PRACE Autumn School 2013 - Industry Oriented HPC Simulations

Sep 23, 2013, 8:30 AM → Sep 27, 2013, 6:45 PM CET

Lecture room IV/2 and computing rooms II/5, III/1, N17 (University of Ljubljana)

Resolving Industrial Engineering Applications with HPC by ANSYS, optiSLang, AVBP, elsA, RBF morph, etc.
In today's competitive product development, high performance computing (HPC) delivers outstanding value and investment return. Parallel computing increases understanding, productivity and accuracy of the simulation - a faster turnaround, reduced costs, systematic design variations and more complex models.
The PRACE Autumn School communicates and discusses issues and perspectives of HPC targeting industrial applications arising from evaluation performance and/or design of products, e.g. equipment and processes, with a particular emphasis on the automotive, aerospace, and energy fields. The PRACE Autumn School places priority on algorithms, simulation strategies, and programming techniques for complex fluid flow simulations, structural mechanics and electro-magnetics which usually have coupled multiphysics requiring intensive use of HPC resources. In case of expertise and software availability, there will be presentations of other issues, e.g. robustness and performance analysis, a brief introduction to pre- and post-processing techniques, e.g. CAD integration, mesh generation and visualization are presented.
The PRACE Autumn School targets at (40-50) experienced industrial users and some academics. Case studies and hands-on tutorials will be carried on the ULFME cluster. The tutorials will be held in parallel and in subsequent sessions, depending on the applicants' interest indicated in the registration questionnaire.

Due to large interest in elsA and AVBP by ONERA - Le centre français de recherche aérospatiale and CERFACS - European Centre for Research and Advanced Training in Scientific Computation we are considering to extend this tutorials (from 3h30' to 5 hours hands-on) by including small turbine case. This means some time changes in Friday's session with little possibilities for sequential with RBF-Morph tutorials.

  Short URL: http://events.prace-ri.eu/event/as13

  

Brochure PRACE-Autumn-School-2013-brochure.pdf

Certificate of participation PRACE-Autumn-School-2013-certificate-of-participation.pdf

Instructions Setting_up_license_and_remote_desktop_access_to_HPCFS.pdf

Poster poster.pdf

Seats arrangements for computer classrooms http://hpc.fs.uni-lj.si/project/as13

Survey http://survey.ipb.ac.rs/index.php?sid=19497&lang=en

Timetable Short timetable flyer

Monday, September 23

8:30 AM → 1:00 PM Registration, Coffee & Poster hanging

Attendees are required to show up and register themselves at the registration desk before the school starts in the afternoon and hang their poster before the introductory poster session at noon.

9:00 AM → 11:30 AM HPC and parallel programming: Optional training

This pre-training is optional and will be held if enough participants will register for it. It covers HPC architecture, schedulers and OpenMP and MPI hands on training.

9:00 AM HPC architectures and MPI, OpenMP programming hands-on

HPC architecture overview with examples of Parallel Programming with - distributed MPI, - shared OpenMP, and - hybrid MPI+OpenMP, memory models.

Speaker: Dr Leon Kos (University of Ljubljana)

Introduction to HPC video with slides http://prace.viidea.net/introduction_to_hpc/ http://videolectures.net/prace2013_kos_hpc_introduction/

Reference

• MPI C quick reference card
• MPI functions
• OpenMP C quick reference card
• OpenMP Fortran quick reference

Slides

• Basic usage of HPCFS
• Introduction to HPC
• Introduction to MPI
• Introduction to OpenMP

Using HPCFS programming-hands-on video with slides http://prace.viidea.net/using_hpcfs_programming_hands_on/ http://videolectures.net/prace2013_kos_using_hpcfs/

12:00 PM → 1:00 PM Posters

Attendees communicate their work and meet each other by presenting (past) posters.

1:00 PM → 1:20 PM Opening and Welcome address

Welcome from University of Ljubljana as prace PRACE coordinator for Slovenia. Industry challenges tackled with HPC and research targeted for products with simulations and optimizations that give concrete answers to many open questions.

Speaker: Prof. Jožef Duhovnik (University of Ljubljana, PRACE coordinator for Slovenia)

Slides Welcome address describing HPC and PRACE systems and quick intro to HPCFS

Video with slides http://prace.viidea.net/introduction_to_hpc_and_prace/ http://videolectures.net/prace2013_duhovnik_hpc_prace_introduction/

1:20 PM → 2:00 PM HPC – the Perspective of a CFD Practitioner

The lecture will address the trends of the relation between Computational Fluid Dynamics (CFD), which is a subsection of Simulation-Based Engineering Science, and High Performance Computing (HPC) applications. See attached extended abstract.

