Computational Science ICCS 2018 18th

International Conference Wuxi China

June 11 13 2018 Proceedings Part II
Yong Shi
Visit to download the full and correct content document:
https://textbookfull.com/product/computational-science-iccs-2018-18th-international-c
onference-wuxi-china-june-11-13-2018-proceedings-part-ii-yong-shi/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Computational Science ICCS 2018 18th International

Conference Wuxi China June 11 13 2018 Proceedings Part
III Yong Shi

https://textbookfull.com/product/computational-science-
iccs-2018-18th-international-conference-wuxi-china-
june-11-13-2018-proceedings-part-iii-yong-shi/

Intelligence Science II Third IFIP TC 12 International

Conference ICIS 2018 Beijing China November 2 5 2018
Proceedings Zhongzhi Shi

https://textbookfull.com/product/intelligence-science-ii-third-
ifip-tc-12-international-conference-icis-2018-beijing-china-
november-2-5-2018-proceedings-zhongzhi-shi/

Computational Science and Its Applications ICCSA 2018

18th International Conference Melbourne VIC Australia
July 2 5 2018 Proceedings Part III Osvaldo Gervasi

https://textbookfull.com/product/computational-science-and-its-
applications-iccsa-2018-18th-international-conference-melbourne-
vic-australia-july-2-5-2018-proceedings-part-iii-osvaldo-gervasi/

Computational Science and Its Applications ICCSA 2018

18th International Conference Melbourne VIC Australia
July 2 5 2018 Proceedings Part V Osvaldo Gervasi

https://textbookfull.com/product/computational-science-and-its-
applications-iccsa-2018-18th-international-conference-melbourne-
vic-australia-july-2-5-2018-proceedings-part-v-osvaldo-gervasi/
Runtime Verification 18th International Conference RV
2018 Limassol Cyprus November 10 13 2018 Proceedings
Christian Colombo

https://textbookfull.com/product/runtime-verification-18th-
international-conference-rv-2018-limassol-cyprus-
november-10-13-2018-proceedings-christian-colombo/

Theory of Cryptography 16th International Conference

TCC 2018 Panaji India November 11 14 2018 Proceedings
Part II Amos Beimel

https://textbookfull.com/product/theory-of-cryptography-16th-
international-conference-tcc-2018-panaji-india-
november-11-14-2018-proceedings-part-ii-amos-beimel/

Neural Information Processing: 25th International

Conference, ICONIP 2018, Siem Reap, Cambodia, December
13–16, 2018, Proceedings, Part II Long Cheng

https://textbookfull.com/product/neural-information-
processing-25th-international-conference-iconip-2018-siem-reap-
cambodia-december-13-16-2018-proceedings-part-ii-long-cheng/

Computational Science ICCS 2020 20th International

Conference Amsterdam The Netherlands June 3 5 2020
Proceedings Part I Valeria V. Krzhizhanovskaya

https://textbookfull.com/product/computational-science-
iccs-2020-20th-international-conference-amsterdam-the-
netherlands-june-3-5-2020-proceedings-part-i-valeria-v-
krzhizhanovskaya/

Computational Science ICCS 2020 20th International

Conference Amsterdam The Netherlands June 3 5 2020
Proceedings Part IV Valeria V. Krzhizhanovskaya

https://textbookfull.com/product/computational-science-
iccs-2020-20th-international-conference-amsterdam-the-
netherlands-june-3-5-2020-proceedings-part-iv-valeria-v-
 Yong Shi · Haohuan Fu · Yingjie Tian
Valeria V. Krzhizhanovskaya
Michael Harold Lees · Jack Dongarra
Peter M. A. Sloot (Eds.)
LNCS 10861

Computational
Science – ICCS 2018
18th International Conference
Wuxi, China, June 11–13, 2018
Proceedings, Part II

123
Lecture Notes in Computer Science 10861
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology Madras, Chennai, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7407
Yong Shi Haohuan Fu
•

Yingjie Tian Valeria V. Krzhizhanovskaya

•

Michael Harold Lees Jack Dongarra

•

Peter M. A. Sloot (Eds.)

Computational
Science – ICCS 2018
18th International Conference
Wuxi, China, June 11–13, 2018
Proceedings, Part II

123
Editors
Yong Shi Michael Harold Lees
Chinese Academy of Sciences University of Amsterdam
Beijing Amsterdam
China The Netherlands
Haohuan Fu Jack Dongarra
National Supercomputing Center in Wuxi University of Tennessee
Wuxi Knoxville, TN
China USA
Yingjie Tian Peter M. A. Sloot
Chinese Academy of Sciences University of Amsterdam
Beijing Amsterdam
China The Netherlands
Valeria V. Krzhizhanovskaya
University of Amsterdam
Amsterdam
The Netherlands

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-319-93700-7 ISBN 978-3-319-93701-4 (eBook)
https://doi.org/10.1007/978-3-319-93701-4

Library of Congress Control Number: 2018947305

LNCS Sublibrary: SL1 – Theoretical Computer Science and General Issues

© Springer International Publishing AG, part of Springer Nature 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by the registered company Springer International Publishing AG
part of Springer Nature
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
 Preface

Welcome to the proceedings of the 18th Annual International Conference on Com-

putational Science (ICCS: https://www.iccs-meeting.org/iccs2018/), held during June
11–13, 2018, in Wuxi, China. Located in the Jiangsu province, Wuxi is bordered by
Changzhou to the west and Suzhou to the east. The city meets the Yangtze River in the
north and is bathed by Lake Tai to the south. Wuxi is home to many parks, gardens,
temples, and the fastest supercomputer in the world, the Sunway TaihuLight. ICCS
2018 was jointly organized by the University of Chinese Academy of Sciences, the
National Supercomputing Center in Wuxi, the University of Amsterdam, NTU
Singapore, and the University of Tennessee.
The International Conference on Computational Science is an annual conference
that brings together researchers and scientists from mathematics and computer science
as basic computing disciplines, researchers from various application areas who are
pioneering computational methods in sciences such as physics, chemistry, life sciences,
and engineering, as well as in arts and humanitarian fields, to discuss problems and
solutions in the area, to identify new issues, and to shape future directions for research.
Since its inception in 2001, ICCS has attracted increasingly higher quality and
numbers of attendees and papers, and this year was no an exception, with over 350
expected participants. The proceedings series have become a major intellectual
resource for computational science researchers, defining and advancing the state of the
art in this field.
ICCS 2018 in Wuxi, China, was the 18th in this series of highly successful conferences.
For the previous 17 meetings, see: http://www.iccs-meeting.org/iccs2018/previous-iccs/.
The theme for ICCS 2018 was “Science at the Intersection of Data, Modelling and
Computation,” to highlight the role of computation as a fundamental method of sci-
entific inquiry and technological discovery tackling problems across scientific domains
and creating synergies between disciplines. This conference was a unique event
focusing on recent developments in: scalable scientific algorithms; advanced software
tools; computational grids; advanced numerical methods; and novel application areas.
These innovative novel models, algorithms, and tools drive new science through effi-
cient application in areas such as physical systems, computational and systems biology,
environmental systems, finance, and others.
ICCS is well known for its excellent line up of keynote speakers. The keynotes for
2018 were:
• Charlie Catlett, Argonne National Laboratory|University of Chicago, USA
• Xiaofei Chen, Southern University of Science and Technology, China
• Liesbet Geris, University of Liège|KU Leuven, Belgium
• Sarika Jalan, Indian Institute of Technology Indore, India
• Petros Koumoutsakos, ETH Zürich, Switzerland
• Xuejun Yang, National University of Defense Technology, China
VI Preface

This year we had 405 submissions (180 submissions to the main track and 225 to the
workshops). In the main track, 51 full papers were accepted (28%). In the workshops,
97 full papers (43%). A high acceptance rate in the workshops is explained by the
nature of these thematic sessions, where many experts in a particular field are per-
sonally invited by workshop organizers to participate in their sessions.
ICCS relies strongly on the vital contributions of our workshop organizers to attract
high-quality papers in many subject areas. We would like to thank all committee
members for the main track and workshops for their contribution toward ensuring a
high standard for the accepted papers. We would also like to thank Springer, Elsevier,
Intellegibilis, Beijing Vastitude Technology Co., Ltd. and Inspur for their support.
Finally, we very much appreciate all the local Organizing Committee members for their
hard work to prepare this conference.
We are proud to note that ICCS is an ERA 2010 A-ranked conference series.

June 2018 Yong Shi

Haohuan Fu
Yingjie Tian
Valeria V. Krzhizhanovskaya
Michael Lees
Jack Dongarra
Peter M. A. Sloot
The ICCS 2018 Organizers
 Organization

Local Organizing Committee

Co-chairs
Yingjie Tian University of Chinese Academy of Sciences, China
Lin Gan National Supercomputing Center in Wuxi, China

Members
Jiming Wu National Supercomputing Center in Wuxi, China
Lingying Wu National Supercomputing Center in Wuxi, China
Jinzhe Yang National Supercomputing Center in Wuxi, China
Bingwei Chen National Supercomputing Center in Wuxi, China
Yuanchun Zheng University of Chinese Academy of Sciences, China
Minglong Lei University of Chinese Academy of Sciences, China
Jia Wu Macquarie University, Australia
Zhengsong Chen University of Chinese Academy of Sciences, China
Limeng Cui University of Chinese Academy of Sciences, China
Jiabin Liu University of Chinese Academy of Sciences, China
Biao Li University of Chinese Academy of Sciences, China
Yunlong Mi University of Chinese Academy of Sciences, China
Wei Dai University of Chinese Academy of Sciences, China

Workshops and Organizers

Advances in High-Performance Computational Earth Sciences: Applications and

Frameworks – IHPCES 2018
Xing Cai, Kohei Fujita, Takashi Shimokawabe
Agent-Based Simulations, Adaptive Algorithms, and Solvers – ABS-AAS 2018
Robert Schaefer, Maciej Paszynski, Victor Calo, David Pardo
Applications of Matrix Methods in Artificial Intelligence and Machine Learning –
AMAIML 2018
Kourosh Modarresi
Architecture, Languages, Compilation, and Hardware Support for Emerging
Manycore Systems – ALCHEMY 2018
Loïc Cudennec, Stéphane Louise
Biomedical and Bioinformatics Challenges for Computer Science – BBC 2018
Giuseppe Agapito, Mario Cannataro, Mauro Castelli, Riccardo Dondi, Rodrigo Weber
dos Santos, Italo Zoppis
VIII Organization

