Download the full version of the textbook now at textbookfull.

com

Database and Expert Systems Applications 29th

International Conference DEXA 2018 Regensburg
Germany September 3 6 2018 Proceedings Part
II Sven Hartmann
https://textbookfull.com/product/database-and-
expert-systems-applications-29th-international-
conference-dexa-2018-regensburg-germany-
september-3-6-2018-proceedings-part-ii-sven-
hartmann/

Explore and download more textbook at https://textbookfull.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Big Data Analytics and Knowledge Discovery 20th

International Conference DaWaK 2018 Regensburg Germany
September 3 6 2018 Proceedings Carlos Ordonez
https://textbookfull.com/product/big-data-analytics-and-knowledge-
discovery-20th-international-conference-dawak-2018-regensburg-germany-
september-3-6-2018-proceedings-carlos-ordonez/
textbookfull.com

Electronic Government and the Information Systems

Perspective: 7th International Conference, EGOVIS 2018,
Regensburg, Germany, September 3–5, 2018, Proceedings
Andrea K■
https://textbookfull.com/product/electronic-government-and-the-
information-systems-perspective-7th-international-conference-
egovis-2018-regensburg-germany-september-3-5-2018-proceedings-andrea-
ko/
textbookfull.com

Computer Vision ECCV 2018 Workshops Munich Germany

September 8 14 2018 Proceedings Part II Laura Leal-Taixé

https://textbookfull.com/product/computer-vision-eccv-2018-workshops-
munich-germany-september-8-14-2018-proceedings-part-ii-laura-leal-
taixe/
textbookfull.com

Stochastic Global Optimization Methods and Applications to

Chemical, Biochemical, Pharmaceutical and Environmental
Processes 1st Edition Ch. Venkateswarlu Satya Eswari
Jujjavarapu
https://textbookfull.com/product/stochastic-global-optimization-
methods-and-applications-to-chemical-biochemical-pharmaceutical-and-
environmental-processes-1st-edition-ch-venkateswarlu-satya-eswari-
jujjavarapu/
textbookfull.com
Payment 1st Edition Olivia Ashers

https://textbookfull.com/product/payment-1st-edition-olivia-ashers/

textbookfull.com

Management Control with Integrated Planning Models and

Implementation for Sustainable Coordination Lukas Rieder

https://textbookfull.com/product/management-control-with-integrated-
planning-models-and-implementation-for-sustainable-coordination-lukas-
rieder/
textbookfull.com

Platform Chemical Biorefinery. Future Green Industry 1st

Edition Satinder Kaur Brar

https://textbookfull.com/product/platform-chemical-biorefinery-future-
green-industry-1st-edition-satinder-kaur-brar/

textbookfull.com

Transforming Negative Reactions to Clients From

Frustration to Compassion 1st Edition Abraham W. Wolf

https://textbookfull.com/product/transforming-negative-reactions-to-
clients-from-frustration-to-compassion-1st-edition-abraham-w-wolf/

textbookfull.com

Advances in Materials Sciences Energy Technology and

Environmental Engineering Proceedings of the International
Conference on Materials Science MSETEE 2016 Zhuhai China
May 28 29 2016 1st Edition Aragona Patty
https://textbookfull.com/product/advances-in-materials-sciences-
energy-technology-and-environmental-engineering-proceedings-of-the-
international-conference-on-materials-science-msetee-2016-zhuhai-
china-may-28-29-2016-1st-edition-arag/
textbookfull.com
The invisible hand? : how market economies have emerged
and declined since AD 500 1st Edition Bas Van Bavel

https://textbookfull.com/product/the-invisible-hand-how-market-
economies-have-emerged-and-declined-since-ad-500-1st-edition-bas-van-
bavel/
textbookfull.com
 Sven Hartmann · Hui Ma
Abdelkader Hameurlain
Günther Pernul
Roland R. Wagner (Eds.)
LNCS 11030

Database and Expert

Systems Applications
29th International Conference, DEXA 2018
Regensburg, Germany, September 3–6, 2018
Proceedings, Part II

123
Lecture Notes in Computer Science 11030
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/7409
Sven Hartmann Hui Ma
•

Abdelkader Hameurlain Günther Pernul

•

Roland R. Wagner (Eds.)

Database and Expert

Systems Applications
29th International Conference, DEXA 2018
Regensburg, Germany, September 3–6, 2018
Proceedings, Part II

123
Editors
Sven Hartmann Günther Pernul
Clausthal University of Technology University of Regensburg
Clausthal-Zellerfeld Regensburg
Germany Germany
Hui Ma Roland R. Wagner
Victoria University of Wellington Johannes Kepler University
Wellington Linz
New Zealand Austria
Abdelkader Hameurlain
Paul Sabatier University
Toulouse
France

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-319-98811-5 ISBN 978-3-319-98812-2 (eBook)
https://doi.org/10.1007/978-3-319-98812-2

Library of Congress Control Number: 2018950775

LNCS Sublibrary: SL3 – Information Systems and Applications, incl. Internet/Web, and HCI

© Springer Nature Switzerland AG 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.

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

This volume contains the papers presented at the 29th International Conference on
Database and Expert Systems Applications (DEXA 2018), which was held in
Regensburg, Germany, during September 3–6, 2018. On behalf of the Program
Committee, we commend these papers to you and hope you find them useful.
Database, information, and knowledge systems have always been a core subject of
computer science. The ever-increasing need to distribute, exchange, and integrate data,
information, and knowledge has added further importance to this subject. Advances in
the field will help facilitate new avenues of communication, to proliferate interdisci-
plinary discovery, and to drive innovation and commercial opportunity.
DEXA is an international conference series that showcases state-of-the-art research
activities in database, information, and knowledge systems. The conference and its
associated workshops provide a premier annual forum to present original research
results and to examine advanced applications in the field. The goal is to bring together
developers, scientists, and users to extensively discuss requirements, challenges, and
solutions in database, information, and knowledge systems.
DEXA 2018 solicited original contributions dealing with any aspect of database,
information, and knowledge systems. Suggested topics included, but were not limited
to:
– Acquisition, Modeling, Management, and Processing of Knowledge
– Authenticity, Privacy, Security, and Trust
– Availability, Reliability, and Fault Tolerance
– Big Data Management and Analytics
– Consistency, Integrity, Quality of Data
– Constraint Modeling and Processing
– Cloud Computing and Database-as-a-Service
– Database Federation and Integration, Interoperability, Multi-Databases
– Data and Information Networks
– Data and Information Semantics
– Data Integration, Metadata Management, and Interoperability
– Data Structures and Data Management Algorithms
– Database and Information System Architecture and Performance
– Data Streams and Sensor Data
– Data Warehousing
– Decision Support Systems and Their Applications
– Dependability, Reliability, and Fault Tolerance
– Digital Libraries and Multimedia Databases
– Distributed, Parallel, P2P, Grid, and Cloud Databases
– Graph Databases
– Incomplete and Uncertain Data
– Information Retrieval
VI Preface

