulations.

This project has been developed as collaboration between LVD Biotech and the Barcelona Supercomputing Center (BSC) within the SHAPE programme of PRACE.

Download paper: PDF

Title: CFD Simulations in HPC Immersion Cooled


  • G. Houzeaux, Barcelona Supercomputing Center, Barcelona, Spain
  • J. Pita, SUBMER Immersion Cooling, Spain
  • O. Lehmkuhl, Barcelona Supercomputing Center, Barcelona, Spain
  • S. Gómez, Barcelona Supercomputing Center, Barcelona, Spain
  • D. Pastrana, Barcelona Supercomputing Center, Barcelona, Spain

Abstract: SUBMER company designs, builds and installs Liquid Immersion cooling solutions for HPC, hyperscaler, datacenters, Edge, AI, deep learning and blockchain applications. SUBMER Immersion Cooling is changing how datacenters are being built from the ground up, to be as efficient as possible and to have little or positive impact on the environment around them (reducing their footprint and reducing their consumption of precious resources such as water). SUBMER was founded in 2015 by Daniel Pope and Pol Valls, to make operating and constructing datacenters more sustainable. This SHAPE project aims to increase the knowledge and expertise of SUBMER employees in the realm of CFD simulations in an HPC environment. The project also aims to demonstrate tangible benefits in terms of improved design, efficiency and effectiveness of their immersion cooling solution. The vision of SUBMER is that being able to model a reasonably close estimation to the operational behavior of their systems is key to avoiding costly mistakes, provide a better product to their clientele and greatly reducing time to market. This white paper presents the methodology employed during the project, as well as some selected results.

Download paper: PDF

Title: Tooling4G – Minimize the Airflow Generated Noise on Automotive HVAC Systems


  • J. F. Caseiro, Technological Center for Mouldmaking Special Tooling and Plastic Industries (CENTIMFE), Marinha-Grande, Portugal
  • P. Alberto, Laboratory for Advanced Computing – University of Coimbra, Coimbra, Portugal

Abstract: The main goal of this work is to assess the gain that can be obtained when considering High-Performance Computing (HPC) for solving complex industrial problems. To that end, a Heat, Ventilation and Air Conditioning (HVAC) system of an automobile is analysed by means of an Aero-Acoustic Computational Fluid Dynamics (CFD) analysis. The performances of the HPC Navigator from the University of Coimbra and a research workstation are compared using Ansys® Fluent. Results show that using the Navigator with 64 cores distributed equally on two nodes reduces by a factor of 12.3 the execution time when compared with a workstation limited to 4 cores. The use of HPC resources is extremely useful for Small and Medium Enterprises (SMEs) that don’t have the required hardware to solve this kind of problems, enabling an increase in productivity and reducing costs associated with shorter project stages.

Download paper: PDF

Title: Deep Learning Based Topic Classification for Sensitivity Assignment to Personal Data


  • Apdullah Yayık, Mobildev Research and Development, Istanbul
  • Hasan Apik, Mobildev Research and Development, Istanbul
  • Ayse Tosun, Faculty of Computer and Informatics Engineering, Istanbul Technical University, Istanbul
  • Enver Ozdemir, National Center for High Performance Computing, Istanbul Technical University, Istanbul

Abstract: Knowing the topic of textual content before performing a natural language processing task enables the design of topic-specific pipelines. Since the topic is represented by all the sentences and words of the document, it can be accepted as a reference point that can describe the document alone. In this partnership, Mobildev successfully completed the construction and deployment of a topic classification model in order to assign a sensitivity level to the extracted personal data within the topic context. Since justice, health, and religion topics are considered as highly sensitive data by Personal Data Protection Rule, it is essential for the model to identify documents in these stated topics. Therefore, a publicly available dataset was chosen and, new document instances from these three important categories were added. Two state-of-the-art machine learning models for natural language processing tasks were assessed on the extended dataset: fasttext and bidirectional encoder representations transformers (BERT). The performance of the models and the computational costs for training at the server side are reported. After testing at the client side the most suitable model for a lightweight client operation is determined and deployed into the Mobildev’s existing platform. The model training and assessment have been successfully completed in collaboration with the National Center for High Performance Computing at Istanbul Technical University.

Download paper: PDF

Title: OPEN ENGINEERING (Belgium): “RF Solver on multi GPUs”


  • S. Costarelli, Open Engineering, Belgium
  • P. De Vincenzo, Open Engineering, Belgium
  • R. Dubois, IDRIS – Orsay, France
  • K. Hasnaoui, IDRIS – Orsay, France

Abstract: The Open Engineering is a high-tech SME which was created in 2001 and located in Liege (Belgium). The company is active in the Computer-Aided Engineering (CAE) market. They design, develop, and sell the OOFELIE::Multiphysics software. OOFELIE::Multiphysics is used to conceptualize, design, analyze, and optimize various types of systems before starting time-consuming and costly build-and-test cycles.