Computational Finance and Business Intelligence – CFBI 2018

Shouyang Wang, Yong Shi, Yingjie Tian
Computational Optimization, Modelling, and Simulation – COMS 2018
Xin-She Yang, Slawomir Koziel, Leifur Leifsson, T. O. Ting
Data-Driven Computational Sciences – DDCS 2018
Craig Douglas, Abani Patra, Ana Cortés, Robert Lodder
Data, Modeling, and Computation in IoT and Smart Systems – DMC-IoT 2018
Julien Bourgeois, Vaidy Sunderam, Hicham Lakhlef
Mathematical Methods and Algorithms for Extreme Scale – MATH-EX 2018
Vassil Alexandrov
Multiscale Modelling and Simulation – MMS 2018
Derek Groen, Lin Gan, Valeria Krzhizhanovskaya, Alfons Hoekstra
Simulations of Flow and Transport: Modeling, Algorithms, and Computation –
SOFTMAC 2018
Shuyu Sun, Jianguo (James) Liu, Jingfa Li
Solving Problems with Uncertainties – SPU 2018
Vassil Alexandrov
Teaching Computational Science – WTCS 2018
Angela B. Shiflet, Alfredo Tirado-Ramos, Nia Alexandrov
Tools for Program Development and Analysis in Computational Science –
TOOLS 2018
Karl Fürlinger, Arndt Bode, Andreas Knüpfer, Dieter Kranzlmüller, Jens Volkert,
Roland Wismüller
Urgent Computing – UC 2018
Marian Bubak, Alexander Boukhanovsky

Program Committee

Ahmad Abdelfattah Ioannis Anagnostou Adam Belloum

David Abramson Michael Antolovich Abdelhak Bentaleb
Giuseppe Agapito Hartwig Anzt Stefano Beretta
Ram Akella Hideo Aochi Daniel Berrar
Elisabete Alberdi Tomasz Arodz Sanjukta Bhowmick
Marco Aldinucci Tomàs Artés Vivancos Anna Bilyatdinova
Nia Alexandrov Victor Azizi Tarksalooyeh Guillaume Blin
Vassil Alexandrov Ebrahim Bagheri Nasri Bo
Saad Alowayyed Bartosz Balis Marcel Boersma
Ilkay Altintas Krzysztof Banas Bartosz Bosak
Stanislaw Jörn Behrens Kris Bubendorfer
Ambroszkiewicz Adrian Bekasiewicz Jérémy Buisson
 Organization IX

Aleksander Byrski Christos Jaap Kaandorp

Wentong Cai Filelis-Papadopoulos Viacheslav Kalashnikov
Xing Cai Karl Frinkle George Kampis
Mario Cannataro Haohuan Fu Drona Kandhai
Yongcan Cao Karl Fuerlinger Aneta Karaivanova
Pedro Cardoso Kohei Fujita Vlad Karbovskii
Mauro Castelli Wlodzimierz Funika Andrey Karsakov
Eduardo Cesar Takashi Furumura Takahiro Katagiri
Imen Chakroun David Gal Wayne Kelly
Huangxin Chen Lin Gan Deepak Khazanchi
Mingyang Chen Robin Gandhi Alexandra Klimova
Zhensong Chen Frédéric Gava Ivan Kondov
Siew Ann Cheong Alex Gerbessiotis Vladimir Korkhov
Lock-Yue Chew Carlos Gershenson Jari Kortelainen
Ana Cortes Domingo Gimenez Ilias Kotsireas
Enrique Frank Giraldo Jisheng Kou
Costa-Montenegro Ivo Gonçalves Sergey Kovalchuk
Carlos Cotta Yuriy Gorbachev Slawomir Koziel
Jean-Francois Couchot Pawel Gorecki Valeria Krzhizhanovskaya
Helene Coullon George Gravvanis Massimo La Rosa
Attila Csikász-Nagy Derek Groen Hicham Lakhlef
Loïc Cudennec Lutz Gross Roberto Lam
Javier Cuenca Kun Guo Anna-Lena Lamprecht
Yifeng Cui Xiaohu Guo Rubin Landau
Ben Czaja Piotr Gurgul Johannes Langguth
Pawel Czarnul Panagiotis Hadjidoukas Vianney Lapotre
Wei Dai Azzam Haidar Jysoo Lee
Lisandro Dalcin Dongxu Han Michael Lees
Bhaskar Dasgupta Raheel Hassan Minglong Lei
Susumu Date Jurjen Rienk Helmus Leifur Leifsson
Quanling Deng Bogumila Hnatkowska Roy Lettieri
Xiaolong Deng Alfons Hoekstra Andrew Lewis
Minh Ngoc Dinh Paul Hofmann Biao Li
Riccardo Dondi Sergey Ivanov Dewei Li
Tingxing Dong Hideya Iwasaki Jingfa Li
Ruggero Donida Labati Takeshi Iwashita Kai Li
Craig C. Douglas Jiří Jaroš Peijia Li
Rafal Drezewski Marco Javarone Wei Li
Jian Du Chao Jin I-Jong Lin
Vitor Duarte Hai Jin Hong Liu
Witold Dzwinel Zhong Jin Hui Liu
Nahid Emad Jingheng James Liu
Christian Engelmann David Johnson Jiabin Liu
Daniel Etiemble Anshul Joshi Piyang Liu
X Organization

Weifeng Liu Marcin Paprzycki Peter Sloot

Weiguo Liu David Pardo Renata Slota
Marcelo Lobosco Anna Paszynska Grażyna Ślusarczyk
Robert Lodder Maciej Paszynski Sucha Smanchat
Wen Long Abani Patra Maciej Smołka
Stephane Louise Dana Petcu Bartlomiej Sniezynski
Frederic Loulergue Eric Petit Sumit Sourabh
Paul Lu Serge Petiton Achim Streit
Sheraton M. V. Gauthier Picard Barbara Strug
Scott MacLachlan Daniela Piccioni Bongwon Suh
Maciej Malawski Yuri Pirola Shuyu Sun
Michalska Malgorzatka Antoniu Pop Martin Swain
Vania Ela Pustulka-Hunt Ryszard Tadeusiewicz
Marangozova-Martin Vladimir Puzyrev Daisuke Takahashi
Tomas Margalef Alexander Pyayt Jingjing Tang
Tiziana Margaria Pei Quan Osamu Tatebe
Svetozar Margenov Rick Quax Andrei Tchernykh
Osni Marques Waldemar Rachowicz Cedric Tedeschi
Pawel Matuszyk Lukasz Rauch Joao Teixeira
Valerie Maxville Alistair Rendell Yonatan Afework
Rahul Mazumder Sophie Robert Tesfahunegn
Valentin Melnikov J. M. F Rodrigues Andrew Thelen
Ivan Merelli Daniel Rodriguez Xin Tian
Doudou Messoud Albert Romkes Yingjie Tian
Yunlong Mi James A. Ross T. O. Ting
Jianyu Miao Debraj Roy Alfredo Tirado-Ramos
John Michopoulos Philip Rutten Stanimire Tomov
Sergey Mityagin Katarzyna Rycerz Ka Wai Tsang
K. Modarresi Alberto Sanchez Britt van Rooij
Kourosh Modarresi Rodrigo Santos Raja Velu
Jânio Monteiro Hitoshi Sato Antonio M. Vidal
Paulo Moura Oliveira Robert Schaefer David Walker
Ignacio Muga Olaf Schenk Jianwu Wang
Hiromichi Nagao Ulf D. Schiller Peng Wang
Kengo Nakajima Bertil Schmidt Yi Wang
Denis Nasonov Hichem Sedjelmaci Josef Weinbub
Philippe Navaux Martha Johanna Mei Wen
Hoang Nguyen Sepulveda Mark Wijzenbroek
Mai Nguyen Yong Shi Maciej Woźniak
Anna Nikishova Angela Shiflet Guoqiang Wu
Lingfeng Niu Takashi Shimokawabe Jia Wu
Mawloud Omar Tan Singyee Qing Wu
Kenji Ono Robert Sinkovits Huilin Xing
Raymond Padmos Vishnu Sivadasan Wei Xue
 Organization XI

Chao-Tung Yang Peng Zhang Jinghui Zhong

Xin-She Yang Yao Zhang Xiaofei Zhou
He Yiwei Zepu Zhang Luyao Zhu
Ce Yu Wenlai Zhao Sotirios Ziavras
Ma Yue Yuanchun Zheng Andrea Zonca
Julija Zavadlav He Zhong Italo Zoppis
Gábor Závodszky Hua Zhong
 Contents – Part II

Track of Advances in High-Performance Computational

Earth Sciences: Applications and Frameworks

Development of Scalable Three-Dimensional Elasto-Plastic Nonlinear

Wave Propagation Analysis Method for Earthquake Damage Estimation
of Soft Grounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Atsushi Yoshiyuki, Kohei Fujita, Tsuyoshi Ichimura, Muneo Hori,
and Lalith Wijerathne

A New Matrix-Free Approach for Large-Scale Geodynamic Simulations

and its Performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Simon Bauer, Markus Huber, Marcus Mohr, Ulrich Rüde,
and Barbara Wohlmuth

Viscoelastic Crustal Deformation Computation Method with Reduced

Random Memory Accesses for GPU-Based Computers . . . . . . . . . . . . . . . . 31
Takuma Yamaguchi, Kohei Fujita, Tsuyoshi Ichimura, Anne Glerum,
Ylona van Dinther, Takane Hori, Olaf Schenk, Muneo Hori,
and Lalith Wijerathne

An Event Detection Framework for Virtual Observation System:

Anomaly Identification for an ACME Land Simulation . . . . . . . . . . . . . . . . 44
Zhuo Yao, Dali Wang, Yifan Wang, and Fengming Yuan

Enabling Adaptive Mesh Refinement for Single Components in ECHAM6. . . 56

Yumeng Chen, Konrad Simon, and Jörn Behrens

Efficient and Accurate Evaluation of Bézier Tensor Product Surfaces . . . . . . 69