– Information and Database Systems and Their Applications

– Mobile, Pervasive, and Ubiquitous Data
– Modeling, Automation, and Optimization of Processes
– NoSQL and NewSQL Databases
– Object, Object-Relational, and Deductive Databases
– Provenance of Data and Information
– Semantic Web and Ontologies
– Social Networks, Social Web, Graph, and Personal Information Management
– Statistical and Scientific Databases
– Temporal, Spatial, and High-Dimensional Databases
– Query Processing and Transaction Management
– User Interfaces to Databases and Information Systems
– Visual Data Analytics, Data Mining, and Knowledge Discovery
– WWW and Databases, Web Services
– Workflow Management and Databases
– XML and Semi-structured Data
Following the call for papers, which yielded 160 submissions, there was a rigorous
review process that saw each submission refereed by three to six international experts.
The 35 submissions judged best by the Program Committee were accepted as full
research papers, yielding an acceptance rate of 22%. A further 40 submissions were
accepted as short research papers.
As is the tradition of DEXA, all accepted papers are published by Springer. Authors
of selected papers presented at the conference were invited to submit substantially
extended versions of their conference papers for publication in the Springer journal
Transactions on Large-Scale Data- and Knowledge-Centered Systems (TLDKS). The
submitted extended versions underwent a further review process.
The success of DEXA 2018 was the result of collegial teamwork from many
individuals. We wish to thank all authors who submitted papers and all conference
participants for the fruitful discussions.
We are grateful to Xiaofang Zhou (The University of Queensland) for his keynote
talk on “Spatial Trajectory Analytics: Past, Present, and Future” and to Tok Wang Ling
(National University of Singapore) for his keynote talk on “Data Models Revisited:
Improving the Quality of Database Schema Design, Integration and Keyword Search
with ORA-Semantics.”
This edition of DEXA also featured three international workshops covering a variety
of specialized topics:
– BDMICS 2018: Third International Workshop on Big Data Management in Cloud
Systems
– BIOKDD 2018: 9th International Workshop on Biological Knowledge Discovery
from Data
– TIR 2018: 15th International Workshop on Technologies for Information Retrieval
We would like to thank the members of the Program Committee and the external
reviewers for their timely expertise in carefully reviewing the submissions. We are
grateful to our general chairs, Abdelkader Hameurlain, Günther Pernul, and
 Preface VII

Roland R. Wagner, to our publication chair, Vladimir Marik, and to our workshop
chairs, A Min Tjoa and Roland R. Wagner.
We wish to express our deep appreciation to Gabriela Wagner of the DEXA con-
ference organization office. Without her outstanding work and excellent support, this
volume would not have seen the light of day.
Finally, we like to thank Günther Pernul and his team for being our hosts during the
wonderful days in Regensburg.

July 2018 Sven Hartmann

Hui Ma
 Organization

General Chairs
Abdelkader Hameurlain IRIT, Paul Sabatier University, Toulouse, France
Günther Pernul University of Regensburg, Germany
Roland R. Wagner Johannes Kepler University Linz, Austria

Program Committee Chairs

Hui Ma Victoria University of Wellington, New Zealand
Sven Hartmann Clausthal University of Technology, Germany

Publication Chair
Vladimir Marik Czech Technical University, Czech Republic

Program Committee
Slim Abdennadher German University, Cairo, Egypt
Hamideh Afsarmanesh University of Amsterdam, The Netherlands
Riccardo Albertoni Institute of Applied Mathematics and Information
Technologies - Italian National Council of Research,
Italy
Idir Amine Amarouche University Houari Boumediene, Algeria
Rachid Anane Coventry University, UK
Annalisa Appice Università degli Studi di Bari, Italy
Mustafa Atay Winston-Salem State University, USA
Faten Atigui CNAM, France
Spiridon Bakiras Hamad bin Khalifa University, Qatar
Zhifeng Bao National University of Singapore, Singapore
Ladjel Bellatreche ENSMA, France
Nadia Bennani INSA Lyon, France
Karim Benouaret Université Claude Bernard Lyon 1, France
Benslimane Djamal Lyon 1 University, France
Morad Benyoucef University of Ottawa, Canada
Catherine Berrut Grenoble University, France
Athman Bouguettaya University of Sydney, Australia
Omar Boussaid University of Lyon/Lyon 2, France
Stephane Bressan National University of Singapore, Singapore
Barbara Catania DISI, University of Genoa, Italy
Michelangelo Ceci University of Bari, Italy
Richard Chbeir UPPA University, France
X Organization

Cindy Chen University of Massachusetts Lowell, USA

Phoebe Chen La Trobe University, Australia
Max Chevalier IRIT - SIG, Université de Toulouse, France
Byron Choi Hong Kong Baptist University, Hong Kong, SAR
China
Soon Ae Chun City University of New York, USA
Deborah Dahl Conversational Technologies, USA
Jérôme Darmont Université de Lyon (ERIC Lyon 2), France
Roberto De Virgilio Università Roma Tre, Italy
Vincenzo Deufemia Università degli Studi di Salerno, Italy
Gayo Diallo Bordeaux University, France
Juliette Dibie-Barthélemy AgroParisTech, France
Dejing Dou University of Oregon, USA
Cedric du Mouza CNAM, France
Johann Eder University of Klagenfurt, Austria
Suzanne Embury The University of Manchester, UK
Markus Endres University of Augsburg, Germany
Noura Faci Lyon 1 University, France
Bettina Fazzinga ICAR-CNR, Italy
Leonidas Fegaras The University of Texas at Arlington, USA
Stefano Ferilli University of Bari, Italy
Flavio Ferrarotti Software Competence Center Hagenberg, Austria
Vladimir Fomichov School of Business Informatics, National Research
University Higher School of Economics, Moscow,
Russian Federation
Flavius Frasincar Erasmus University Rotterdam, The Netherlands
Bernhard Freudenthaler Software Competence Center Hagenberg, Austria
Hiroaki Fukuda Shibaura Institute of Technology, Japan
Steven Furnell Plymouth University, UK
Joy Garfield University of Worcester, UK
Claudio Gennaro ISTI-CNR, Italy
Manolis Gergatsoulis Ionian University, Greece
Javad Ghofrani Leibniz Universität Hannover, Germany
Fabio Grandi University of Bologna, Italy
Carmine Gravino University of Salerno, Italy
Sven Groppe Lübeck University, Germany
Jerzy Grzymala-Busse University of Kansas, USA
Francesco Guerra Università degli Studi di Modena e Reggio Emilia, Italy
Giovanna Guerrini University of Genoa, Italy
Allel Hadjali ENSMA, Poitiers, France
Abdelkader Hameurlain Paul Sabatier University, France
Ibrahim Hamidah Universiti Putra Malaysia, Malaysia
Takahiro Hara Osaka University, Japan
Sven Hartmann Clausthal University of Technology, Germany
Wynne Hsu National University of Singapore, Singapore
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
 Organization XI

