Workshop for high-end HPC users for upcoming Exetrme-scale Industry Demonstrators

Size: px
Start display at page:

Download "Workshop for high-end HPC users for upcoming Exetrme-scale Industry Demonstrators"

Transcription

1 Workshop for high-end HPC users for upcoming Exetrme-scale Industry Demonstrators Dr.-Ing. Fouad el KHALDI Industry Strategy & Innovation General Manager ESI Group 1

2 Content Trends & Drivers Demonstrators driven by Industry Applications Hybrid Twin Crash & Safety Virtual Manufacturing : Additive Mfg. Overview HPC Challenges & Needs 2 Contexts: Open Sources Proprietary Applications 2

3 Moving Forward Immersive Virtual Engineering: Business Transformation Digital Transformation Disruptive Benefits Virtual Engineering Hybrid Twin : Holistic & interactive Virtual Prototyping Predictive Key Enablers : Exetrme-scale Modelling Multi Scales Multi Physics Co-Simulation HPC / HPDA Real Time Life Time etc Virtual Test Pro-active Better test Reactive Time 3

4 Portfolio: Multi-Prespectives Value Chain Virtual Manufacturing End-to-End Solutions Casting Composites Sheet Metal Forming Welding & Assembly Virtual Performance Solution Virtual Seat Solution Virtual Performance Crash & Safety NVH & Dynamics Interior Acoustics Statics & Strength Virtual Environment CFD & Multiphysics Electromagnetics Virtual Integration Platform Virtual Systems & Controls Virtual Reality Data Analytics Cybersecurity Cloud Solutions Vibro-Acoustics Data Mining Systems Modeling Mechatronic Models Engineering Services Combinational innovation Multi-Domain Simulation Decision Support Collaborative Decision-making Democratizing CAE Consulting Services Images courtesy of Fonderia Casati, AP&T, Škoda Auto, Volkswagen AG, EADS Casa Espacio and Herrenknecht. 4

5 Industry Digital Transformation: Virtual Prototyping Hybrid Twin Holistic view of the Value Chain (including Supply chain): Barrier (Safety Regulations / Region) Airbag & Dummies (Safety Regulations / Region) BIW Full integrated BIW : M-Material, M-Functions, Weld & Assembly,.. Full integrated Vehicle (BIW, Closers, Engine, transmission, Seat, ) Virtual Prototype Upfront Reliable Collaborative Decision Support for the Value Chain along Life Cycle Access to Details Design & Mfg. when it is needed Devil in Details Reliable Trade-offs (cause & effects) Data Design Test / Validation / Mfg Production / IoT : Operation, Services, Hybrid Twin Hybrid Twin to support Product Life Cycle Performance 5

6 6 High Scalability for Industrial Crash Models Challenge Ever growing model Complexity : virtual prototype needs Ever increasing model Size : approaching 10 millions of elements today, and reaching 20 millions within next 5 years. Ever reducing elapsed waiting Time for engineers : twice parameter runs in the day. Objective Reach below 5 hours for a full safety deviced car crash simulation (Airbag, dummies, stamping effect import etc) on cores. Scope Multi-Scale / Multi Physicss Scaling Mesh adaptivity (shell-shell; Shell-solid remeshing) FPM method scalability Dynamically adjustable load balance 6

7 ESI Scilab Applied Mathematics for Modeling, Simulation and Data analytics 7 Copyright ESI Copyright Group, ESI All Group, rights reserved All rights reserved.

8 Scilab First R&D Area Make simulation accessible Bring HPC compute power through Scilab/Scilab Cloud for European SMEs A large number of European SMEs would benefit from having access to large-scale HPC-powered simulation and HPDA-powered analytics. As of today however, HPC and HPDA services require a specific expertise which is rarely available in SMEs. The contemplated research and development will focus on building simulation and data analytics solution which are easy-to-use yet powerful through the use of HPC & HPDA compute cycles. The objective is to hide all the complexity of HPC and HPDA to SMEs while still providing them with state-of-the-art simulation precision. 8

9 Scilab Second R&D Area Industrial IoT Real-time operational Digital Twins powered by reduced-order models Digital twins are derived both from simulation models and from data-driven models Digital twins become extremely valuable in Industrial IoT use cases Model-order reduction and machine-learning techniques allow digital twin models to be used in operation as their simulation computation times are much lower than full-scale models. Model-order reduction and machine-learning techniques require data derived from computeintensive operations (both HPC and HPDA) The contemplated research and development will focus on how to select and perform regularly HPC-powered simulations and HPDA computation cycles to continue finetuning the reducedorder models used in operations. 9

10 Product Performance Lifecycle for Industrial IoT Hybrid Virtual Twins (physics-based and data-driven) Physics-based Virtual Twin Simulation Data Data-driven Virtual Twin Sensor Data Lake Instrumented Physical Asset Wrapping elements Update Auto-learning Math-based models System models Detailed 3D models Machine Learning & Reduced order models Apply Operational Predictive Models Act Mechanics Sensors Controllers Modeling & Simulation Software HPC Industrial Data Analytics Software HPDA Decision 10

11 HPC for Hybrid Twins 11

12 HPC for ROM & HT Biomedical Engineering 12

13 HPC for ROM in Industry

14 Additive Manufacturing HPC perspectives 14

15 Additive Manufacturing / Powder Bed Process Pre-scan process window Spreading feed stock Confirm + Defects? Melting Distortions & Residual Stresses Homogenization 15

16 HPC Challenges Thermal Model Processing of complete workpiece Coupling with residual stress modelling (activation models) Residual Stresses & Distortion Coupling with metallurgy Grain Growth models for complete workpiece 16 Development Scope & Target + Suggested Partners

17 HPC : Needs Spreading Massive MPI for DEM modelling Target: Model processing table of up to 1m length 100e6 particles Uni. Swansea: 2 PhDs Melting Maximize performance Target: 4 hatches in 1 day Multilayer: DEM-Melting: 1 layer/day 17 Development Scope & Target + Suggested Partners

18 18