Jing Lan, Hao Jiang, and Peibing Du

Track of Agent-Based Simulations, Adaptive Algorithms and Solvers

Hybrid Swarm and Agent-Based Evolutionary Optimization . . . . . . . . . . . . . 89

Leszek Placzkiewicz, Marcin Sendera, Adam Szlachta,
Mateusz Paciorek, Aleksander Byrski, Marek Kisiel-Dorohinicki,
and Mateusz Godzik

Data-Driven Agent-Based Simulation for Pedestrian Capacity Analysis . . . . . 103

Sing Kuang Tan, Nan Hu, and Wentong Cai
XIV Contents – Part II

A Novel Agent-Based Modeling Approach for Image Coding and Lossless

Compression Based on the Wolf-Sheep Predation Model . . . . . . . . . . . . . . . 117
Khaldoon Dhou

Planning Optimal Path Networks Using Dynamic Behavioral Modeling . . . . . 129

Sergei Kudinov, Egor Smirnov, Gavriil Malyshev, and Ivan Khodnenko

Multiagent Context-Dependent Model of Opinion Dynamics

in a Virtual Society . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
Ivan Derevitskii, Oksana Severiukhina, Klavdiya Bochenina,
Daniil Voloshin, Anastasia Lantseva, and Alexander Boukhanovsky

An Algorithm for Tensor Product Approximation of Three-Dimensional

Material Data for Implicit Dynamics Simulations . . . . . . . . . . . . . . . . . . . . 156
Krzysztof Podsiadło, Marcin Łoś, Leszek Siwik, and Maciej Woźniak

Track of Applications of Matrix Methods in Artificial Intelligence

and Machine Learning

On Two Kinds of Dataset Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . 171

Pavel Emelyanov

A Graph-Based Algorithm for Supervised Image Classification . . . . . . . . . . . 184

Ke Du, Jinlong Liu, Xingrui Zhang, Jianying Feng, Yudong Guan,
and Stéphane Domas

An Adversarial Training Framework for Relation Classification . . . . . . . . . . 194

Wenpeng Liu, Yanan Cao, Cong Cao, Yanbing Liu, Yue Hu, and Li Guo

Topic-Based Microblog Polarity Classification Based on Cascaded Model . . . 206

Quanchao Liu, Yue Hu, Yangfan Lei, Xiangpeng Wei, Guangyong Liu,
and Wei Bi

An Efficient Deep Learning Model for Recommender Systems . . . . . . . . . . . 221

Kourosh Modarresi and Jamie Diner

Standardization of Featureless Variables for Machine Learning Models

Using Natural Language Processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
Kourosh Modarresi and Abdurrahman Munir

Generalized Variable Conversion Using K-means Clustering

and Web Scraping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Kourosh Modarresi and Abdurrahman Munir

Parallel Latent Dirichlet Allocation on GPUs . . . . . . . . . . . . . . . . . . . . . . . 259

Gordon E. Moon, Israt Nisa, Aravind Sukumaran-Rajam,
Bortik Bandyopadhyay, Srinivasan Parthasarathy, and P. Sadayappan
 Contents – Part II XV

Improving Search Through A3C Reinforcement Learning Based

Conversational Agent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Milan Aggarwal, Aarushi Arora, Shagun Sodhani,
and Balaji Krishnamurthy

Track of Architecture, Languages, Compilation and Hardware

Support for Emerging ManYcore Systems

Architecture Emulation and Simulation of Future Many-Core Epiphany

RISC Array Processors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
David A. Richie and James A. Ross

Automatic Mapping for OpenCL-Programs on CPU/GPU Heterogeneous

Platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Konrad Moren and Diana Göhringer

Track of Biomedical and Bioinformatics Challenges

for Computer Science

Combining Data Mining Techniques to Enhance Cardiac Arrhythmia

Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
Christian Gomes, Alan Cardoso, Thiago Silveira, Diego Dias,
Elisa Tuler, Renato Ferreira, and Leonardo Rocha

CT Medical Imaging Reconstruction Using Direct Algebraic Methods

with Few Projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
Mónica Chillarón, Vicente Vidal, Gumersindo Verdú, and Josep Arnal

On Blood Viscosity and Its Correlation with Biological Parameters . . . . . . . . 347

Patrizia Vizza, Giuseppe Tradigo, Marianna Parrilla,
Pietro Hiram Guzzi, Agostino Gnasso, and Pierangelo Veltri

Development of Octree-Based High-Quality Mesh Generation Method

for Biomedical Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
Keisuke Katsushima, Kohei Fujita, Tsuyoshi Ichimura, Muneo Hori,
and Lalith Maddegedara

1,000x Faster Than PLINK: Genome-Wide Epistasis Detection

with Logistic Regression Using Combined FPGA and GPU Accelerators . . . . 368
Lars Wienbrandt, Jan Christian Kässens, Matthias Hübenthal,
and David Ellinghaus
XVI Contents – Part II

Track of Computational Finance and Business Intelligence

Deep Learning and Wavelets for High-Frequency Price Forecasting . . . . . . . 385

Andrés Arévalo, Jaime Nino, Diego León, German Hernandez,
and Javier Sandoval

Kernel Extreme Learning Machine for Learning from Label Proportions . . . . 400
Hao Yuan, Bo Wang, and Lingfeng Niu

Extreme Market Prediction for Trading Signal with Deep Recurrent

Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
Zhichen Lu, Wen Long, and Ying Guo

Multi-view Multi-task Support Vector Machine. . . . . . . . . . . . . . . . . . . . . . 419

Jiashuai Zhang, Yiwei He, and Jingjing Tang

Research on Stock Price Forecast Based on News Sentiment

Analysis—A Case Study of Alibaba . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
Lingling Zhang, Saiji Fu, and Bochen Li

Parallel Harris Corner Detection on Heterogeneous Architecture . . . . . . . . . . 443

Yiwei He, Yue Ma, Dalian Liu, and Xiaohua Chen

A New Method for Structured Learning with Privileged Information . . . . . . . 453

Shiding Sun, Chunhua Zhang, and Yingjie Tian

An Effective Model Between Mobile Phone Usage and P2P

Default Behavior. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
Huan Liu, Lin Ma, Xi Zhao, and Jianhua Zou

A Novel Data Mining Approach Towards Human Resource

Performance Appraisal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
Pei Quan, Ying Liu, Tianlin Zhang, Yueran Wen, Kaichao Wu,
Hongbo He, and Yong Shi

Word Similarity Fails in Multiple Sense Word Embedding . . . . . . . . . . . . . . 489

Yong Shi, Yuanchun Zheng, Kun Guo, Wei Li, and Luyao Zhu

Track of Computational Optimization, Modelling and Simulation

A Hybrid Optimization Algorithm for Electric Motor Design . . . . . . . . . . . . 501

Mokhtar Essaid, Lhassane Idoumghar, Julien Lepagnot,
Mathieu Brévilliers, and Daniel Fodorean

Dynamic Current Distribution in the Electrodes of Submerged Arc Furnace

Using Scalar and Vector Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
Yonatan Afework Tesfahunegn, Thordur Magnusson, Merete Tangstad,
and Gudrun Saevarsdottir
 Contents – Part II XVII

Optimising Deep Learning by Hyper-heuristic Approach for Classifying

Good Quality Images. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
Muneeb ul Hassan, Nasser R. Sabar, and Andy Song

An Agent-Based Distributed Approach for Bike Sharing Systems . . . . . . . . . 540

Ningkui Wang, Hayfa Zgaya, Philippe Mathieu, and Slim Hammadi

A Fast Vertex-Swap Operator for the Prize-Collecting Steiner Tree Problem . . . 553
Yi-Fei Ming, Si-Bo Chen, Yong-Quan Chen, and Zhang-Hua Fu

Solving CSS-Sprite Packing Problem Using a Transformation

to the Probabilistic Non-oriented Bin Packing Problem . . . . . . . . . . . . . . . . 561
Soumaya Sassi Mahfoudh, Monia Bellalouna, and Leila Horchani

Optimization of Resources Selection for Jobs Scheduling in Heterogeneous

Distributed Computing Environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 574
Victor Toporkov and Dmitry Yemelyanov

Explicit Size-Reduction-Oriented Design of a Compact Microstrip

Rat-Race Coupler Using Surrogate-Based Optimization Methods. . . . . . . . . . 584
Slawomir Koziel, Adrian Bekasiewicz, Leifur Leifsson, Xiaosong Du,
and Yonatan Tesfahunegn

Stochastic-Expansions-Based Model-Assisted Probability of Detection

Analysis of the Spherically-Void-Defect Benchmark Problem . . . . . . . . . . . . 593
Xiaosong Du, Praveen Gurrala, Leifur Leifsson, Jiming Song,
William Meeker, Ronald Roberts, Slawomir Koziel,
and Yonatan Tesfahunegn

Accelerating Optical Absorption Spectra and Exciton Energy Computation

via Interpolative Separable Density Fitting . . . . . . . . . . . . . . . . . . . . . . . . . 604
Wei Hu, Meiyue Shao, Andrea Cepellotti, Felipe H. da Jornada, Lin Lin,
Kyle Thicke, Chao Yang, and Steven G. Louie

Model-Assisted Probability of Detection for Structural Health Monitoring

of Flat Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 618
Xiaosong Du, Jin Yan, Simon Laflamme, Leifur Leifsson,
Yonatan Tesfahunegn, and Slawomir Koziel

Track of Data, Modeling, and Computation in IoT and Smart Systems

Anomalous Trajectory Detection Between Regions of Interest Based

on ANPR System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
Gao Ying, Nie Yiwen, Yang Wei, Xu Hongli, and Huang Liusheng
XVIII Contents – Part II

Dynamic Real-Time Infrastructure Planning and Deployment for Disaster

Early Warning Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
Huan Zhou, Arie Taal, Spiros Koulouzis, Junchao Wang, Yang Hu,
George Suciu Jr., Vlad Poenaru, Cees de Laat, and Zhiming Zhao

Calibration and Monitoring of IoT Devices by Means of Embedded

Scientific Visualization Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655
Konstantin Ryabinin, Svetlana Chuprina, and Mariia Kolesnik

Gated Convolutional LSTM for Speech Commands Recognition . . . . . . . . . . 669

Dong Wang, Shaohe Lv, Xiaodong Wang, and Xinye Lin

Enabling Machine Learning on Resource Constrained Devices

by Source Code Generation of the Learned Models . . . . . . . . . . . . . . . . . . . 682
Tomasz Szydlo, Joanna Sendorek, and Robert Brzoza-Woch