Yu Hua Huazhong University of Science and Technology,

China
San-Yih Hwang National Sun Yat-Sen University, Taiwan
Theo Härder TU Kaiserslautern, Germany
Ionut Emil Iacob Georgia Southern University, USA
Sergio Ilarri University of Zaragoza, Spain
Abdessamad Imine Inria Grand Nancy, France
Yasunori Ishihara Nanzan University, Japan
Peiquan Jin University of Science and Technology of China, China
Anne Kao Boeing, USA
Dimitris Karagiannis University of Vienna, Austria
Stefan Katzenbeisser Technische Universität Darmstadt, Germany
Anne Kayem Hasso Plattner Institute, Germany
Carsten Kleiner University of Applied Sciences and Arts Hannover,
Germany
Henning Koehler Massey University, New Zealand
Harald Kosch University of Passau, Germany
Michal Krátký Technical University of Ostrava, Czech Republic
Petr Kremen Czech Technical University in Prague, Czech Republic
Sachin Kulkarni Macquarie Global Services, USA
Josef Küng University of Linz, Austria
Gianfranco Lamperti University of Brescia, Italy
Anne Laurent LIRMM, University of Montpellier 2, France
Lenka Lhotska Czech Technical University, Czech Republic
Yuchen Li Singapore Management University, Singapore
Wenxin Liang Dalian University of Technology, China
Tok Wang Ling National University of Singapore, Singapore
Sebastian Link The University of Auckland, New Zealand
Chuan-Ming Liu National Taipei University of Technology, Taiwan
Hong-Cheu Liu University of South Australia, Australia
Jorge Lloret Gazo University of Zaragoza, Spain
Alessandra Lumini University of Bologna, Italy
Hui Ma Victoria University of Wellington, New Zealand
Qiang Ma Kyoto University, Japan
Stephane Maag TELECOM SudParis, France
Zakaria Maamar Zayed University, United Arab Emirates
Elio Masciari ICAR-CNR, Università della Calabria, Italy
Brahim Medjahed University of Michigan - Dearborn, USA
Harekrishna Mishra Institute of Rural Management Anand, India
Lars Moench University of Hagen, Germany
Riad Mokadem IRIT, Paul Sabatier University, France
Yang-Sae Moon Kangwon National University, South Korea
Franck Morvan IRIT, Paul Sabatier University, France
Dariusz Mrozek Silesian University of Technology, Poland
Francesc Munoz-Escoi Universitat Politecnica de Valencia, Spain
Ismael Navas-Delgado University of Málaga, Spain
XII Organization

Wilfred Ng Hong Kong University of Science and Technology,

Hong Kong, SAR China
Javier Nieves Acedo IK4-Azterlan, Spain
Mourad Oussalah University of Nantes, France
George Pallis University of Cyprus, Cyprus
Ingrid Pappel Tallinn University of Technology, Estonia
Marcin Paprzycki Polish Academy of Sciences, Warsaw Management
Academy, Poland
Oscar Pastor Lopez Universitat Politecnica de Valencia, Spain
Francesco Piccialli University of Naples Federico II, Italy
Clara Pizzuti Institute for High Performance Computing and
Networking (ICAR)-National Research Council
(CNR), Italy
Pascal Poncelet LIRMM, France
Elaheh Pourabbas National Research Council, Italy
Claudia Raibulet Università degli Studi di Milano-Bicocca, Italy
Praveen Rao University of Missouri-Kansas City, USA
Rodolfo Resende Federal University of Minas Gerais, Brazil
Claudia Roncancio Grenoble University/LIG, France
Massimo Ruffolo ICAR-CNR, Italy
Simonas Saltenis Aalborg University, Denmark
N. L. Sarda I.I.T. Bombay, India
Marinette Savonnet University of Burgundy, France
Florence Sedes IRIT, Paul Sabatier University, Toulouse, France
Nazha Selmaoui University of New Caledonia, New Caledonia
Michael Sheng Macquarie University, Australia
Patrick Siarry Université Paris 12 (LiSSi), France
Gheorghe Cosmin Silaghi Babes-Bolyai University of Cluj-Napoca, Romania
Hala Skaf-Molli Nantes University, France
Bala Srinivasan Retried, Monash University, Australia
Umberto Straccia ISTI - CNR, Italy
Maguelonne Teisseire Irstea - TETIS, France
Sergio Tessaris Free University of Bozen-Bolzano, Italy
Olivier Teste IRIT, University of Toulouse, France
Stephanie Teufel University of Fribourg, Switzerland
Jukka Teuhola University of Turku, Finland
Jean-Marc Thevenin University of Toulouse 1 Capitole, France
A Min Tjoa Vienna University of Technology, Austria
Vicenc Torra University of Skövde, Sweden
Traian Marius Truta Northern Kentucky University, USA
Theodoros Tzouramanis University of the Aegean, Greece
Lucia Vaira University of Salento, Italy
Ismini Vasileiou University of Plymouth, UK
Krishnamurthy Vidyasankar Memorial University of Newfoundland, Canada
Marco Vieira University of Coimbra, Portugal
Junhu Wang Griffith University, Brisbane, Australia
 Organization XIII

Wendy Hui Wang Stevens Institute of Technology, USA

Piotr Wisniewski Nicolaus Copernicus University, Poland
Ming Hour Yang Chung Yuan Christian University, Taiwan
Yang, Xiaochun Northeastern University, China
Yanchang Zhao CSIRO, Australia
Qiang Zhu The University of Michigan, USA
Marcin Zimniak Leipzig University, Germany
Ester Zumpano University of Calabria, Italy