Speaker: Prof. Antonio C.M. Sousa (University of Aveiro and New Brunswick)

Paper Introductory_Lecture.pdf

Slides HPC - the Perspective of a CFD Practitioner

Video with slides http://prace.viidea.net/hpc_the_perspective_of_a_cfd_practitioner/ http://videolectures.net/prace2013_sousa_cfd_practitioner/

2:00 PM → 3:00 PM Discover your design quicker as before with HPC

An overview over current technology, recent benchmark results from a user perspective, as well as new ways of calculating large models and extensive design variations will be presented. - High Performance Computing (HPC). approaches overview - Success stories from structural mechanics and electromagnetics - Overview of the solvers across all physics - Ongoing HPC initiatives

Speaker: Alexander Dopf (CADFEM)

Slides PRACE-Autumn-School-2013-Dopf-CADFEM-Discover.pdf

Video with slides http://prace.viidea.net/discover_your_design_quicker_as_before_with_hpc/ http://videolectures.net/prace2013_dopf_discover_design/

3:00 PM → 7:00 PM Structural Mechanics: Best practices for efficient HPC performance with large models

3:00 PM Hands On Training with HPC

Prerequisites: MPI – HP or Intel How to set up HPC: solver handler, number of CPUs, and DMP Handling of large models: DM Prep in v14.5; mechanical filters, tags and named selections

Speaker: Dr Bernhard Hössl (CADFEM)

Slides Best practices for efficient HPC performance with large models

Video with slides http://prace.viidea.net/best_practices_for_efficient_hpc_performance/ http://videolectures.net/prace2013_hoessl_hpc_large_models/

4:30 PM Break

4:45 PM Hands On Training with HPC (cont.)

Mesh controls for large models: curvature and proximity How to cut analysis time: Proper contact and analysis settings Evaluation of results: results tracker; post command snippets; max tag adjust to visible

Speaker: Dr Bernhard Hössl (CADFEM)

Tuesday, September 24

9:00 AM → 5:00 PM Structural Mechanics: Robust Design Optimization

9:00 AM Robust Design Optimization – from the idea to the optimized product

Introduction Motivation for parametric variations Parametric workflows Introduction into optiSLang Systematic variation using optiSLang a) Sensitivity Analysis and Metamodeling b) Tutorial: Sensitivity analysis of a notch

Speaker: Andreas Veiz (CADFEM)

Video with slides http://prace.viidea.net/understand_your_design/ http://videolectures.net/prace2013_veiz_robust_design_optimization/

11:30 AM Lunch break

1:00 PM Sensitivity analysis of a notch (cont.)

Getting started in design improvement: a) How to improve a design b) Tutorial: Improve the notch Typical Questions in postprocessing results Outlook: Efficient performance of extensive design variation

Speaker: Andreas Veiz (CADFEM)

Sample notch.wbpz

Slides

• 01_Understand_your_Design_Agenda.pdf
• 02_Understand_you_Dresign_Motivation.pdf
• 03_Understand_your_Design_parametric_workflow.pdf
• 04_Understand_your_Design_Manual_variation.pdf
• 05_Understand_your_Design_Systematic_variation_using_optiSLang_inside_Workbench.pdf
• 06_Understand_your_Design_Typical_Questions.pdf
• 07_Understand_your_Design_Performance_in_Computing.pdf
• 08_Understand_your_Design_Theory_in_Optimization.pdf
• 09_Understand_your_Design_Outlook_Seminars.pdf

Video with slides http://prace.viidea.net/sensitivity_analysis_of_a_notch/ http://videolectures.net/prace2013_veiz_sensitivity_analysis/

Wednesday, September 25

9:00 AM → 5:00 PM Multiphysics: Electromechanical and Mechatronic Systems: Hands on with EM

Maxwell and HFSS training

9:00 AM EM Simulation Overview (tools and numerical methods)

Speakers: Dr Christian RöMELSBERGER (CADFEM), Jens Otto (CADFEM)

Video with slides http://prace.viidea.net/em_simulation_overview/ http://videolectures.net/prace2013_roemelsberger_electromagnetic_simulation/

10:00 AM Coffe break

Attendees reposition into Computing room for EM tutorials or continue with CFD lectures until lunch.

