research-article

Machine learned domain decomposition scheme applied to parallel multi-scale muscle simulation

Authors: Miloš Ivanović, Ana Kaplarević-Mališić, Boban Stojanović, Marina Svičević, Srboljub M MijailovichAuthors Info & Claims

The International Journal of High Performance Computing Applications, Volume 33, Issue 5

Pages 885 - 896

https://doi.org/10.1177/1094342019833151

Published: 01 September 2019 Publication History

Abstract

Since multi-scale models of muscles rely on the integration of physical and biochemical properties across multiple length and time scales, they are highly processor and memory intensive. Consequently, their practical implementation and usage in real-world applications is limited by high computational requirements. There are various reported solutions to the problem of parallel computation of various multi-scale models, but due to their inherent complexity, load balancing remains a challenging task. In this article, we present a novel load balancing method for multi-scale simulations based on finite element (FE) method. The method employs a computationally simple single-scale model and machine learning in order to predict computational weights of the integration points within a complex multi-scale model. Employing the obtained weights, it is possible to improve the domain decomposition prior to the complex multi-scale simulation run and consequently reduce computation time. The method is applied to a two-scale muscle model, where the FE on macroscale is coupled with Huxley’s model of cross-bridge kinetics on the microscale. Our massive parallel solution is based on the static domain decomposition policy and operates in a heterogeneous (central processing units + graphics processing units) environment. The approach has been verified on a real-world example of the human tongue, showing high utilization of all processors and ensuring high scalability, owing to the proposed load balancing scheme. The performance analysis shows that the inclusion of the prediction of the computational weights reduces execution time by about 40% compared to the run which uses a trivial load balancer which assumes identical computational weights of all micro-models. The proposed domain decomposition approach possesses a high capability to be applied in a variety of multi-scale models based on the FE method.

References

[1]

Bathe KJ (1982) Finite Element Procedures in Engineering Analysis. Upper Saddle River: Prentice-Hall.

Google Scholar

[2]

Bathe K and Kojic M (2005) Inelastic Analysis of Solids and Structures. Berlin: Springer.

Google Scholar

[3]

Bernst I, Kaufmann C, and Frochte J (2015) Learning load balancing for simulation in heterogeneous systems. In: Proceedings of the 12th international conference applied computing (ed Weghorn H), Maynooth, Greater Dublin, Ireland, 24–26 October 2015, pp. 121–128. IADIS.

Google Scholar

[4]

Brady TF and Yellig E (2005) Simulation data mining: a new form of computer simulation output. In: Proceedings of the 37th conference on Winter simulation. Winter simulation conference, Orlando, Florida, 4–7 December 2005, pp. 285–289. Winter Simulation Conference ©2005.

Google Scholar

[5]

Burrows S, Frochte J, and Völske M, et al. (2013) Learning overlap optimization for domain decomposition methods. In: Pacific-Asia conference on knowledge discovery and data mining (eds Pei J, Vincent S, Tseng, Cao L, Motoda H, and Guandong X), Gold Coast, Australia, 14–17 April 2013, pp. 438–449. New York, NY: Springer.

Google Scholar

[6]

Burrows S, Stein B, and Frochte J, et al. (2011) Simulation data mining for supporting bridge design. In: Proceedings of the Ninth Australasian data mining conference-Volume 121 (eds Vamplew P, Stranieri A, Ong K-L, Christen P, and Kennedy PJ), Ballarat, Australia, 1–2 December 2011, pp. 163–170. Darlinghurst, Australia: Australian Computer Society, Inc.

Digital Library

Google Scholar

[7]

Chopard B, Falcone JL, and Hoekstra AG, et al. (2011) A framework for multiscale and multiscience modeling and numerical simulations. In: International Conference on Unconventional Computation (eds Calude CS, Kari J, Petre I, and Rozenberg G), Turku, Finland, 6–11 June, 2011, pp. 2–8. Berlin, Heidelberg: Springer.

Google Scholar

[8]

Gordon A, Huxley AF, and Julian F (1966) The variation in isometric tension with sarcomere length in vertebrate muscle fibres. The Journal of Physiology 184(1): 170.

Google Scholar

[9]

Heidlauf T and Röhrle O (2013) Modeling the chemoelectromechanical behavior of skeletal muscle using the parallel open-source software library OpenCMISS. Computational and Mathematical Methods in Medicine 2013: 517287.

Google Scholar

[10]

Hill A (1938) The heat of shortening and the dynamic constants of muscle. Proceedings of the Royal Society of London B: Biological Sciences 126(843): 136–195.