Track of Data-Driven Computational Sciences

Fast Retrieval of Weather Analogues in a Multi-petabytes Archive

Using Wavelet-Based Fingerprints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 697
Baudouin Raoult, Giuseppe Di Fatta, Florian Pappenberger,
and Bryan Lawrence

Assimilation of Fire Perimeters and Satellite Detections by Minimization

of the Residual in a Fire Spread Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 711
Angel Farguell Caus, James Haley, Adam K. Kochanski,
Ana Cortés Fité, and Jan Mandel

Analyzing Complex Models Using Data and Statistics . . . . . . . . . . . . . . . . . 724

Abani K. Patra, Andrea Bevilacqua, and Ali Akhavan Safei

Research on Technology Foresight Method Based on Intelligent

Convergence in Open Network Environment . . . . . . . . . . . . . . . . . . . . . . . 737
Zhao Minghui, Zhang Lingling, Zhang Libin, and Wang Feng

Prediction of Blasting Vibration Intensity by Improved PSO-SVR

on Apache Spark Cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
Yunlan Wang, Jing Wang, Xingshe Zhou, Tianhai Zhao, and Jianhua Gu

Bisections-Weighted-by-Element-Size-and-Order Algorithm to Optimize

Direct Solver Performance on 3D hp-adaptive Grids . . . . . . . . . . . . . . . . . . 760
H. AbouEisha, V. M. Calo, K. Jopek, M. Moshkov, A. Paszyńska,
and M. Paszyński

Establishing EDI for a Clinical Trial of a Treatment for Chikungunya . . . . . . 773

Cynthia Dickerson, Mark Ensor, and Robert A. Lodder
 Contents – Part II XIX

Static Analysis and Symbolic Execution for Deadlock Detection

in MPI Programs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783
Craig C. Douglas and Krishanthan Krishnamoorthy

Track of Mathematical-Methods-and-Algorithms for Extreme Scale

Reproducible Roulette Wheel Sampling for Message Passing Environments . . . 799

Balazs Nemeth, Tom Haber, Jori Liesenborgs, and Wim Lamotte

Speedup of Bicubic Spline Interpolation. . . . . . . . . . . . . . . . . . . . . . . . . . . 806

Viliam Kačala and Csaba Török

Track of Multiscale Modelling and Simulation

Optimized Eigenvalue Solvers for the Neutron Transport Equation . . . . . . . . 823

Antoni Vidal-Ferràndiz, Sebastián González-Pintor, Damián Ginestar,
Amanda Carreño, and Gumersindo Verdú

Multiscale Homogenization of Pre-treatment Rapid and Slow Filtration

Processes with Experimental and Computational Validations . . . . . . . . . . . . 833
Alvin Wei Ze Chew and Adrian Wing-Keung Law

The Solution of the Lambda Modes Problem Using Block

Iterative Eigensolvers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 846
A. Carreño, A. Vidal-Ferràndiz, D. Ginestar, and G. Verdú

A Versatile Hybrid Agent-Based, Particle and Partial Differential Equations

Method to Analyze Vascular Adaptation . . . . . . . . . . . . . . . . . . . . . . . . . . 856
Marc Garbey, Stefano Casarin, and Scott Berceli

Development of a Multiscale Simulation Approach for Forced Migration . . . . 869

Derek Groen

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 877

Track of Advances in High-Performance
Computational Earth Sciences:
Applications and Frameworks
 Development of Scalable
Three-Dimensional Elasto-Plastic
Nonlinear Wave Propagation Analysis
Method for Earthquake Damage
Estimation of Soft Grounds

Atsushi Yoshiyuki(B) , Kohei Fujita, Tsuyoshi Ichimura, Muneo Hori,

and Lalith Wijerathne

Earthquake Research Institute and Department of Civil Engineering,

The University of Tokyo, Bunkyō, Japan
{y-atsu,fujita,ichimura,hori,lalith}@eri.u-tokyo.ac.jp

Abstract. In soft complex grounds, earthquakes cause damages with

large deformation such as landslides and subsidence. Use of elasto-plastic
models as the constitutive equation of soils is suitable for evaluation of
nonlinear wave propagation with large ground deformation. However,
there is no example of elasto-plastic nonlinear wave propagation analysis
method capable of simulating a large-scale soil deformation problem. In
this study, we developed a scalable elasto-plastic nonlinear wave prop-
agation analysis program based on three-dimensional nonlinear finite-
element method. The program attains 86.2% strong scaling efficiency
from 240 CPU cores to 3840 CPU cores of PRIMEHPC FX10 based
Oakleaf-FX [1], with 8.85 TFLOPS (15.6% of peak) performance on 3840
CPU cores. We verified the elasto-plastic nonlinear wave propagation
program through convergence analysis, and conducted an analysis with
large deformation for an actual soft ground modeled using 47,813,250
degrees-of-freedom.

1 Introduction

Large earthquakes often cause severe damage in cut-and-fill land developed for
housing. It is said that earthquake waves are amplified locally by impedance con-
trast between the cut layer and fill layer, which causes damage. To evaluate this
wave amplification, 3D wave propagation analysis with high spatial resolution
considering nonlinearity of soil properties is required. Finite-element methods
(FEM) are suitable for solving problems with complex geometry, and nonlinear
constitutive relations can be implemented. However, large-scale finite-element
analysis is computational expensive to assure convergence of the numerical solu-
tion.
Efficient use of high performance computers is effective for solving this prob-
lem [2,3]. For example, Ichimura et al. [4] developed a fast and scalable 3D
c Springer International Publishing AG, part of Springer Nature 2018
Y. Shi et al. (Eds.): ICCS 2018, LNCS 10861, pp. 3–16, 2018.
https://doi.org/10.1007/978-3-319-93701-4_1
4 A. Yoshiyuki et al.

nonlinear wave propagation analysis method based on nonlinear FEM, and was
selected as a Gordon Bell Prize Finalist in SC14. Here, computational methods
for speeding up the iterative solver was developed, which enabled large-scale
analysis on distributed-shared memory parallel supercomputers such as the K
computer [5]. In this method, a simple nonlinear model (Ramberg-Osgood model
[6] with the Masing rule [7]) was used for the constitutive equation of soils, and
the program was used for estimating earthquake damage at sites with complex
grounds [8]. However, this simple constitutive equation is insufficient for simu-
lating permanent ground displacement; 3D elasto-plastic constitutive equations
are required to conduct reliable nonlinear wave propagation analysis for soft
grounds. On the other hand, existing elasto-plastic nonlinear wave propagation
analysis programs based on nonlinear FEM for seismic response of soils are not
designed for high performance computers, and thus they cannot be used for large
scale analyses.
In this study, we develop a scalable 3D elasto-plastic nonlinear wave prop-
agation analysis method based on the highly efficient FEM solver described in
[4]. Here, we incorporate a standard 3D elasto-plastic constitutive equation for
soft soils (i.e., super-subloading surface Sekiguchi-Ohta EC model [9–11]) into
this FEM solver. The FEM solver is also extended to conduct self-weight analy-
sis, which is essential for conducting elasto-plastic analysis. This enables large-
scale 3D elasto-plastic nonlinear wave propagation analysis, which is required
for assuring numerical convergence when computing seismic response of soft
grounds.
The rest of the paper is organized as follows. In Sect. 2, we describe the
target equation and the developed nonlinear wave propagation analysis method.
In Sect. 3, we verify the method through a convergence test, apply the method
to an actual site, and measure the computational performance of the method.
Section 4 concludes the paper.

2 Methodology

Previous wave propagation analysis based on nonlinear FEM [4] used the
Ramberg-Osgood model and Masing rule for the constitutive equation of soils.
Instead, we apply an elasto-plastic model (super-subloading surface Sekiguchi-
Ohta EC model) to this FEM solver for analyzing large ground deformation. In
elasto-plastic nonlinear wave propagation analysis, we first find an initial stress
state by conducting initial stress analysis considering gravitational forces, and
then conduct nonlinear wave propagation analysis by inputting seismic waves.
Since the previous FEM implementation was not able to carry out initial stress
analysis and nonlinear wave propagation analysis successively, we extended the
solver. In this section, we first describe the target wave propagation problem with
the super-subloading surface Sekiguchi-Ohta EC model, and then we describe the
developed scalable elasto-plastic nonlinear wave propagation analysis method.
 Development of Scalable Three-Dimensional Elasto-Plastic Nonlinear Wave 5

2.1 Target Problem

We use the following equation obtained by discretizing the nonlinear wave equa-
tion in the spatial domain by FEM and the time domain by the Newmark-β
method:
 
4 2 n n
M + C + K δun
dt2 dt
 
4
= f n − qn−1 + Cn vn−1 + M an−1 + vn−1 , (1)
dt

with ⎧ n
⎪
⎪q = qn−1 + Kn δun ,
⎪ n
⎨u = un−1 + δun ,
(2)
⎪
⎪vn = −vn−1 + dt
2
δun ,
⎪
⎩ n
a = −an−1 − dt
4 n−1
v + 4 n
dt2 δu .

