1 / 14

Kiran Narayanan a i n collaboration with Angel Mora a , Nicholas Allsopp a,b , Tamer El Sayed a

Optimization of the impact performance of a metal/ polyurea composite plate via coupling of a genetic algorithm and a finite element code. Kiran Narayanan a i n collaboration with Angel Mora a , Nicholas Allsopp a,b , Tamer El Sayed a. b Cray Computing Deutschland GmbH

ivan
Download Presentation

Kiran Narayanan a i n collaboration with Angel Mora a , Nicholas Allsopp a,b , Tamer El Sayed a

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Optimization of the impact performance of a metal/polyurea composite plate via coupling of a genetic algorithm and a finite element code KiranNarayanana in collaboration with Angel Moraa, Nicholas Allsoppa,b, Tamer El Sayeda b Cray Computing Deutschland GmbH University of Stuttgart, HLRS Nobelstrasse 19 D-70569 Stuttgart, Germany a Physical Sciences and Engineering Division King Abdullah University of Science and Technology (KAUST) Building 4, 4700 KAUST Thuwal 23955-6900, KSA

  2. High velocity impact : Scenarios Composite materials are used for Micrometeoroid Orbital Debris (MOD) ballistic-shielding Extra-vehicular activities in space Space shuttle window pit from orbital debris impact Low High International Space Station: Impact Risk Picture credits: NASA K Narayanan, ECCOMAS 2012

  3. Problem specification Composite Plate Vzero Velocity Impactor Vz-0 Large No. of Function Evaluations Polyurea Steel (HSS) X Z Y Finite Element model adapted from El Sayed et al (2009) K Narayanan, ECCOMAS 2012

  4. 16 Racks, 32x32x16 Cabled 8x8x16 Rack Blue Gene/P 32 Node Cards 222 TFlop/s 64TB Node Card (32 chips 4x4x2) 32 compute, 0-2 IO cards 13.9 TFlop/s 4 TB Compute Card 1 chip, 20 DRAMs 435 GF/s 128GB Chip 4 processors 13.6 GF/s 4.0 GB DDR2 13.6 GF/s 8 MB EDRAM Computational Resources Master: Single x86 node on a cluster Workers: BG/P With permission from KAUST Supercomputing Laboratory K Narayanan, ECCOMAS 2012

  5. Initialization START Function evaluation Write parameter cycle file Crossover Mutation n=n+1 Function evaluation Fitness assessment Write parameter cycle file Replacement Convergence NO Write results file STOP Parallelization Strategy Single Objective Genetic Algorithm (SOGA) iterator from DAKOTAa is used to compute optimal value of velocity V0 that is used as input to the FE simulation. The fitness function for the GA is |Vz-final|. aAdams et al (2009) YES Flowchart - GA K Narayanan, ECCOMAS 2012

  6. GA parameter file Computational Steering(GA/FE Coupling) Create FE input Return value to GA Submit to LL on BG/P Submission successful Set velocity to 5.555e-6 m/s NO YES Return value to GA Simulation complete NO Set velocity to 6.666e-6 m/s YES NO K Narayanan, ECCOMAS 2012 Clean termination Wall clock limit NO YES YES Write results Write results

  7. Results K Narayanan, ECCOMAS 2012

  8. Strong Scaling of FE code Simulation run time T(p’) [secs] Simulation times of the dynamic problem reached at wall clock limit of 24 hrs No. of processors (p’) Computational time to reach a predetermined dynamic simulation time Optimal number of MPI-enabled jobs for each objective function evaluation was determined to be 128 K Narayanan, ECCOMAS 2012

  9. Relative Efficiency of hybrid parallelization b Eldred and Hart (1998) Number of processors (p) vs Relative Efficiency E(p) Concurrency=5 Concurrency=16 Concurrency=32 K Narayanan, ECCOMAS 2012

  10. Optimal impact velocity of projectile Max. Concurrency=16, Total Evaluations=32 Max. Concurrency=5, Total Evaluations=25 K Narayanan, ECCOMAS 2012

  11. Performance of Genetic Algorithm Impact velocities closest to optimal value Computational time to calculate optimal impact velocity K Narayanan, ECCOMAS 2012

  12. Conclusions • GA/FE coupling was used to facilitate guided computation of optimal impact velocity with reduced computational time • The methodology of coupling and selection of parameters is applicable to simulation-based optimizations in general and is scalable K Narayanan, ECCOMAS 2012

  13. Acknowledgements • For computer time, this research used the resources of the Supercomputing Laboratory at King Abdullah University of Science & Technology (KAUST) in Thuwal, Saudi Arabia • This work was fully funded by the KAUST baseline fund K Narayanan, ECCOMAS 2012

  14. References Narayanan K, Mora A, Allsopp N, El Sayed T, A hybrid massively parallel implementation of a genetic algorithm for optimization of the impact performance of a metal/polymer composite plate, International Journal of High Performance Computing Applications, Published online before print July 17, 2012, DOI: 10.1177/1094342012451474 El Sayed T, Willis M Jr., Mota A, Fraternalli F, Ortiz M, Computational Assessment of ballistic impact on a high strength structural steel/polyurea composite plate, Computational Mechanics, 43, 525-534 (2009) Adams BM, Bohnhoff WJ, Dalbey KR, Eddy JP, Eldred MS, Gay DM, Haskell K, Hough PD, Swiler LP, DAKOTA, A Multilevel Parallel Object-Oriented Framework for Design Optimization, Parameter Estimation, Uncertainty Quantification, and Sensitivity Analysis: Version 5.0 User's Manual, Sandia Technical Report SAND2010-2183 (2009) Eldred MS, Hart WE, Design and implementation of multilevel parallel optimization on the Intel teraflops, Paper AIAA-98-4707 in Proceedings of the 7th AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and Optimization, 44-54 (1998) K Narayanan, ECCOMAS 2012

More Related