Additional Reviewers
Valentyna Tsap Tallinn University of Technology, Estonia
Liliana Ibanescu AgroParisTech, France
Cyril Labbé Université Grenoble-Alpes, France
Zouhaier Brahmia University of Sfax, Tunisia
Dunren Che Southern Illinois University, USA
Feng George Yu Youngstown State University, USA
Gang Qian University of Central Oklahoma, USA
Lubomir Stanchev Cal Poly, USA
Jorge Martinez-Gil Software Competence Center Hagenberg, Austria
Loredana Caruccio University of Salerno, Italy
Valentina Indelli Pisano University of Salerno, Italy
Jorge Bernardino Polytechnic Institute of Coimbra, Portugal
Bruno Cabral University of Coimbra, Portugal
Paulo Nunes Polytechnic Institute of Guarda, Portugal
William Ferng Boeing, USA
Amin Mesmoudi LIAS/University of Poitiers, France
Sabeur Aridhi LORIA, University of Lorraine - TELECOM Nancy,
France
Julius Köpke Alpen Adria Universität Klagenfurt, Austria
Marco Franceschetti Alpen Adria Universität Klagenfurt, Austria
Meriem Laifa Bordj-Bouarreridj University, Algeria
Sheik Mohammad University of Sydney, Australia
Mostakim Fattah
Mohammed Nasser University of Sydney, Australia
Mohammed Ba-hutair
Ali Hamdi Fergani Ali University of Sydney, Australia
Masoud Salehpour University of Sydney, Australia
Adnan Mahmood Macquarie University, Australia
Wei Emma Zhang Macquarie University, Australia
Zawar Hussain Macquarie University, Australia
Hui Luo RMIT University, Australia
Sheng Wang RMIT University, Australia
Lucile Sautot AgroParisTech, France
Jacques Fize Cirad, Irstea, France
XIV Organization

María del Carmen Technological Institute of Aragón, Spain

Rodríguez-Hernández
Ramón Hermoso University of Zaragoza, Spain
Senen Gonzalez Software Competence Center Hagenberg, Austria
Ermelinda Oro High Performance and Computing Institute of the
National Research Council (ICAR-CNR), Italy
Shaoyi Yin Paul Sabatier University, France
Jannai Tokotoko ISEA University of New Caledonia, New Caledonia
Xiaotian Hao Hong Kong University of Science and Technology,
Hong Kong, SAR China
Ji Cheng Hong Kong University of Science & Technology,
Hong Kong, China
Radim Bača Technical University of Ostrava, Czech Republic
Petr Lukáš Technical University of Ostrava, Czech Republic
Peter Chovanec Technical University of Ostrava, Czech Republic
Galicia Auyon Jorge ISAE-ENSMA, Poitiers, France
Armando
Nabila Berkani ESI, Algiers, Algeria
Amine Roukh Mostaganem University, Algeria
Chourouk Belheouane USTHB, Algiers, Algeria
Angelo Impedovo University of Bari, Italy
Emanuele Pio Barracchia University of Bari, Italy
Arpita Chatterjee Georgia Southern University, USA
Stephen Carden Georgia Southern University, USA
Tharanga Wickramarachchi U.S. Bank, USA
Divine Wanduku Georgia Southern University, USA
Lama Saeeda Czech Technical University in Prague, Czech Republic
Michal Med Czech Technical University in Prague, Czech Republic
Franck Ravat Université Toulouse 1 Capitole - IRIT, France
Julien Aligon Université Toulouse 1 Capitole - IRIT, France
Matthew Damigos Ionian University, Greece
Eleftherios Kalogeros Ionian University, Greece
Srini Bhagavan IBM, USA
Monica Senapati University of Missouri-Kansas City, USA
Khulud Alsultan University of Missouri-Kansas City, USA
Anas Katib University of Missouri-Kansas City, USA
Jose Alvarez Telecom SudParis, France
Sarah Dahab Telecom SudParis, France
Dietrich Steinmetz Clausthal University of Technology, Germany
 Contents – Part II

Information Retrieval

Template Trees: Extracting Actionable Information from

Machine Generated Emails . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Manoj K. Agarwal and Jitendra Singh

Parameter Free Mixed-Type Density-Based Clustering . . . . . . . . . . . . . . . . . 19

Sahar Behzadi, Mahmoud Abdelmottaleb Ibrahim,
and Claudia Plant

CROP: An Efficient Cross-Platform Event Popularity Prediction Model

for Online Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Mingding Liao, Xiaofeng Gao, Xuezheng Peng, and Guihai Chen

Probabilistic Classification of Skeleton Sequences . . . . . . . . . . . . . . . . . . . . 50

Jan Sedmidubsky and Pavel Zezula

Uncertain Information

A Fuzzy Unified Framework for Imprecise Knowledge . . . . . . . . . . . . . . . . 69

Soumaya Moussa and Saoussen Bel Hadj Kacem

Frequent Itemset Mining on Correlated Probabilistic Databases . . . . . . . . . . . 84

Yasemin Asan Kalaz and Rajeev Raman

Leveraging Data Relationships to Resolve Conflicts from Disparate

Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Romila Pradhan, Walid G. Aref, and Sunil Prabhakar

Data Warehouses and Recommender Systems

Direct Conversion of Early Information to Multi-dimensional Model . . . . . . . 119

Deepika Prakash

OLAP Queries Context-Aware Recommender System . . . . . . . . . . . . . . . . . 127

Elsa Negre, Franck Ravat, and Olivier Teste

Combining Web and Enterprise Data for Lightweight Data

Mart Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Suzanne McCarthy, Andrew McCarren, and Mark Roantree
XVI Contents – Part II

FairGRecs: Fair Group Recommendations by Exploiting Personal

Health Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Maria Stratigi, Haridimos Kondylakis, and Kostas Stefanidis

Data Streams

Big Log Data Stream Processing: Adapting an Anomaly Detection

Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Marietheres Dietz and Günther Pernul

Information Filtering Method for Twitter Streaming Data Using

Human-in-the-Loop Machine Learning. . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Yu Suzuki and Satoshi Nakamura

Parallel n-of-N Skyline Queries over Uncertain Data Streams . . . . . . . . . . . . 176

Jun Liu, Xiaoyong Li, Kaijun Ren, Junqiang Song,
and Zongshuo Zhang

A Recommender System with Advanced Time Series Medical Data

Analysis for Diabetes Patients in a Telehealth Environment . . . . . . . . . . . . . 185
Raid Lafta, Ji Zhang, Xiaohui Tao, Jerry Chun-Wei Lin, Fulong Chen,
Yonglong Luo, and Xiaoyao Zheng

Information Networks and Algorithms

Edit Distance Based Similarity Search of Heterogeneous Information

Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Jianhua Lu, Ningyun Lu, Sipei Ma, and Baili Zhang

An Approximate Nearest Neighbor Search Algorithm Using

Distance-Based Hashing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Yuri Itotani, Shin’ichi Wakabayashi, Shinobu Nagayama,
and Masato Inagi