10:15 AM EM Example 1: Position Sensor Design (Hall based) with Maxwell 3D

Speaker: Jens Otto (CADFEM)

Geometry SENSOR_GEOM.x_t

Project file sensor_1.wbpz

Slides HANDS-ON-Position-Sensor-Introduction-to-ANSYS_Maxwell_-_PRACE_2013.pdf Position-Sensor-Introduction-to-ANSYS_Maxwell_-_PRACE_2013.pdf

Video with slides http://prace.viidea.net/em_example_1_position_sensor_design_hall_based_wit/ http://videolectures.net/prace2013_otto_position_sensor_simulation/

11:30 AM Lunch break

1:00 PM EM Example 2: WLAN-Antenna Optimization with HFSS

Problem Description Analysis Setup and Basic Simulation Results Design Optimization

Speaker: Dr Christian Römelsberger (CADFEM)

HFSS projects

• Dual_Band_Slot_4.wbpz
• Dual_Band_Slot_Antenna.hfss
• Dual_Band_Slot_For_Optimization.hfss
• Dual_Band_Slot_Full.hfss

Paper Theorie.pdf

Slides Optimierung_Dual_Band_Slot_Antenna.pdf WLAN_Workshop.pdf

Video with slides http://prace.viidea.net/em_example_2_wlanantenna_optimization_with_hfss/ http://videolectures.net/prace2013_roemelsberger_optimization_antenna/

2:30 PM Break

2:50 PM EM Example 3: Starter-Generator Design with Maxwell 3D/ Mechanical

Problem Description Analysis Setup and Basic Simulation Results Efficency Calculation and Loss Assessment Temperature Evaluation

Speaker: Jens Otto (CADFEM)

Geometry cpg_geom_PRACE_no_ISURF.x_t

Project file

• claw2.sph
• claw_workon_geom.wbpz
• cpg.amcp

Slides Vibroacoustic_Simulation_of_Claw_Pole_Generator_-_PRACE_2013.pdf

Video with slides http://prace.viidea.net/em_example_3_vibroacoustic_simulation_of_claw_pole/ http://videolectures.net/prace2013_otto_vibroacoustic_simulation/

4:30 PM EM: Summary and discussion

Speakers: Dr Christian Römelsberger (CADFEM), Jens Otto (CADFEM)

10:15 AM → 5:00 PM Computational Fluid Dynamics

10:15 AM Sailing Yachts CFD

Case-history on the balance between computational costs & insight benefits obtained by using advanced CFD modeling (Delayed Detached Eddy Simulations vs Reynolds Averaged NS ) in sailing yachts aerodynamics

Speaker: Dr Raffaele Ponzini (CINECA)

Slides Sailing yacht CFD

Video with slides http://prace.viidea.net/delayed_detached_eddy_simulation_of_yacht_sails/ http://videolectures.net/prace2013_ponzini_yacht_sails/

11:00 AM BIO-CFD Case-history: Introduction on CFD in hemodynamics

Wide range of haemodynamics CFD application is presented (medical implantable devices, blood filters, multi-scale-models,non-Newtonian rheology,...) to perform in silico health-care research and medical device design. The main expected benefits of the lecture+tutorial for attendants are: - enlarge their knowledge on state of art methods and algorithms; - apply best practices on state-of-the-art software deployment; - experience technical tips-and-tricks throughout the different phases of the tutorial; - learn from case history and practical applications told by CFD expert.

Speaker: Dr Raffaele Ponzini (CINECA)

Slides BIO CFD case-history

Video with slides http://prace.viidea.net/biocfd_tutorial_1/ http://videolectures.net/prace2013_ponzini_bio_cfd/

12:00 PM Lunch break

1:30 PM BIO-CFD Tutorial

a) Basic modeling using Fluent for carotid bifurcation (GUI in fluent for case setup from mesh import to first steady state, Newtonian simulation running; TUI and case journaling in Fluent for batch execution; post-processing in Fluent) b) Intermediate modeling using Fluent for carotid bifurcation haemodynamics (unsteady BC in Fluent; non-Newtonian custom models; multi-scale models coupling 3D and 0D in Fluent; customized post-processing)

Speaker: Dr Raffaele Ponzini (CINECA)