Here, δu, u, v, a, and f are vectors describing incremental displacement, displace-

ment, velocity, acceleration, and external force, respectively. M, C, and K are
the mass, damping, and stiffness matrices. dt, and n are the time step increment
and the time step number, respectively. In the case that nonlinearity occurs,
C, K change every time steps. Rayleigh damping is used for the damping matrix
C, where the element damping matrix Cne is calculated using the element mass
matrix Me and the element stiffness matrix Kne as follows:

Cne = α∗ Me + β ∗ Kne ,

The coefficients α∗ and β ∗ are determined by solving the following least-squares

equation, 
fmax   2
1 α∗ ∗
minimize h −
n
+ 2πf β df .
fmin 2 2πf
where fmax and fmin are the maximum and minimum target frequencies and
hn is the damping ratio at time step n. Small elements are locally generated
when modeling complex geometry with solid elements, and therefore satisfy-
ing the Courant condition when using explicit time integration methods (e.g.,
central difference method) leads to small time increments and considerable com-
putational cost. Thus, the Newmark-β method is used for time integration with
β = 1/4, δ = 1/2 (β and δ are parameters of the Newmark-β method). By
applying Semi-infinite absorbing boundary conditions to the bottom and side
boundaries of the simulation domain, we take dissipation character and semi-
infinite character into consideration.
Next we summarize the super-subloading surface Sekiguchi-Ohta EC model
[9–11], which is one of the 3D elasto-plastic constitutive equations used in nonlin-
ear wave propagation analysis of soils. The super-subloading surface Sekiguchi-
Ohta EC model is described using subloading and superloading surfaces sum-
marized in Fig. 1. The subloading surface is a yield surface defined inside of the
normal yield surface. It is similar in shape to the normal yield surface and a
6 A. Yoshiyuki et al.

current stress state is always on it. We can take into account plastic deforma-
tion in the normal yield surface and reproduce smooth change from elastic state
to plastic state by introducing the subloading surface. On the other hand, the
superloading surface is a yield surface defined outside of the normal yield surface.
It is similar in shape to the normal yield surface and the subloading surface. Rel-
ative contraction of the superloading surface (i.e., the expansion of the normal
yield surface) describes the decay of the structure as plastic deformation pro-
ceeds. At the end, the superloading surface and the normal yield surface become
identical. Similarity ratios of the subloading surface to the superloading surface,
of the normal yield surface to the superloading surface are denoted by R, R∗ ,
respectively (0 < R ≤ 1, 0 < R∗ ≤ 1). 1/R is overconsolidation ratio and R is
the index of degree of structure. As plastic deformation proceeds, the subloading
surface expands and the superloading surface relatively contracts. The expan-
sion speed Ṙ and contraction speed R˙∗ are calculated as in Fig. 1. D, ˙p are the
coefficient of dilatancy, the plastic volumetric strain speed and m, a, b, c are the
degradation parameters of overconsolidated state and structures state, respec-
tively. Using this R and R∗ , a yield function of the subloading surface is described
as f (σ  , v p ) in Fig. 1. Here, M, nE , σ  , σ0  are the critical state parameter, the
fitting parameter, the effective stress tensor, the effective initial stress tensor
and η ∗ , p , q are the stress parameter proposed by Sekiguchi and Ohta, the effec-
tive mean stress, the deviatoric stress. The following stress-strain relationship is
obtained by solving the simultaneous equations in Fig. 1.
⎛ ⎞
⎜ e Ce : ∂f ⊗ ∂f
 : C
e
˙ 
σ = ⎝C − ∂σ ∂σ
    ⎟
⎠ : ˙ ,
∂f e : ∂f − ∂f ∂f m (ln R) ∂f  ∂f  ∂f  ∂f 
: C ∂v p ∂p
+  ∂σ   − a (R∗ )b (1 − R∗ )c ∂R ∗  ∂σ  
∂σ  ∂σ  D ∂R

ep
= C : ˙ , (3)

where,
 
2
e
Cijkl = K − G δij δkl + G (δik δjl + δil δjk ) ,
3
Λ 3 (1 − 2ν  )
K= p , G = K,
M D (1 − Λ) 2 (1 + ν  )

Ce (Cijkl
e
), Cep are the elasticity tensor, the elasto-plasticity tensor and
K, G, Λ, ν  are the bulk modulus, the shear modulus, the irreversibility ratio,
the effective Poisson’s ratio, respectively.

2.2 Fast and Scalable Elasto-Plastic Nonlinear Analysis Method

In this subsection, we first summarize the solver algorithm in [4] following
Algorithm 1. By changing the K matrix in Algorithm 1 according to the change
in the constitutive model, we can expect high computational efficiency when con-
ducting elasto-plastic analyses. In the latter part of the subsection, we describe
the initial stress analysis and nonlinear wave propagation analysis procedure.
The majority of the cost in conducting finite-element analysis is in solving
the linear equation in Eq. (1). The solver in [4] enables fast and scalable solving
 Development of Scalable Three-Dimensional Elasto-Plastic Nonlinear Wave 7

Fig. 1. Governing equation of stress-strain relation and relation of yield surfaces

of Eq. (1) by using adaptive conjugate gradient (CG) method with multi-grid
preconditioning, mixed precision arithmetics, and fast matrix-vector multiplica-
tion based on the Element-by-Element method [12,13]. Instead of storing a fixed
preconditioning matrix, the preconditioning equation is solved roughly using an
another CG solver. In Algorithm 1, outer loop means the iterative calculation of
the CG method solving Ax = b, and the inner loop means the computation of
preconditioning equation (solving z = A−1 r by CG method). Since the precondi-
tioning equation needs only be solved roughly, single-precision arithmetic is used
in the preconditioner, while double precision arithmetic is used in the outer loop.
Furthermore, the multi-grid method is used in the preconditioner to improve
convergence in the inner loop itself. Here, a two-step grid with second-order
tetrahedral mesh (FEMmodel) and first-order tetrahedral mesh (FEMmodelc )
is used. Specifically, an initial solution of z = A−1 r is estimated by computing
zc = Ac −1 rc , which reduces the number of iterations in solving z = A−1 r. In
order to reduce memory footprint, memory transfer sizes, and improve load bal-
ance, a matrix-free method is used to compute matrix-vector products instead
of storing the global matrix on memory. This algorithm is implemented using
MPI/OpenMP for computation on distributed-shared memory computers.
We enable initial stress analysis and nonlinear wave propagation analysis
successively by changing the right hand side of Eq. (1). The calculation algorithm
for each time step of the elasto-plastic nonlinear wave propagation analysis is
shown in Algorithm 2. Here, the same algorithm is used for both the initial
stress analysis and the wave propagation analysis. In the following, we describe
initial stress analysis and nonlinear wave propagation analysis after initial stress
analysis.
In this study, we use self-weight analysis as initial stress analysis. Gravity is
considered by calculating the external force vector in Eq. (1) as

fn = fn + ρgNdV, (4)
8 A. Yoshiyuki et al.

Algorithm 1. Algorithm for solving Ax = b. The matrix-vector multiplication

Ay is computed using an Element-by-Element method. diag[ ], (¯) and  indicate
the 3 × 3 block Jacobi of [ ], single-precision variable, and tolerance for relative
error, respectively. ( )c indicates the calculation related to FEMmodelc , and the
other is related to calculation of the FEMmodel. ( )in indicates the value in the
inner loop. P̄ is a mapping matrix, from FEMmodelc to FEMmodel, which is
defined by interpolating the displacement in each element of FEMmodelc .
1: set b according to boundary condition
2: x⇐0
3: B̄ ⇐ diag[A]
4: B̄c ⇐ diag[Ac ]
5: r⇐b
6: β⇐0
7: i ⇐1
8: (*outer loop start*)
9: while r2 /b2 ≥  do
10: (*inner loop start*)
11: r̄ ⇐ r
12: z̄ ⇐ B−1 r
13: r̄c ⇐ P̄T r̄
14: z̄c ⇐ P̄T z̄
15: z̄c ⇐ Ā−1c r̄c (*Inner coarse loop: solved on FEMmodelc with c
in
and initial
solution z̄c *)
16: z̄ ⇐ P̄z̄c
17: z̄ ⇐ Ā−1 r̄ (*Inner fine loop: solved on FEMmodel with in and initial solution
z̄*)
18: z ⇐ z̄
19: (*inner loop end*)
20: if i > 1 then
21: β ⇐ (z, q)/ρ
22: end if
23: p ⇐ z + βp
24: q ⇐ Ap
25: ρ ⇐ (z, r)
26: α ⇐ ρ/(p, q)
27: q ⇐ −αq
28: r⇐r+q
29: x ⇐ x + αp
30: i⇐i+1
31: end while
32: (*outer loop end*)

where ρ, g, and N are density, gravitational acceleration and the shape func-
tion, respectively. We apply the Dirichlet boundary condition by fixing vertical
displacement at bottom nodes of the model.
During nonlinear wave propagation analysis, waves are inputted from the
bottom of the model. Thus, instead of using Dirichlet boundary conditions at
 Development of Scalable Three-Dimensional Elasto-Plastic Nonlinear Wave 9

Algorithm 2. Algorithm for elasto-plastic nonlinear wave propagation analysis

in each time step. D, ε, σ and  indicate the constitutive tensor, strain, stress and
tolerance for error, respectively. ( )n (i) indicates the value during i-th iteration
in the n-th time step.
1: calculate Kn , Cn by using Dn
2: calculate δun (1) by solving Eq. (1) taking Eq. (4) and Eq. (5) into account
3: update each value by Eq. (2)
4: i ⇐1
5: δun (0) ⇐ ∞
6: (*iteration start*)
7: while max |δun (i) − δun (i−1) | ≥  do
8: calculate εn (i) by using δun (i)
9: δεn (i) ⇐ εn (i) − εn−1
10: calculate δσ n (i) and Dn (i)
11: re-evaluate Kn , Cn by using Dn (i)
12: re-calculate δun (i+1) by solving Eq. (1)
13: re-update each value by Eq. (2)
14: i ⇐i +1
15: end while
16: (*iteration end*)
17: σ n ⇐ σ n−1 + δσ n (i−1)
18: Dn+1 ⇐ Dn (i−1)

the bottom of the model, we balance gravitational forces by adding reaction force
to the bottom of the model obtained at the last step of initial stress analysis
(step t0 ). Here, the reaction force

− f t0 + qt0 −1 , (5)

is added to the bottom nodes of the model in Eq. (1). Here, f n is calculated as
in Eq. (4).

3 Numerical Experiments
3.1 Verification of Proposed Method
As we cannot obtain analytical solutions for elasto-plastic nonlinear wave propa-
gation analysis, we cannot verify the developed program by comparing numerical
solutions with analytical solutions. However, we can compare 1D numerical anal-
ysis results with the same elasto-plastic constitutive models with 3D numerical
analysis results on a horizontally stratified soil structure to verify the consistency
between the 1D and 3D analyses as well as the numerical convergence with fine
discretization of the analyses. As we use the results of the 1D analysis (stress and
velocity) with the same elasto-plastic models as the boundary condition at base
and side faces of the 3D model for 3D analyses, we can check the consistency
between the 3D and 1D analyses and their numerical convergence by checking
the uniformity of 3D analysis results in the x − y plane.
10 A. Yoshiyuki et al.