Crossref

Google Scholar

[11]

Huxley AF (1957) Muscle structure and theories of contraction. Progress in Biophysics and Biophysical Chemistry 7: 255–318.

Crossref

Google Scholar

[12]

Ivanović M, Stojanović B, and Kaplarević-Mališić A, et al. (2016) Distributed multi-scale muscle simulation in a hybrid MPI–CUDA computational environment. Simulation 92(1): 19–31.

Google Scholar

[13]

Karabasov S, Nerukh D, and Hoekstra A, et al. (2014) Multiscale modelling: approaches and challenges. Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering Sciences 372(2021): 20130390.

Google Scholar

[14]

Kojic M, Mijailovic S, and Zdravkovic N (1998) Modelling of muscle behaviour by the finite element method using hill’s three-element model. International Journal for Numerical Methods in Engineering 43(5): 941–953.

Crossref

Google Scholar

[15]

Kuhn M (2008) Building Predictive Models in R Using the caret Package. Journal of Statistical Software 28(5): 1–26.

Crossref

Google Scholar

[16]

Lister M (1960) The numerical solution of hyperbolic partial differential equations by the method of characteristics. Mathematical Methods for Digital Computers 1: 165–179.

Google Scholar

[17]

McMahon TA (1984) Muscles, Reflexes, and Locomotion. Princeton: Princeton University Press.

Google Scholar

[18]

Mei L and Thole CA (2008) Data analysis for parallel car-crash simulation results and model optimization. Simulation Modelling Practice and Theory 16(3): 329–337.

Crossref

Google Scholar

[19]

Mijailovich SM, Fredberg JJ, and Butler JP (1996) On the theory of muscle contraction: filament extensibility and the development of isometric force and stiffness. Biophysical Journal 71(3): 1475–1484.

Google Scholar

[20]

Razumova MV, Bukatina AE, and Campbell KB (1999) Stiffness-distortion sarcomere model for muscle simulation. Journal of Applied Physiology 87(5): 1861–1876.

Google Scholar

[21]

Stein B and Curatolo D (1998) Selection of numerical methods in specific simulation applications. In: International conference on industrial, engineering and other applications of applied intelligent systems (eds del Pobil AP, Mira J, and Ali M), Castellón, Spain, 1–4 June 1998, pp. 918–927. Berlin, Heidelberg: Springer.

Google Scholar

[22]

Stojanovic B, Kojic M, and Rosic M, et al. (2007) An extension of hill’s three-component model to include different fibre types in finite element modelling of muscle. International Journal for Numerical Methods in Engineering 71(7): 801–817.

Google Scholar

[23]

Torelli A (1997) Study of a mathematical model for muscle contraction with deformable elements. Rendiconti del Seminario Matematico della Università di Padova (Torino) 55(3): 241–271.

Google Scholar

[24]

Zajac FE (1989) Muscle and tendon properties models scaling and application to biomechanics and motor. Critical Reviews in Biomedical Engineering 17(4): 359–411.

Google Scholar

Author biographies

Miloš Ivanović received a PhD in computer science from the Faculty of Science, University of Kragujevac, Serbia, in 2010. He is an associate professor in the Department of Mathematics and Informatics at the Faculty of Science, University of Kragujevac. His main research interests include numerical modeling using mesh-free methods, soft-computing, the application of shared and distributed memory parallelism, GPU computing, and grid and cloud computing. He has participated in several national-, European-, and US-funded projects; EU-funded research projects include FP7-224297 ARTreat and ongoing H2020-777204 SilicoFCM. He authored over 15 publications in international journals and numerous conference publications.

Ana Kaplarevic-Mališić received a PhD in informatics from the Faculty of Science, University of Kragujevac, Serbia, in 2016. She is an assistant professor at the Department of Mathematics and Informatics, Faculty of Science, University of Kragujevac, Serbia. Her main research interests include HPC, the application of shared and distributed memory parallelism, and mathematical modeling and simulation. She has participated in several national and European funded projects. She authored a number of publications in peer-reviewed journals and conference proceedings.

Boban Stojanović received a PhD in technical sciences from the University of Kragujevac, Serbia, in 2007. He is an associate professor at the Department of Mathematics and Informatics, Faculty of Science, University of Kragujevac, Serbia. His research interests are in the area of computer modeling and numerical simulations, especially in the field of bioengineering. He is author and coauthor of few monographs, a number of publications in peer-reviewed journals, and several simulation software. He has participated in several international scientific projects funded by European Commission, US NIH, and so on.