Slides BIO CFD Tutorial 1 BIO CFD Tutorial 2

Video with slides http://prace.viidea.net/biocfd_tutorial_2/ http://videolectures.net/prace2013_ponzini_bio_cfd/

Thursday, September 26

9:00 AM → 10:15 AM Examples of Solving Industrial Cases with Fluent

Several industrial cases will be presented, executed by SimTec or its customers with the aid of ANSYS Fluent general-purpose 3D CFD package. The examples are taken from industrial applications in SE Europe countries but emphasis will be given on the Slovenian cases. The applications cover the industry sectors of marine, constructions (HVAC), chemical, pharmaceutical, hydraulic, metal, power generation, amongst others. Each case presentation includes: (a) modelling targets, (b) description of the physical model, (c) results, (d) modelling conclusions.

Speaker: Dr Dimitrios Sofialidis (SimTec)

Slides Simulation_Stories_in_SE-Europe.pdf

Video with slides http://prace.viidea.net/examples_of_solving_industrial_cases_with_fluent/ http://videolectures.net/prace2013_sofialidis_simulation_projects/

10:15 AM → 10:30 AM Coffee break

10:30 AM → 3:00 PM Computational Fluid Dynamics: Express Introductory Training in Fluent

10:30 AM Introduction to CFD

Speaker: Dr Dimitrios Sofialidis (SimTec)

Slides Fluent-Intro_14.5_L01_IntroCFD.pdf

Video with slides http://prace.viidea.net/introduction_to_cfd_metodology/ http://videolectures.net/prace2013_sofialidis_cfd_methodology/

11:15 AM Fluent Workshop 1: Fluid Flow & Heat Transfer in a Mixing Tee

Speaker: Dr Dimitrios Sofialidis (SimTec)

Slides Fluent-Intro_14.5_WS01_Mixing_Tee.pdf

12:30 PM Lunch break

1:30 PM Boundary Conditions and Solver Settings (Convergence & Accuracy)

The presentation will cover and explain the basic steps of a CFD analysis. Important topics such as Boundary Conditions, Convergence of the solution and increasing the Accuracy of the results will be discussed in depth.

Speaker: Dr Dimitrios Sofialidis (SimTec)

Slides L02_Boundary Conditions & Solver Settings

2:00 PM Turbulence Modelling, Heat Transfer & Transient Calculations

As most flow problems exhibit turbulent nature and most engineering applications involve heat transfer, these two subjects will be covered. In addition, as many fluid flows are not steady-state phenomena, transient simulations and solution and relative issues like time step size will be discussed.

Speaker: Dr Dimitrios Sofialidis (SimTec)

Slides Turbulence & HeatTransfer & Transient

2:30 PM Second Fluent workshop (hands on)

Each participant will choose 1 among 3 alternative workshops. These workshops will be formed after the list of participants is closed, in order to cover as much as possible application fields declared by the participants.

Speaker: Dr Dimitrios Sofialidis (SimTec)

Printouts poll Printouts poll

Slides

• WS02_Discrete_Phase
• WS03_Multi_Species_Postprocessing
• WS04_Airfoil
• WS05_Electronics_Cooling
• WS06_Sliding_Mesh
• WS07_Tank_Flush
• WS08_Vortex_Shedding

3:00 PM → 6:00 PM High-Performance Computing with Fluent (hands-on)

An industrial CFD case will be executed as a workshop. The setup, solution and postprocessing will span through the first half of the session. The second half will be spent with experimentation of alternative solutions (e.g. changing BCs, material properties, mesh size, etc.) and discussion regarding best practices, error reduction and accuracy, tips and tricks, parallel computing efficiency, etc.

Speaker: Dr Dimitrios Sofialidis (SimTec)

Slides HPC_with-ANSYS-Fluent.pdf

Video with slides http://prace.viidea.net/highperformance_computing_with_fluent/ http://videolectures.net/prace2013_sofialidis_hpc_ansys_fluent/

Friday, September 27

9:00 AM → 6:45 PM Computational Fluid Dynamics: Custom CFD codes and plugins

9:00 AM High Performance Computing of gas turbine flows: current and future trends