(a) Whole view (b) Enlarged view

(c) Ground property. Vp , Vs and hmax are the P-wave velocity,

the S-wave velocity and the maximum damping ratio.

(d) Elasto-plastic property of soft layer

Fig. 2. Horizontally layered model and ground property

We conducted numerical tests on a horizontally stratified ground structure

with soft layer of 10 m thickness on top of bedrock of 40 m thickness. The size
of the 3D model was 0 ≤ x ≤ 16 m, 0 ≤ y ≤ 16 m, 0 ≤ z ≤ 50 m (Fig. 2). The
ground properties of each layer and elasto-plastic parameters of the soft layer are
described in Fig. 2. Here, Ki and K0 are the coefficient of initial earth pressure at
rest and the coefficient of earth pressure at rest, respectively. We used hmax ×0.01
for Rayleigh damping of the soft layer. Following previous studies [8], we chose
element size ds such that it satisfies
Vs
ds ≤ . (6)
χfmax
Here, fmax and χ are the maximum target frequency and the number of ele-
ments per wavelength, respectively. χ is set to χ > 10 for nonlinear layers and
χ > 5 for linear layers for numerical convergence of the solution. Taking the
above conditions into account, we considered two models whose minimum ele-
ment size is 1 m and 2 m, respectively, and the maximum element size is 8 m
in both 1D analysis and 3D analysis. We used the seismic wave observed at
the Kobe Marine Meteorological Observatory during the Great Hanshin Earth-
 Development of Scalable Three-Dimensional Elasto-Plastic Nonlinear Wave 11

(a) Kobe wave (b) Mashiki wave

Fig. 3. Input wave

quake in 1995 (Fig. 3, Kobe wave). We pull back this wave to the bedrock and
input it to the bottom of the 3D model. Since the major components of the
response is influenced by waves below 2.5 Hz, we conduct analysis targeting fre-
quency range between 0.1 and 2.5 Hz. We first conduct self-weight analysis with
dt = 0.001 s × 700,000 time steps, and then conduct nonlinear wave propagation
analysis with dt = 0.001 s × 40,000 time steps using the Kobe wave. Instead of
loading the full gravitational force at the initial step, we increased the gravi-
tational force by 0.000002 times every time step until 500,000 time steps for
both the 1D and 3D analyses. For the 3D analysis, we used the Oakleaf-FX
system at the University of Tokyo consisting of 4,800 computing nodes each
with single 16 core SPARC64 IXfx CPUs (Fujitsu’s PRIMEHPC FX10 mas-
sively parallel supercomputer with a peak performance of 1.13 PFLOPS). For
the model with minimum element size of 1 m, the degrees-of-freedom was 85,839,
and the 3D analysis took 20,619 s using 576 CPU cores (72 MPI processes × 8
OpenMP threads). For the model with minimum element size of 2 m, the degrees-
of-freedom was 14,427, and the 3D analysis took 12,278 s by using 64 CPU cores
(8 MPI processes × 8 OpenMP threads).
Results of the 1D and 3D analyses are shown in Figs. 4 and 5. From Fig. 4, we
can see that the time history of displacement on ground surface for each analysis
are almost identical. Figure 5 shows the displacement distribution at surface
of the 3D analysis. We can see that the difference of displacement values at
each point is converged within about 0.75%. Although not shown, the maximum
difference was about 2% for the case with element size of 2 m. We can see that
the 3D analysis results converge to the 1D analysis results by using sufficiently
small elements (in this case, 1 m elements).
Another random document with
no related content on Scribd:
 ¹⁰And Cush begat Nimrod: he began to be a
mighty one in the earth.
10. And Cush begat Nimrod] From the parallel passages in
Genesis (x. 10, 11) it is apparent that Nimrod is the name of an
individual, the traditional founder of the Babylonian-Assyrian Empire.
As Cush is here called the father of Nimrod and in verse 8 is the son
of Ham, Hebrew tradition would appear to have regarded Hamites as
the founders of the Babylonian power. Possibly the Redactor of
Genesis who combined these verses which belong to the tradition of
J with verses 5‒9 which are from “P” may have thought so. But in the
independent “J” narrative it is very probable that Cush, father of
Nimrod, represents the third or Kassite dynasty (Κοσσαῖοι) which
held sway in Babylon from about 1750‒1200 b.c. Even so, the
identification of Nimrod himself remains a puzzle, and it is not yet
possible to say whether he is a legendary or an historical character,
or partly both.

began to be a mighty one in the earth] i.e. was the first grand
monarch (for the idiom, compare Genesis ix. 20). In Genesis x. 9, he
is further and quaintly described as “a mighty hunter before the
Lord.”

¹¹And Mizraim begat Ludim, and Anamim, and

Lehabim, and Naphtuhim,
11. Ludim] reckoned in Jeremiah xlvi. 9 and Ezekiel xxx. 5
(Revised Version “Lud”) among the auxiliary troops of Egypt
(Mizraim). Probably not the Lydians of Asia Minor are meant, but a
people of north Africa not yet known. Both this word and Lehabim
may be variants for the Libyans, tribes west of Cyrene (compare 2
Chronicles xii. 3, xvi. 8). See also verse 17, note on Lud. Of the
Anamim, Naphtuhim, nothing is certainly known.

¹²and Pathrusim, and Casluhim (from whence

came the Philistines ¹), and Caphtorim.
 ¹ Hebrew Pelishtim.

12. Pathrusim] the inhabitants of Pathros (Isaiah xi. 11), i.e.

Upper Egypt.

Casluhim] not identified.

from whence came the Philistines] Elsewhere (Jeremiah xlvii. 4;

Amos ix. 7; compare Deuteronomy ii. 23) the Philistines are said to
have come from Caphtor. It is natural therefore to think that an
accidental transposition has taken place, and that this clause,
whence ... Philistines, originally followed Caphtorim. Note, however,
that the same order is found in Genesis x. 14.

Caphtorim] i.e. the inhabitants of Caphtor, which has usually

been taken to mean the island of Crete, but is also plausibly
identified with “Keftiu,” the south-west coastlands of Asia Minor.
Compare Macalister, The Philistines, pp. 4 ff.

¹³And Canaan begat Zidon his firstborn, and

Heth;
13. Canaan begat] Of the four sons of Ham—viz. Cush, Mizraim,
Put, Canaan—note that the sons of Put are omitted. After the sons of
Cush (verse 9), and of Mizraim (verse 11), we here pass to the sons
of Canaan.

Zidon his firstborn] From the time of David downwards Tyre takes
precedence of Zidon in any mention of the Phoenician cities in the
Old Testament, but Zidon was the older of the two cities, as is here
implied and as the Roman historian Justin (xviii. 3) asserts. So we
find the Phoenicians in the earlier books of the Old Testament called
Zidonians, not Tyrians (e.g. Judges iii. 3; 1 Kings v. 6). Homer also
refers not to Tyre but to Zidon.
 Heth] i.e. the Hittites, a northern non-Semitic race, who from
about 1800‒700 b.c. were a great power, extending over part of Asia
Minor and northern Syria from the Orontes to the Euphrates. The
references to them in the Old Testament make it probable that Hittite
settlements were to be found in various parts of Palestine. This fact
and their dominant influence, circa 1300 b.c., throughout Canaan
and Phoenicia probably accounts for their inclusion as a “son” of
Canaan.

¹⁴and the Jebusite, and the Amorite, and the

Girgashite;
14. the Jebusite] the ancient population of Jerusalem, compare
Judges i. 21; 2 Samuel v. 6.

the Amorite] compare Numbers xiii. 29, xxi. 21; Judges i. 35. The
name (probably a racial one) was frequently used of the pre-
Israelitish inhabitants of Canaan (“Canaanites” being the
geographical description). In a more restricted sense it was used to
denote the people of Sihon, east of the Jordan.

¹⁵and the Hivite, and the Arkite, and the Sinite;

¹⁶and the Arvadite, and the Zemarite, and the
Hamathite.
15. the Hivite] In Joshua xi. 3, the Hivites are placed in the
extreme north of the land, “the Hivite under Hermon,” but the word
may be an error for Hittite (see above verse 13). In Joshua ix. 7 and
Genesis xxxiv. 2 they are located at Gibeon and Shechem. The
Arkite and Sinite lived in Lebanon, the Arvadite (compare Ezekiel
xxvii. 8) on the sea-coast north of Gebal (Byblus), the Zemarite a
little to the south of the Arvadite, and the Hamathite furthest to the
north on the Orontes.

17 (= Genesis x. 22, 23).

The Sons of Shem.
 ¹⁷The sons of Shem; Elam, and Asshur, and
Arpachshad, and Lud, and Aram, and Uz, and
Hul, and Gether, and Meshech ¹.
¹ In Genesis x. 23, Mash.

17. The sons of Shem] These occupied the middle geographical

“zone.”

Elam] is the name of a land and nation north of the Persian Gulf
and east of Babylonia, and is often referred to in the Old Testament.
Though settled by Semites at a very early date, it was subsequently,
circa 2280 b.c., possessed by a non-Semitic race, who even
extended their power over Babylonia itself. The inclusion of Elam
among the Semites is doubtless due to its proximity to Asshur, and,
though not strictly correct, is very natural.

Asshur] The Assyrians, who are so frequently referred to in the

Old Testament, were mainly, if not entirely Semitic: a martial and
ruthless people whose conquests in the 14th‒7th centuries have
made them world-famous.

Arpachshad] a somewhat obscure name. In the last part (chshad)

the same consonants occur as in the name “Chasdim,” the
“Chaldees” of the Old Testament. Possibly two names have been run
together, the second being that of the Chaldees or Chaldeans, a
Semitic race who from circa 900 b.c. dominated Babylonia,
assimilating with the earlier Semitic inhabitants. This conjecture has
some support in the surprising fact that the Chaldeans are not
otherwise mentioned in the table; it is opposed by the fact that
Arpachshad occurs elsewhere, verse 24; Genesis x. 24, xi. 10 ff.

Lud] the name suggests the Lydians, but how this non-Semitic
people situated on the west coast of Asia Minor comes to be
included with Asshur and Aram as a son of Shem is a mystery.
Possibly therefore a Semitic region, called Lubdu, between Tigris
and Euphrates is meant.