Approximate Set Similarity Join Using Many-Core Processors . . . . . . . . . . . 214

Kenta Sugano, Toshiyuki Amagasa, and Hiroyuki Kitagawa

Mining Graph Pattern Association Rules . . . . . . . . . . . . . . . . . . . . . . . . . . 223

Xin Wang and Yang Xu

Database System Architecture and Performance

Cost Effective Load-Balancing Approach for Range-Partitioned

Main-Memory Resident Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Djahida Belayadi, Khaled-Walid Hidouci, Ladjel Bellatreche,
and Carlos Ordonez
 Contents – Part II XVII

Adaptive Workload-Based Partitioning and Replication for RDF Graphs . . . . 250

Ahmed Al-Ghezi and Lena Wiese

QUIOW: A Keyword-Based Query Processing Tool for RDF Datasets

and Relational Databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Yenier T. Izquierdo, Grettel M. García, Elisa S. Menendez,
Marco A. Casanova, Frederic Dartayre, and Carlos H. Levy

An Abstract Machine for Push Bottom-Up Evaluation of Datalog . . . . . . . . . 270

Stefan Brass and Mario Wenzel

Novel Database Solutions

What Lies Beyond Structured Data? A Comparison Study for Metric

Data Storage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Pedro H. B. Siqueira, Paulo H. Oliveira, Marcos V. N. Bedo,
and Daniel S. Kaster

A Native Operator for Process Discovery . . . . . . . . . . . . . . . . . . . . . . . . . . 292

Alifah Syamsiyah, Boudewijn F. van Dongen, and Remco M. Dijkman

Implementation of the Aggregated R-Tree for Phase Change Memory . . . . . . 301

Maciej Jurga and Wojciech Macyna

Modeling Query Energy Costs in Analytical Database Systems with

Processor Speed Scaling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
Boming Luo, Yuto Hayamizu, Kazuo Goda, and Masaru Kitsuregawa

Graph Querying and Databases

Sprouter: Dynamic Graph Processing over Data Streams at Scale . . . . . . . . . 321

Tariq Abughofa and Farhana Zulkernine

A Hybrid Approach of Subgraph Isomorphism and Graph Simulation

for Graph Pattern Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Kazunori Sugawara and Nobutaka Suzuki

Time Complexity and Parallel Speedup of Relational Queries to Solve

Graph Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
Carlos Ordonez and Predrag T. Tosic

Using Functional Dependencies in Conversion of Relational Databases

to Graph Databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
Youmna A. Megid, Neamat El-Tazi, and Aly Fahmy
XVIII Contents – Part II

Learning

A Two-Level Attentive Pooling Based Hybrid Network for Question

Answer Matching Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
Zhenhua Huang, Guangxu Shan, Jiujun Cheng, and Juan Ni

Features’ Associations in Fuzzy Ensemble Classifiers . . . . . . . . . . . . . . . . . 369

Ilef Ben Slima and Amel Borgi

Learning Ranking Functions by Genetic Programming Revisited . . . . . . . . . . 378

Ricardo Baeza-Yates, Alfredo Cuzzocrea, Domenico Crea,
and Giovanni Lo Bianco

A Comparative Study of Synthetic Dataset Generation Techniques . . . . . . . . 387

Ashish Dandekar, Remmy A. M. Zen, and Stéphane Bressan

Emerging Applications

The Impact of Rainfall and Temperature on Peak and Off-Peak

Urban Traffic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
Aniekan Essien, Ilias Petrounias, Pedro Sampaio, and Sandra Sampaio

Fast Identification of Interesting Spatial Regions with Applications in

Human Development Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
Carl Duffy, Deepak P., Cheng Long, M. Satish Kumar, Amit Thorat,
and Amaresh Dubey

Creating Time Series-Based Metadata for Semantic IoT Web Services. . . . . . 417
Kasper Apajalahti

Topic Detection with Danmaku: A Time-Sync Joint NMF Approach . . . . . . . 428

Qingchun Bai, Qinmin Hu, Faming Fang, and Liang He

Data Mining

Combine Value Clustering and Weighted Value Coupling Learning for

Outlier Detection in Categorical Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
Hongzuo Xu, Yongjun Wang, Zhiyue Wu, Xingkong Ma,
and Zhiquan Qin

Mining Local High Utility Itemsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450

Philippe Fournier-Viger, Yimin Zhang, Jerry Chun-Wei Lin,
Hamido Fujita, and Yun Sing Koh

Mining Trending High Utility Itemsets from Temporal Transaction

Databases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461
Acquah Hackman, Yu Huang, and Vincent S. Tseng
 Contents – Part II XIX

Social Media vs. News Media: Analyzing Real-World Events from

Different Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 471
Liqiang Wang, Ziyu Guo, Yafang Wang, Zeyuan Cui, Shijun Liu,
and Gerard de Melo

Privacy

Differential Privacy for Regularised Linear Regression. . . . . . . . . . . . . . . . . 483

Ashish Dandekar, Debabrota Basu, and Stéphane Bressan

A Metaheuristic Algorithm for Hiding Sensitive Itemsets . . . . . . . . . . . . . . . 492

Jerry Chun-Wei Lin, Yuyu Zhang, Philippe Fournier-Viger,
Youcef Djenouri, and Ji Zhang

Text Processing

Constructing Multiple Domain Taxonomy for Text Processing Tasks. . . . . . . 501

Yihong Zhang, Yongrui Qin, and Longkun Guo

Combining Bilingual Lexicons Extracted from Comparable Corpora:

The Complementary Approach Between Word Embedding and
Text Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
Sourour Belhaj Rhouma, Chiraz Latiri, and Catherine Berrut

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519

 Contents – Part I

Big Data Analytics

Scalable Vertical Mining for Big Data Analytics of Frequent Itemsets . . . . . . 3

Carson K. Leung, Hao Zhang, Joglas Souza, and Wookey Lee

ScaleSCAN: Scalable Density-Based Graph Clustering . . . . . . . . . . . . . . . . 18

Hiroaki Shiokawa, Tomokatsu Takahashi, and Hiroyuki Kitagawa

Sequence-Based Approaches to Course Recommender Systems. . . . . . . . . . . 35

Ren Wang and Osmar R. Zaïane

Data Integrity and Privacy

BFASTDC: A Bitwise Algorithm for Mining Denial Constraints. . . . . . . . . . 53

Eduardo H. M. Pena and Eduardo Cunha de Almeida

BOUNCER: Privacy-Aware Query Processing over Federations

of RDF Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Kemele M. Endris, Zuhair Almhithawi, Ioanna Lytra,
Maria-Esther Vidal, and Sören Auer