Nicolas Gourdain presents elsA for aerodynamics used by many by industrial partners such as Airbus, SAFRAN, etc. and AVBP for combustion, used by SAFRAN, IFP, etc. The use of Computational Fluid Dynamics (CFD) is mandatory today both for scientific investigation of complex flows and industrial design of gas turbines and aircraft. Further improvement and optimization of such systems currently present one of the most formidable challenges in modern engineering research due to the very stringent requirements for efficiency, pollutant emissions, reliability and safety. Efficient numerical tools coupled with high performance computers, have become a key element of the design process in the fields of energy supply and transportation. However flow phenomena that occur in complex systems such as gas turbines and aircrafts are still not understood mainly because of the models that are needed. In fact, most CFD predictions as found today in industry focus on a reduced or simplified version of the real system (such as a periodic sector) and are usually solved with a steady-state assumption. This course discusses how to overcome such barriers and how this challenges can be addressed by developing flow solvers running on high-end computing platforms, using concurrently thousands of computing cores. Parallel strategies used by modern flow solvers are presented with particular emphases on mesh-partitioning, load balancing and communication.

Speaker: Dr Nicolas Gourdain (CERFACS)

Slides HPC for gas turbines by CERFACS and ONERA

Video with slides http://prace.viidea.net/high_performance_computing_of_gas_turbine_flows_cu/ http://videolectures.net/prace2013_gourdain_hpc_gas_turbine_flows/

10:00 AM RBF morph theory and applications case-history

An overview about mesh morphing and its benefits will be given with a specific focus on Radial Basis Functions (RBF) methods and the industrial tool RBF Morph, currently available as an add-on for the CFD solver ANSYS Fluent. Standard applications of mesh morphing will be first explained; the tool is here used as an effective way to makes the original CFD model parametric with respect to the shape. Details on how such parametric models can be introduced in typical industrial workflow (shape or set-up optimization) will be given including CAD connection (STL targets and back to CAD) and flow sculpting (using adjoint solver of Fluent). Advanced applications will be covered as well including: fluid structure interaction using 2-ways FSI and modal superposition, ice/snow accretion modeling, transient morphing with desired time histories.

Speaker: Prof. Marco Evangelos Biancolini ("Tor Vergata" University of Rome)

Slides

• RBF morph
• RBF Morph Advanced Mesh Morphing for optimization and multi-physics
• RBF morph training agenda

Video with slides http://prace.viidea.net/rbf_morph_theory_and_applications_casehistory/ http://videolectures.net/prace2013_biancolini_rbf_morph/

10:45 AM Coffe break

11:00 AM RBF morph hands-on

The first session will cover basic exercises to set-up shape modifications on simple geometries (a cube immersed in a wind tunnel and a straight pipe). A step by step set-up will be demonstrated to gain confidence with the GUI of RBF Morph (set-up stage). An advanced session addressed to industrial meshes and complex set-up will follow. Pre computed set-up will be available so that the students can deepen their knowledge about advanced feature of RBF Morph for check trouble shooting the set-up (preview, mesh quality).

Speaker: Prof. Marco Evangelos BIANCOLINI ("Tor Vergata" University of Rome)

Slides RBFmorph hands-on

Video with slides http://prace.viidea.net/rbf_morph_handson/ http://videolectures.net/prace2013_biancolini_rbf_morph_hands_on/

12:30 PM Lunch break

1:30 PM Multi-block structured code (elsA) and an unstructured code (AVBP) hands-on

The aim will be to test some of the requirements presented in the course (such as mesh-partitioning and load balancing) to achieve high-scalability on massively parallel computers (maybe also on GPUs). Two examples are used to illustrate these concepts: a multi-block structured code (elsA) and an unstructured code (AVBP). Parallel computing strategies used with both flow solvers are detailed and compared. This comparison indicates that mesh-partitioning and load balancing are more straightforward with unstructured grids than with multi-block structured meshes. However, the mesh-partitioning stage can be challenging for unstructured grids, mainly due to memory limitations of the newly developed massively parallel architectures. Finally, detailed investigations show that the impact of mesh-partitioning on the numerical CFD solutions, due to rounding errors and block splitting, may be of importance and should be accurately addressed before qualifying massively parallel CFD tools for a routine industrial use.

Speakers: Dr Michel Gazaix (ONERA), Dr Nicolas Gourdain (CERFACS)

Slides HPC with elsA

Video with slides http://prace.viidea.net/multiblock_structured_code_elsa_handson/ http://videolectures.net/prace2013_gourdain_hpc_elsa_hands_on/