Aram] the “Syrians” of the Authorized Version; better called

Arameans. They were widely settled in the lands to the north and
north-east of Palestine, with important centres in Damascus (Syria
proper) and the north of the Euphrates valley (the Aram-Naharaim of
the Old Testament). So great and lasting was their influence on
Israel that the Aramean dialect eventually superseded Hebrew and
was the ordinary language of Palestine in the time of Christ.

Uz] From Genesis x. 23 it appears that in Chronicles the words

“And the children of Aram ¹” have dropped out, so that “Uz” etc.
appear as the immediate descendants of Shem.

¹ The Alexandrine MS. (A) of the LXX. has the words.

Neither Uz nor the three following names have been satisfactorily

identified. For “Meshech” Genesis x. 23 (Hebrew but not LXX.) reads
“Mash.”

18‒23 (= Genesis x. 24‒29).

Appendix to the Sons of Shem.
South Arabian Tribes.

¹⁸And Arpachshad begat Shelah, and Shelah

begat Eber.
18. Eber] The Hebrew word usually means “the land beyond” and
may have originated as a personification of the population beyond
the Euphrates. It is further possible that Eber is an eponym, not
merely of the Hebrews, but of the Habiri, a much wider stock of
Semitic nomads, of whom the Hebrews formed an element, and who
overran and harassed the settled peoples of Palestine in the fifteenth
century b.c.
 ¹⁹And unto Eber were born two sons: the
name of the one was Peleg; for in his days the
earth was divided; and his brother’s name was
Joktan.
19. two sons] one (Peleg) representing, roughly speaking, the
northern or Mesopotamian Semites; the other (Joktan), the south
Arabian tribes.

Peleg] see below on verse 25.

²⁰And Joktan begat Almodad, and Sheleph,

and Hazarmaveth, and Jerah; ²¹and Hadoram,
and Uzal, and Diklah; ²²and Ebal ¹, and
Abimael, and Sheba; ²³and Ophir, and
Havilah, and Jobab. All these were the sons of
Joktan.
¹ In Genesis x. 28, Obal.

20. Joktan begat Almodad] All the names of the sons of Joktan
here given, so far as they have been identified, represent peoples
situated in south Arabia or on the west coast of the Red Sea lying
over against south Arabia. The only familiar name is that of the
unidentified Ophir, which possibly but not certainly may be the “El
Dorado” to which Solomon sent his fleet for gold.

24‒27.
The Descent of Abraham from Shem.

These verses are compressed within the smallest limits from

Genesis xi. 10‒26. For another example of this extreme abbreviation
compare verses 1‒4 (= Genesis v. 3‒32).
 ²⁴Shem, Arpachshad, Shelah; ²⁵Eber,
Peleg, Reu; ²⁶Serug, Nahor, Terah; ²⁷Abram
(the same is Abraham).
25. Peleg] the name perhaps signifies “Division” (see verse 19),
and may refer to some great period of migration among the Semitic
tribes.

28‒31 (= Genesis xxv. 12‒16).

The Descent of the Ishmaelite Tribes from Abraham

²⁸The sons of Abraham; Isaac, and Ishmael.

²⁹These are their generations: the firstborn
of Ishmael, Nebaioth; then Kedar, and Adbeel,
and Mibsam,
29. Nebaioth] Compare Isaiah lx. 7.

Kedar] Isaiah xxi. 13‒17.

³⁰Mishma, and Dumah, Massa; Hadad, and

Tema,
30. Dumah] Isaiah xxi. 11.

Massa] Proverbs xxxi. 1 (Revised Version margin).

Hadad] The name begins with the Hebrew letter Ḥēth and
therefore differs from the Hadad of verse 46 and of verse 50 and of 2
Chronicles xvi. 2 in which the first letter is Hē, a softer guttural than
Ḥeth.

Tema] Isaiah xxi. 14.

 ³¹Jetur, Naphish, and Kedemah. These are the
sons of Ishmael.
31. Jetur, Naphish] compare v. 18‒22.

32, 33 (= Genesis xxv. 1‒4).

The Descent of Arabian Tribes from Abraham through
Keturah

³²And the sons of Keturah, Abraham’s

concubine: she bare Zimran, and Jokshan,
and Medan, and Midian, and Ishbak, and
Shuah. And the sons of Jokshan; Sheba, and
Dedan.
32. Keturah] called a wife of Abraham in Genesis xxv. 1. The
Chronicler by calling her a concubine may imply that he considered
that the tribes descended from her were not so closely akin to Israel
as the Ishmaelites, or possibly he held that Sarah ought to be the
only wife of Abraham, and “corrects” his source accordingly.

Medan, Midian] Kindred tribes often bore names only slightly

differing in form.

Midian] In Judges viii. 14 the Midianites are reckoned as

Ishmaelites.

Sheba, and Dedan] Sheba and Dedan in verse 9 (which belongs

to the same source P) are included among the Hamitic peoples.
Doubtless the names in the present passage, which comes from J,
refer to the same tribes; but J follows a different tradition as to their
origin. Possibly there is truth in both views, and the people of Sheba
were of mixed African and Arabian descent.
 ³³And the sons of Midian; Ephah, and Epher,
and Hanoch, and Abida, and Eldaah. All these
were the sons of Keturah.
33. Ephah] Isaiah lx. 6.

Hanoch] as Genesis xxv. 4. Compare verse 3.

34‒37 (compare Genesis xxxvi. 10‒14).

The Descent of the Tribes of Edom from Abraham.

³⁴And Abraham begat Isaac. The sons of

Isaac; Esau, and Israel.
³⁵The sons of Esau; Eliphaz, Reuel, and
Jeush, and Jalam, and Korah.
34. Esau] “Esau is Edom,” Genesis xxxvi. 1, 8.

³⁶The sons of Eliphaz; Teman, and Omar,

Zephi ¹, and Gatam, Kenaz, and Timna, and
Amalek. ³⁷The sons of Reuel; Nahath, Zerah,
Shammah, and Mizzah.
¹ In Genesis xxxvi. 11, Zepho.

36. Teman] Amos i. 11, 12; Habakkuk iii. 13. The word means
South, and is applied in the first passage to Edom itself, in the
second to the wilderness of Edom, both being south of Canaan.

Kenaz] Other references (Judges i. 13, iii. 9, 11) show a close

connection with Caleb, which in turn implies that the Calebites were
closely related to the Edomites (compare iv. 13).
 Amalek] the eponymous ancestor of the Amalekites who lived in
the south and south-east of Palestine, see iv. 42 f.

38‒42 (compare Genesis xxxvi. 20‒28).

The Genealogy of the Horite Inhabitants of Seir.

³⁸And the sons of Seir; Lotan and Shobal and

Zibeon and Anah, and Dishon and Ezer and
Dishan. ³⁹And the sons of Lotan; Hori and
Homam ¹: and Timna was Lotan’s sister.
¹ In Genesis xxxvi. 22, Hemam.

38. The sons of Seir] Chronicles omits the further description

given in Genesis “the Horite, the inhabitants of the land,” words
which show clearly that these “sons of Seir” were not descendants of
Esau, but aboriginal inhabitants of the land.

Lotan] perhaps to be connected with Lot, a name anciently

associated with the land or people dwelling east of the Jordan
(compare Genesis xix. 30).

⁴⁰The sons of Shobal; Alian ¹ and Manahath

and Ebal, Shephi ² and Onam. And the sons of
Zibeon; Aiah and Anah. ⁴¹The sons of Anah;
Dishon. And the sons of Dishon; Hamran ³ and
Eshban and Ithran and Cheran. ⁴²The sons of
Ezer; Bilhan and Zaavan, Jaakan ⁴. The sons
of Dishan; Uz and Aran.
¹ In Genesis xxxvi. 23, Alvan.

² In Genesis xxxvi. 23, Shepho.

 ³ In Genesis xxxvi. 26, Hemdan.

⁴ In Genesis xxxvi. 27, and Akan.

40. Aiah and Anah] See Genesis xxxvi. 24.

43‒51a (compare Genesis xxxvi. 31‒39).

The early Kings of Edom.

⁴³Now these are the kings that reigned in the

land of Edom, before there reigned any king
over the children of Israel: Bela the son of
Beor; and the name of his city was Dinhabah.
⁴⁴And Bela died, and Jobab the son of Zerah
of Bozrah reigned in his stead. ⁴⁵And Jobab
died, and Husham of the land of the
Temanites reigned in his stead.
43. kings] Note that the kings are of different families and
localities. They may be compared with the “judges” of early Israel.

in the land of Edom] In early times the mountainous region of

Seir, extending from the south-east of the Dead Sea to the Gulf of
Akaba, but the precise territory of the Edomites is uncertain and of
course must have varied from time to time. In the post-exilic period
Edomites (Idumeans) pressed up into the south of Judah (compare
ii. 42), and Edom (Idumea) continued to play an important and often
sinister part in the history of Israel till long after the Chronicler’s
lifetime. See (e.g.) 1 Maccabees v. 65; 2 Maccabees x. 14‒17. The
Herods were of Edomite descent.

before ... Israel] i.e. before Saul; or possibly “before David,” if the
phrase means before the reign of the first Israelitish king over Edom.
For the use made of this statement in the discussion of the date of
the Hexateuch, see Chapman, Introduction to the Pentateuch, p. 40,
in this series.

Bela the son of Beor] possibly the same as the familiar Balaam
son of Beor, the consonants of the names differing in Hebrew only by
the final m. See, however, Gray, Numbers (International Critical
Commentary), pp. 315, 324.

⁴⁶And Husham died, and Hadad the son of

Bedad, which smote Midian in the field of
Moab, reigned in his stead: and the name of
his city was Avith. ⁴⁷And Hadad died, and
Samlah of Masrekah reigned in his stead.
46. smote Midian in the field of Moab] An isolated historical
notice, interesting as showing the power of Edom at some period.
The Midianites centred round the lands east of the Gulf of Akaba, but
bands of them were constantly pushing northwards and harassing
the territories of Edom, Moab, and Israel (compare Numbers xxii. 4;
Judges vi.; etc.).