The goal of this SHAPE project is to develop in partnership with the IDRIS team, a multi GPUs version of the CUDA RF Solver in order to address problems that do not fit in the main memory of a single GPU; and to analyse its performance on an HPC multi-GPUs host.

Download paper: PDF

Title: SHAPE Project: Global Surface Intelligence Optimisation of national Vegetation Mapping Workflows


  • N. Banglawala, EPCC, The University of Edinburgh, UK
  • K. Stratford, EPCC, The University of Edinburgh, UK
  • M. Howie, Global Surface Intelligence Ltd., UK

Abstract: Global Surface Intelligence (GSI) uses earth observation to assess natural resources such as forestry, to monitor agricultural assets, and to provide business intelligence on land use and sustainability across the globe. Its operation, involving large volumes of satellite and other data, depends on efficient and effective software tools for machine learning and artificial intelligence workflows. The purpose of the work described here was twofold: first, to improve software effectiveness and software sustainability by making GSI’s standard workflows more robust and easier to use by non-experts and second, to improve parallel efficiency in a small number of specific data analysis workflows.

Download paper: PDF

Title: SHAPE Project NIER: Deep Learning for Video and Time Series Analysis


  • Laura Rocchi, NIER Ingegneria S.P.A
  • Cesare Sassoli, NIER Ingegneria S.P.A
  • Roberto Da Vià, CINECA
  • Eric Pascolo, CINECA

Abstract: In the present study we evaluated the possibility of automatically identifying the level of attention / inattention of a car driver, using deep learning techniques to classify the images recorded by a specific equipment able to monitor the driver’s gaze. An Eye Tracking system returned videos representing the scene seen by the drivers with a superimposed red cross corresponding to their gaze. The images were labelled by road safety experts in 5 classes, depending on the objects looked at (car, street, rear mirror, background, car interior). We developed a Convolutional Neural Network model (CNN) that was trained and tested on the images to detect the five classes. Specific video pre-processing was necessary. Analyses were performed with OpenCV software library, Keras and Python. The parallelisation of the code was investigated as a proof of concept (PoC) to evaluate a good performance of the computation when moving from this pilot study (that included only three videos) to the use “in production”. The code parallelisation was performed on the pre-processing and on the CNN code. In the first case both a multi-thread and a multi-task strategy were tested for the code parallelisation with the latter more performant in small sets of frames case. As for the CNN, we tested a multi-task implementation, using Horovod (4 GPU), and a multi-thread strategy using the mirrored strategy of Tensorflow (2 GPU). Even if a larger database is necessary for further evaluation, the Horovod parallelisation strategy is recommended at this stage. The CNN performance is promising, since, depending on the different combinations of training and testing set, accuracy was always larger than 70%, frequently over 80%. For the specific application this is a really appropriate results. Further studies are necessary to increase the dataset, to tune the models and to finalise the PoC, towards a ready-to-use application.

Download paper: PDF

Title: High resolution calculation for the optimisation of commercial LED spectra in health applications


  • Adrià Huguet-Ferran, Kumux, Enlighting Technologies SL, Baldiri Reixac 4, Parc Científic de Barcelona, Torre R, Planta 2, B1, 08028 Barcelona, Spain and Departament d’Enginyeries: Electrònica, Universitat de Barcelona, Facultat de Física, Martí i Franquès 1, 08028 Barcelona, Spain
  • Cristian Morales Pérez, Barcelona Supercomputing Center (BSC), Barcelona, Spain


The proposed project is related to the installation of bioadapted lamps at a hospital facility. The different partners involved in the project are: the scientific team in charge of the hospital itself, a manufacturer of light sources, a lamp control platform and Kumux – Enlighting Technologies (ET). In this context, ET will be responsible for performing the high-resolution calculations of the LED system in the lamp to be able to choose different light scenarios taking into account the visual, non-visual and power properties of light, such as luminous and radiant flux or the energy efficiency of the particular light scenario. The main goal of the project is, once the results of the simulation have been processed and replicated to the lamps, to be able to quantitatively demonstrate the incidence of bioadapted light in the recovery of patients, since there are preliminary studies that show very significant reductions in the hospitalisation period of the patient.

With the computer means available in Kumux, it is not possible to perform high-resolution calculations of the lamps used in the project. Our computational system only allows calculations at low resolution, so in this SHAPE project, we aim to simulate a lighting scenario at high resolution, which would take months of computation with our current code and computer means.

Download paper: PDF