Marina Svičević received an MSc degree in mathematics and informatics from the Faculty of Science, University of Kragujevac, Serbia, in 2009. She is PhD student and teaching and research assistant at the Department of Mathematics and Informatics, Faculty of Science, University of Kragujevac, Serbia. Her main research interests include development of numerical methods and software for simulation of biomechanical muscle behavior on multiple scales. She has participated in several national and international projects with partner institutions in the United States and the United Kingdom. She has coauthored publications in international journals and numerous conference publications.

Srboljub M Mijailovich received his PhD in mechanical engineering from the Massachusetts Institute of Technology (MIT), Cambridge, Massachusetts, in 1991. He is a research professor for computational mechanobiology at the Department of Chemistry and Chemical Biology, Northeastern University. His current research interests are in developing quantitative approaches to study biological systems at multiple levels of organization (i.e. multi-scale modeling). In particular, he is leading the development of a theoretical framework that will advance our understanding of how cellular and subcellular phenomena integrate to dynamic behavior of physiological systems, based on the kinetics of underlying molecular processes. These theoretical advances are the foundation for the development of computational platforms to study the interplay between mechanical forces, cell biology, and integrated organ physiology.

Cited By

View all

Milićević BIvanović MStojanović BMilošević MKojić MFilipović N(2022)Huxley muscle model surrogates for high-speed multi-scale simulations of cardiac contractionComputers in Biology and Medicine10.1016/j.compbiomed.2022.105963149:COnline publication date: 14-Dec-2022
https://dl.acm.org/doi/10.1016/j.compbiomed.2022.105963

Index Terms

Machine learned domain decomposition scheme applied to parallel multi-scale muscle simulation

Index terms have been assigned to the content through auto-classification.

Recommendations

Computational homogenization with million-way parallelism using domain decomposition methods
Abstract
Parallel computational homogenization using the well-knwon ${FE}^{2}$ approach is described and combined with domain decomposition and algebraic multigrid solvers. It is the purpose of this paper to show that and how the ${FE}^{2}$ method can take advantage of ... $^{}$ $^{}$
Multi-scale Modeling with Cellular Automata: The Complex Automata Approach
ACRI '08: Proceedings of the 8th international conference on Cellular Automata for Reseach and Industry

Cellular Automata are commonly used to describe complex natural phenomena. In many cases it is required to capture the multi-scale nature of these phenomena. A single Cellular Automata model may not be able to efficiently simulate a wide range of ...
Parallel implementation and scalability analysis of 3D Fast Fourier Transform using 2D domain decomposition

3D FFT is computationally intensive and at the same time requires global or collective communication patterns. The efficient implementation of FFT on extreme scale computers is one of the grand challenges in scientific computing. On parallel computers ...

Comments

Please enable JavaScript to view thecomments powered by Disqus.

Information & Contributors

Information

Published In

cover image International Journal of High Performance Computing Applications

International Journal of High Performance Computing Applications Volume 33, Issue 5

Sep 2019

303 pages

ISSN:1094-3420

Issue’s Table of Contents

Publisher

Sage Publications, Inc.

United States

Publication History

Published: 01 September 2019

Author Tags

Qualifiers

Research-article

Contributors

Other Metrics

View Article Metrics

Bibliometrics & Citations

Bibliometrics

Article Metrics

1
Total Citations
View Citations
0
Total Downloads

Downloads (Last 12 months)0
Downloads (Last 6 weeks)0

Reflects downloads up to 18 Dec 2024

Other Metrics

View Author Metrics

Citations

Cited By

View all

Milićević BIvanović MStojanović BMilošević MKojić MFilipović N(2022)Huxley muscle model surrogates for high-speed multi-scale simulations of cardiac contractionComputers in Biology and Medicine10.1016/j.compbiomed.2022.105963149:COnline publication date: 14-Dec-2022
https://dl.acm.org/doi/10.1016/j.compbiomed.2022.105963

Abstract

References

Author biographies

Cited By

Index Terms

Recommendations

Computational homogenization with million-way parallelism using domain decomposition methods

Multi-scale Modeling with Cellular Automata: The Complex Automata Approach

Parallel implementation and scalability analysis of 3D Fast Fourier Transform using 2D domain decomposition

Comments

Information

Published In

Publisher

Publication History

Author Tags

Qualifiers

Contributors

Other Metrics

Bibliometrics

Article Metrics

Other Metrics

Citations

Cited By

View options

Figures

Other

Share

Share this Publication link

Share on social media

Affiliations