Minimising Information Loss on Anonymised High Dimensional Data

with Greedy In-Memory Processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Nikolai J. Podlesny, Anne V. D. M. Kayem, Stephan von Schorlemer,
and Matthias Uflacker

Decision Support Systems

A Diversification-Aware Itemset Placement Framework for Long-Term

Sustainability of Retail Businesses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Parul Chaudhary, Anirban Mondal, and Polepalli Krishna Reddy

Global Analysis of Factors by Considering Trends to Investment Support . . . 119

Makoto Kirihata and Qiang Ma

Efficient Aggregation Query Processing for Large-Scale Multidimensional

Data by Combining RDB and KVS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Yuya Watari, Atsushi Keyaki, Jun Miyazaki, and Masahide Nakamura
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
XXII Contents – Part I

Data Semantics

Learning Interpretable Entity Representation in Linked Data. . . . . . . . . . . . . 153

Takahiro Komamizu

GARUM: A Semantic Similarity Measure Based on Machine Learning

and Entity Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Ignacio Traverso-Ribón and Maria-Esther Vidal

Knowledge Graphs for Semantically Integrating Cyber-Physical Systems . . . . 184

Irlán Grangel-González, Lavdim Halilaj, Maria-Esther Vidal,
Omar Rana, Steffen Lohmann, Sören Auer, and Andreas W. Müller

Cloud Data Processing

Efficient Top-k Cloud Services Query Processing Using Trust and QoS . . . . . 203
Karim Benouaret, Idir Benouaret, Mahmoud Barhamgi,
and Djamal Benslimane

Answering Top-k Queries over Outsourced Sensitive Data in the Cloud. . . . . 218
Sakina Mahboubi, Reza Akbarinia, and Patrick Valduriez

R2 -Tree: An Efficient Indexing Scheme for Server-Centric Data

Center Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
Yin Lin, Xinyi Chen, Xiaofeng Gao, Bin Yao, and Guihai Chen

Time Series Data

Monitoring Range Motif on Streaming Time-Series . . . . . . . . . . . . . . . . . . . 251

Shinya Kato, Daichi Amagata, Shunya Nishio, and Takahiro Hara

MTSC: An Effective Multiple Time Series Compressing Approach . . . . . . . . 267

Ningting Pan, Peng Wang, Jiaye Wu, and Wei Wang

DANCINGLINES: An Analytical Scheme to Depict Cross-Platform

Event Popularity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Tianxiang Gao, Weiming Bao, Jinning Li, Xiaofeng Gao, Boyuan Kong,
Yan Tang, Guihai Chen, and Xuan Li

Social Networks

Community Structure Based Shortest Path Finding for Social Networks . . . . . 303
Yale Chai, Chunyao Song, Peng Nie, Xiaojie Yuan, and Yao Ge
 Contents – Part I XXIII

On Link Stability Detection for Online Social Networks . . . . . . . . . . . . . . . 320

Ji Zhang, Xiaohui Tao, Leonard Tan, Jerry Chun-Wei Lin, Hongzhou Li,
and Liang Chang

EPOC: A Survival Perspective Early Pattern Detection Model for

Outbreak Cascades . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
Chaoqi Yang, Qitian Wu, Xiaofeng Gao, and Guihai Chen

Temporal and Spatial Databases

Analyzing Temporal Keyword Queries for Interactive Search over

Temporal Databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
Qiao Gao, Mong Li Lee, Tok Wang Ling, Gillian Dobbie,
and Zhong Zeng

Implicit Representation of Bigranular Rules for Multigranular Data . . . . . . . . 372

Stephen J. Hegner and M. Andrea Rodríguez

QDR-Tree: An Efficient Index Scheme for Complex Spatial

Keyword Query . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 390
Xinshi Zang, Peiwen Hao, Xiaofeng Gao, Bin Yao, and Guihai Chen

Graph Data and Road Networks

Approximating Diversified Top-k Graph Pattern Matching . . . . . . . . . . . . . . 407

Xin Wang and Huayi Zhan

Boosting PageRank Scores by Optimizing Internal Link Structure . . . . . . . . . 424

Naoto Ohsaka, Tomohiro Sonobe, Naonori Kakimura,
Takuro Fukunaga, Sumio Fujita, and Ken-ichi Kawarabayashi

Finding the Most Navigable Path in Road Networks: A Summary

of Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440
Ramneek Kaur, Vikram Goyal, and Venkata M. V. Gunturi

Load Balancing in Network Voronoi Diagrams Under Overload Penalties . . . 457

Ankita Mehta, Kapish Malik, Venkata M. V. Gunturi, Anurag Goel,
Pooja Sethia, and Aditi Aggarwal

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477

Information Retrieval
 Template Trees: Extracting Actionable
Information from Machine Generated Emails

Manoj K. Agarwal(&) and Jitendra Singh

AI & Research, Microsoft – India, Hyderabad 500032, India

{agarwalm,jisingh}@microsoft.com

Abstract. Many machine generated emails carry important information which

must be acted upon at scheduled time by the recipient. Thus, it becomes a
natural goal to automatically extract such actionable information from these
emails and communicate to the users. These emails are generated for many
different domains, providing different types of services. However, such emails
carry personal information, therefore, it becomes difficult to get access to large
corpus of labeled data for supervised information extraction methods.
In this paper, we propose a novel method to automatically identify part of the
email containing actionable information, called core region of the email, with
the aid of a domain dictionary. Domain dictionary is generated based on the
public information of the domain. The core regions are stored as template trees -
a template tree is a sub-tree embedded in the email’s HTML DOM tree.
Our experiments over real data show, structure of the core region of the email,
containing all the information of our interest, is very simple and it is 85%–98%
smaller compared to the original email. Further, our experiments also show that
the template trees are highly repetitive across diverse set of emails from a given
service provider.

Keywords: Email
Templates
Trees
Information retrieval

Algorithm

1 Introduction

In a recent study, it is showed that 90% of Web mail traffic is machine generated [1, 7,
8, 10]. As mentioned in [1], “A common characteristics of these machine generated
messages is that most of them are highly structured documents, with rich HTML
formatting, and they are repeated over and over, modulo minor variations, in the
global mail corpus. These characteristics clearly facilitate the application of auto-
mated data extraction and learning methods at a very large scale”. With a significant
chunk of web mail traffic composed of machine generated emails, it becomes a natural
goal to automatically extract the information from these emails, which must be acted
upon by the recipient by scheduled time, e.g., payment of utility bill by the due date.
Therefore, we define actionable information as “information that must be acted upon by
scheduled time, by the recipient”. The requirement to extract actionable information
from machine generated emails is present for many domains, such as flight information,
shipment arrival, payment due date, etc., [5].