⁴⁸And Samlah died, and Shaul of Rehoboth by

the River reigned in his stead. ⁴⁹And Shaul
died, and Baal-hanan the son of Achbor
reigned in his stead.
48. Rehoboth by the River] not “the River,” par excellence (i.e.
the Euphrates), as the Revised Version translators supposed; but
either the Wady el-Arish, the stream on the boundary of Egypt or
Palestine; or else a river in north Edom, Rehoboth being
distinguished from other places of the same name by being the city
on its banks.
 ⁵⁰And Baal-hanan died, and Hadad ¹ reigned in
his stead; and the name of his city was Pai ²:
and his wife’s name was Mehetabel, the
daughter of Matred, the daughter of Me-
zahab.
¹ In Genesis xxxvi. 39, Hadar.

² In Genesis xxxvi. 39, Pau.

50. Hadad] As in verse 46; in Genesis xxxvi. 39, “Hadar.”

Possibly the king whom David overthrew, 2 Samuel viii. 14, compare
1 Kings xi. 14 (perhaps a son of this Hadad).

⁵¹And Hadad died.

51a. And Hadad died] repeated by a copyist’s error from verse
47; the words are not found in Genesis.

51b‒54 (compare Genesis xxxvi. 40‒43).

The “Dukes” of Edom.

And the dukes of Edom were; duke Timna,

duke Aliah ¹, duke Jetheth; ⁵²duke Oholibamah,
duke Elah, duke Pinon; ⁵³duke Kenaz, duke
Teman, duke Mibzar; ⁵⁴duke Magdiel, duke
Iram. These are the dukes of Edom.
¹ In Genesis xxxvi. 40, Alvah.

51b. dukes] The word means “leader of a thousand.” The list

which follows is probably topographical, not chronological. It seems
to give the names of the districts into which Edom was divided at the
time when the list was drawn up.

duke Timna, etc.] Render, the duke of Timna, etc.

Aliah] In Genesis xxxvi. 40, “Alvah.”

 Chapters II.‒VIII.
The Genealogies of the Tribes of Israel.
Attention is now narrowed down to those in the true line of
descent, from Abraham through Isaac (“in Isaac shall thy seed be
called,” Genesis xxi. 12) and from Isaac through Jacob = Israel,
Genesis xxxii. 28 (compare Genesis xxvi. 2‒4).

The Chronicler deals very unequally with the tribes in their

genealogies; as the following table shows:

ii. 1‒iv. 23. Judah (102 verses).

iv. 24‒43. Simeon (20 verses).

v. 1‒26. Reuben, Gad, and Eastern Manasseh (26 verses).

vi. 1‒81. Levi (81 verses).

vii. 1‒40. Issachar, Zebulun, and Dan (according to a

correction of the text, vii. 6‒11, and 12), Naphtali,
Eastern Manasseh (again), Ephraim, and Asher (40
verses).

viii. 1‒40. Benjamin (40 verses).

It may easily be seen that the tribes in which the Chronicler is

really interested are the three southern tribes, Judah, Simeon, and
Benjamin, together with the priestly tribe, Levi.

The order in which the tribes are mentioned is geographical,

Judah and Simeon the southern tribes first, then the eastern tribes,
Reuben, Gad, Manasseh; then (conveniently) Levi, and then the
northern tribes of western Palestine, ending with Benjamin (viii., ix.
35‒44) and the list of inhabitants of Jerusalem (in ix. 1‒34).

Chapter II.
1, 2 (compare Genesis xxxv. 22b‒26).
The Sons of Israel.

¹These are the sons of Israel; Reuben,

Simeon, Levi, and Judah, Issachar and
Zebulun; ²Dan, Joseph and Benjamin,
Naphtali, Gad and Asher.
II. 3‒IV. 23.
Genealogies of Judah.

3‒17.
Descendants of Judah to the Sons of Jesse.

³The sons of Judah; Er, and Onan, and

Shelah: which three were born unto him of
Bath-shua the Canaanitess. And Er, Judah’s
firstborn, was wicked in the sight of the Lord;
and he slew him. ⁴And Tamar his daughter in
law bare him Perez and Zerah. All the sons of
Judah were five. ⁵The sons of Perez; Hezron
and Hamul.
 5. The sons of Perez; Hezron and Hamul] So Genesis xlvi. 12.
The only reference in the Old Testament to Hamulites is Numbers
xxvi. 21. On the other hand Hezron, a south Judean tribe (Joshua xv.
3), is a clan of the first importance in the genealogies. From Hezron
are descended not only the family of David (verse 15), but also the
great Calebite and Jerahmeelite clans (verses 18‒24, 25‒33, etc.).
The name Hezron might bear the significance “an enclosed place” as
opposed to movable encampments, and Atarah (verse 26) who is
said to be the mother of certain Jerahmeelite families has much the
same meaning. Both names therefore may not be eponymous either
of individuals or places, but may originate in the desire to preserve
the fact that the families named as their sons were nomads who had
abandoned wandering for settled life. If so, it might help to explain
the fact that Hezron (compare Carmi, ii. 7, iv. 1, v. 3) is also
mentioned as a son of Reuben (v. 3; Genesis xlvi. 9, etc.).

⁶And the sons of Zerah; Zimri ¹, and Ethan, and

Heman, and Calcol, and Dara ²: five of them in
all.
¹ In Joshua vii. 1, Zabdi.

² Many ancient authorities read, Darda. See 1 Kings iv. 31.

6. the sons of Zerah] This genealogy appears only in Chronicles.

Zimri] LXX. (B) Ζαμβρεί (β being merely euphonic) here and also
Joshua vii. 1 where Hebrew has “Zabdi.” LXX. is probably right in
identifying the two. Either form might arise from the other by easy
textual corruption.

Ethan ... Dara] Read, Darda with Vulgate, Targum, Peshitṭa The
same four names in the same order occur 1 Kings iv. 31 as the
names of wise men whom Solomon surpassed in wisdom. They are
there called sons of “Mahol” who may have been either a nearer or
remoter ancestor than Zerah. Ethan however is there called the
Ezrahite (= probably “son of Zerah”). [Psalms lxxxviii., lxxxix. bear
respectively the names “Heman the Ezrahite,” “Ethan the Ezrahite,”
but these (it seems) were Levites (compare xv. 17, 19, where see
note).]

⁷And the sons of Carmi; Achar ¹, the troubler of

Israel, who committed a trespass in the
devoted thing. ⁸And the sons of Ethan;
Azariah.
¹ In Joshua vii. 1, Achan.

7. the sons of Carmi] Carmi is probably to be taken as the son of

Zimri (= Zabdi, Joshua vii. 1). Targum however has “Carmi who is
Zimri.” See note on Zimri, verse 6.

Achar] This form of the name (instead of “Achan,” Joshua vii. 1)

is used by the Chronicler to bring out better the play on the Hebrew
word for “troubler.” The Hebrew runs, “Achar ocher Israel.”

⁹The sons also of Hezron, that were born unto

him; Jerahmeel, and Ram, and Chelubai.
9. Jerahmeel] For his descendants see verses 25‒41. The
descendants of his younger brother Ram are given first. They
purport to be the ancestry of David and his family.

Chelubai] Another form of “Caleb”; see note on verse 42.

¹⁰And Ram begat Amminadab; and

Amminadab begat Nahshon, prince of the
children of Judah; ¹¹and Nahshon begat
Salma, and Salma begat Boaz; ¹²and Boaz
begat Obed, and Obed begat Jesse;
 10. Ram] The descent of David from Judah is given also in Ruth
iv. 18‒22 and Matthew i. 3‒6. Ram as a clan parallel with the great
clans of Caleb and Jerahmeel is strange; for it is not known
otherwise in the Old Testament Further, as the descendants of Ram
given in verses 10‒12 are the family tree of David (contrast the clans
and cities in the lines of Caleb and Jerahmeel) it may be supposed
that Ram owes his position here simply to the Chronicler’s desire to
incorporate Ruth iv. 19, where also this pedigree of David is given.
Note also that in verse 25 a Ram is mentioned as a son of
Jerahmeel and grandson of Hezron.

Nahshon, prince, etc.] See Numbers i. 4, 7, ii. 3.

¹³and Jesse begat his firstborn Eliab, and

Abinadab the second, and Shimea the third;
13. Shimea] so also xx. 7; but “Shammah” 1 Samuel xvi. 9.

¹⁴Nethanel the fourth, Raddai the fifth;

14. Nethanel] the same name as Nathanael (John i. 45). The
fourth, fifth and sixth brothers are not elsewhere named.

¹⁵Ozem the sixth, David the seventh:

15. David the seventh] Jesse had eight sons (1 Samuel xvii. 12;
compare xvi. 10, 11). Here one seems deliberately passed over,
perhaps because he had no children. (The Elihu “one of David’s
brethren” of 1 Chronicles xxvii. 18 is probably to be identified with
Eliab and not to be regarded as an eighth brother.)

¹⁶and their sisters were Zeruiah and Abigail.

And the sons of Zeruiah; Abishai ¹, and Joab,
and Asahel, three.
 ¹ Hebrew Abshai.

16. sons of Zeruiah] Joab and his brothers are always thus
named after their mother; perhaps their father died while they were
young, or we may have a relic here of the ancient method of tracing
kinship through the mother.

¹⁷And Abigail bare Amasa: and the father of

Amasa was Jether the Ishmaelite.
17. the Ishmaelite] 2 Samuel xvii. 25, “the Israelite,” an error
yielding no satisfactory sense.

18‒24 (compare verses 42‒55.)

Descendants of Caleb.

¹⁸And Caleb the son of Hezron begat children

of Azubah his wife, and of Jerioth: and these
were her sons; Jesher, and Shobab, and
Ardon. ¹⁹And Azubah died, and Caleb took
unto him Ephrath, which bare him Hur.
18. Caleb] a clan dwelling in southern Judea, and probably
distinct from Judah in the time of David (1 Samuel xxv. 3, xxx. 14).
Other references to them or rather their reputed founder Caleb ben
Jephunneh the Kenizzite (Numbers xxxii. 12; Joshua xiv. 6, 14; 1
Chronicles i. 36, where see note on Kenaz) point to an original
connection with the Edomites. Their importance in these lists is
explained by the fact that they were incorporated in Judah, and, after
the exile, occupied townships close to Jerusalem (verses 50‒55)
“forming possibly the bulk of the tribe in post-exilic Judah, since the
Chronicler knows so few other families” (Curtis, Chronicles p. 89).
See also W. R. Smith, Old Testament in the Jewish Church, p. 279
ad fin.