© Springer Nature Switzerland AG 2018

S. Hartmann et al. (Eds.): DEXA 2018, LNCS 11030, pp. 3–18, 2018.
https://doi.org/10.1007/978-3-319-98812-2_1
4 M. K. Agarwal and J. Singh

A big challenge faced by such systems is: such emails contain personal informa-
tion. Therefore, it becomes difficult to get access to large corpus of labeled data to build
annotators based on supervised methods [1, 3]. Moreover, even though machine
generated emails are structured, their structure changes over time. Further, new service
providers join continuously. Hence, it becomes difficult for supervised methods to
model the tail traffic. For example, hotel reservation emails have more than 8000
different small providers representing more than 50% of the total such emails in our
database. Similarly, we have over one thousand small service providers for flight data,
representing more than 40% of all flight reservation emails. There exist supervised
methods [8, 9] that work on labeled data with varying degree of success.
In this paper, we present a novel method to extract actionable information from
emails. We show that the region in the email containing actionable information can be
represented as combination of one or more small template trees. These template trees
have significantly simpler structure compared to the original email. Further, these
template trees are highly repetitive across a diverse set of email from a given service
provider and contain all the information of our interest. We develop a principled and
scalable approach which exploits the semi-structured format of the emails to extract
these template trees. Requirement of training data, for building supervised annotators
over email data, can be reduced significantly, if such small and well-structured snippets
from the emails can be identified for information extraction.
Only information needed by our system is a domain specific dictionary. We call it
domain knowledge. The domain knowledge has been used in earlier systems to
automatically extract the information from HTML web pages [15]. Similarly, there are
existing systems to extract wrappers from HTML pages [16]. However, unlike a
wrapper a template tree is a sub-tree in the email HTML DOM tree. In [17], authors
present a system to extract templates of an entire web pages in an unsupervised manner.
On the other hand, we templatize only the structure of the core region of the emails,
containing the information of our interest, with the aid of domain knowledge.
However, the work in [16, 17] exposes that machine generated HTML documents
(web pages or emails) have a fixed structure, encoded with the actual data. We present
a novel system, that builds on these ideas to extract actionable information from emails.
We use the term ‘domain’ for an entire service. For instance, for extracting
information from flight related emails, ‘flight’ is the domain. A specific vendor in a
domain is called service provider, or just ‘provider’. In ‘flight’ domain, each airline is a
provider (e.g., ‘American Airlines’). We will use the emails from flight domain as
running example throughout the paper.
Domain specific dictionaries contain keywords and regex patterns specific to that
domain. Dictionaries are applicable on entire domain and are not specific to any
provider. Therefore, they are built using public information and it is much easier to
acquire and learn domain specific dictionaries as opposed to acquiring labelled data for
every provider (cf. Sect. 3.2).
The technique described in this paper is applicable across many domains for which
domain specific information can be represented in a dictionary. This is a highly generic
requirement, applicable across most domains, producing machine generated emails.
Random documents with unrelated
content Scribd suggests to you:
When greater strength was needed the thickness of the side walls
was increased to 30 ins. and that of the arch to six rings of brick.

Cross Section. Longitudinal Section.

 Cross Section. Cross Section.

Figs. 159 and 160.—Relining Timber-Lined Tunnel.

The first plan adopted in putting in the masonry was to remove

all the timbering; but owing to the large number of falls and slides
this was abandoned, and the plan followed was to leave in the three
roof segments of the timbering with the overlying cord-wood
packing and débris. In carrying on the work the first step was to
remove the side timbers. This was done by supporting the roof
timbers, as shown in Fig. 159; that is, the first and fourth arch rib of
an 8-ft. section containing four arch ribs were supported by
temporary posts. The intermediate arch ribs were supported against
the downward pressure by 6 × 6 in. timbers, extending from the
side ribs near the tops of the temporary posts to the opposite sides
of the intermediate roof segments, as shown in the longitudinal
section, Fig. 160. To resist the pressure from the sides, 4 × 6 in.
braces were placed across the tunnel from near the center of the
intermediate segments to the upper ends of the hip segments, as
shown in the cross-section, Fig. 159. The hip segments were then
sawed off below the notch, and the side timbering removed and the
masonry built.
The stone was conveyed into the tunnel on flat cars, and laid by
means of small derricks located on the cars. Two derricks were used,
one for each side wall, and the work on both walls was carried on
simultaneously.
The arch was built upon a centering, the ribs of which were 51⁄2
ins. less in diameter than the distance between the side walls, so as
to permit the use of 23⁄4 ins. lagging. Each center had three ribs,
made in 1-in. or 2-in. board segments, 10 ins. thick and 14 ins.
deep. These ribs were mounted on frames, which followed the
opposite walls, and were 4 ft. apart, making the total length of the
center out to out about 9 ft. The frames, upon which the ribs were
supported, are shown in Fig. 161. As will be seen, they were
mounted on dollys to enable the center to be moved from one
section to another. Jacks were used to raise and lower the center
into its proper position.
 Cross Section. Longitudinal Section.

Fig. 161.—Relining Timber-Lined Tunnel, Great Northern Ry.

The arch was built up from the springing lines on both sides at
the same time, four masons being employed. The rings were built
beginning with the intrados, which was brought up, say, a distance
of about 2 ft. from the springing line. Then the back of the ring was
well plastered with from 3⁄8 in. to 1⁄2 in. of mortar, and the second
ring brought up to the same height and plastered on the back, and
so on until the last ring was laid. After bringing the full width of the
arch up some distance, new laggings were placed on the ribs for an
additional height of 2 ft. and the same process was repeated. All the
space between the extrados of the masonry arch and the old lining
was compactly filled with dry rubble. When high enough so that the
hip segments had a foot or more bearing on the masonry the
segments were securely wedged and blocked up against the
brickwork, and the longitudinal 4 × 6 in. timbers removed. The
remaining space was now clear for completion of the arch, and both
sides were brought up until there was not sufficient space for four
masons to work, when the keying was completed by two masons
beginning at the completed and working back toward the toothed
end. The brickwork was built from the top of a staging-car.

Cross Section. Longitudinal Section.

Fig. 162.—Relining Timber-Lined Tunnel, Great Northern Ry.

In a few instances where slides occurred after the removal of the

slide timbering, the method of re timbering the tunnel shown in Fig.
162 was adopted. Two side drifts were first run 21⁄2 ft. wide by 4 ft.
high, and the plate timbers placed in position and blocked. Cross
drifts were then run, and the roof segments placed, and the core
down to the level of the bottoms of the side drifts taken out. The
lower wall plates were then placed and the hip segments inserted.
The bench was then taken down by degrees, the side plates being
held by jacks, and the posts placed one at a time. As the masonry at
the points where slides occur consists of 30-in. walls and six-ring
arch, the timbering was 22 ft. wide in the clear, with other
dimensions as shown in Fig. 162.
Only a single crew of brick and stone masons was employed. In
order to prepare the sections for these masons it was necessary to
have timber and trimming crews at work throughout the whole day
of 24 hours, so that an engine and two train crews were in constant
attendance. The single mason crews were able to complete 8 ft. of
side wall and arch in 24 hours. The number of men actually
employed at the tunnel was 35. This included electric-light
maintenance, and all other labor pertaining to the work. The tunnel
was lighted by an Edison dynamo of 20 arc light capacity, one arc
light being placed on each side of the tunnel at all working-places.
Each lamp carried a coil of wire 20 or 30 ft. long to allow it to be
shifted from place to place without delay.
Mullan Tunnel.—This tunnel is 3,850 ft. long, and crosses the
main range of the Rocky Mountains, about 20 miles west of Helena,
Mont. The tunnel is on a tangent throughout, and has a grade of
20% falling toward the east. The summit of the grade, west of the
tunnel, is 5,548 ft. above sea level, and the mountain above the line
of the tunnel rises to an elevation of 5,855 ft. Owing to the
treacherous nature of the material through which the tunnel passed,
it had been a constant menace to traffic ever since its construction
in 1883, and numerous delays to trains had been caused by the falls
of rock and fires in the timber lining. For these reasons it was finally
decided to build a permanent masonry lining, and work on this was
begun in July, 1892.
With Wall Plates. Without Wall Plates.

Old Timber Sections.

Minimum Section. Average Section.

Permanent Work.
 Fig. 163.—Relining Timber Lined Tunnel, Great Northern Ry.

The original timbering consisted of sets spaced 4 ft. apart c. to

c., with 12 × 12 in. posts supporting wall plates, and a five-segment
arch of 12 × 12 in. timbers joined by 11⁄2-in. dowels. The arch was
covered with 4-in. lagging, and the space between this and the roof
was filled with cordwood. Except where the width had been reduced
by timbering placed inside the original timbering to increase the
strength, the clear width was 16 ft., and the clear height 20 ft.
above the top of the rail. Fig. 163 shows the timbering and also the
form of masonry lining adopted. The side walls are of concrete and
the arch of brick. This new masonry, of course, required the removal
of all the original timbering. The manner of doing this work is as
follows: A 7-ft. section, A B, Fig. 164, was first prepared by removing
one post and supporting the arch by struts, S S. After clearing away
any backing, and excavating for the foundation of the side wall, two
temporary posts, F F, were set up, and fastened by hook bolts. Fig.
146, L, and a lagging was built to form a mold for the concrete.
Several of these 7-ft. sections were prepared at a time, each two
being separated by a 5-ft. section of timbering.
Section, with Concrete Car.
 With Wall Plate. Without Wall
Plate.

Longitudinal Section.
Fig. 164.—Construction of Centering Mullan Tunnel.

The mortar car was then run along, and enough mortar (1
cement to 3 sand) was run by the chute into each section to make
an 8-in. layer of concrete. As the car passed along to each section,
broken stone was shoveled into the last preceding section until all
the mortar was taken up. The walls were thus built up in 8-in.
layers, and became hard enough to support the arches in about 10
to 14 days. The arches were then allowed to rest on the wall, and
the posts of the remaining 5-ft. sections were removed, and the
concrete wall built up in the same way as before.
The average progress per working-day was 30 ft. of side wall, or
about 45 cu. yds.; and the average cost, including all work required
in removing the timber work, train service, lights and tools,
engineering and superintendence, and interest on plant, was $8 per
cubic yard.

Fig. 165.—Centering Mullan Tunnel.

The centering used for putting in the brick arches is shown in Fig.
165. From 3 ft. to 9 ft. of arch was put in at a time, the length
depending upon the nature of the ground. To remove the old timber
arch, one of the segments was partly sawed through; and then a
small charge of giant powder was exploded in it, the resulting
débris, cordwood, rock, etc., being caught by a platform car
extending underneath. From this car the débris was removed to
another car, which conveyed it out of the tunnel. The center was
then placed and the brickwork begun, the cement car shown in Fig.
164 being used for mixing the mortar. The size of the bricks used
was 21⁄2 + 21⁄2 + 9 ins., four rings making a 20-in. arch and giving
1.62 cu. yds. of masonry in the arch per lin. ft. of tunnel. The bricks
were laid in rowlock bond, two gangs, of three bricklayers and six
helpers each, laying about 12 lin. ft. per day. The brickwork cost
about $17 per cu. yd. The total cost of the new lining averaged
about $50 per lin. ft.

Cross Section. Longitudinal Section.

Fig. 166.—Relining Timber-Lined Tunnel, Norfolk and Western Ry.

Larger illustration

Little Tom Tunnel.—The tunnel has a total length of 1,902 ft.,

but only 1,410 ft. of it were originally lined with timber. This old
timber lining consists of bents spaced 3 ft. apart, and located as
shown by the dotted lines in the cross-section, Fig. 166. Instead of
renewing this timber, it was decided to replace it with a brick lining.
Although the tunnel was constructed through rock, this rock is of a
seamy character, and in some portions of the tunnel it disintegrates
on exposure to the air. In removing the timber to make place for the
new lining some of the roof was found close to the lagging, but
often also considerable sections showed breakages in the roof
extending to a height varying from 1 ft. to 12 ft. above the upper
side of the timbering. This dangerous condition of the roof made it
necessary that only a small section of the timber lining should be
removed at one time. It made it necessary, also, that the brick arch
should be built quickly to close this opening, and finally that all
details of centers, etc., should be arranged so as to furnish ample
clearance to trains. The accompanying illustrations show the solution
of the problem which was arrived at.

Fig. 167.—Relining Timber-Lined Tunnel, Norfolk and Western Ry.

 Referring to the transverse and longitudinal sections shown by
Fig. 166, it will be seen that two side trestles were built to carry an
adjustable centering for the roof arch. Two sections of these trestles
and centerings were used alternately, one being carried ahead and
set up to remove the timbering while the masons were at work on
the other. The manner of setting up and adjusting the trestles and
centerings is shown by Fig. 166 and also by Fig. 167, which is an
enlarged detail drawing of the set screw and rollers for the centering
ribs. The following is the bill of material required for one set of
trestles and one center:

Trestles:
Caps and sills 8 pieces 8× 8 ins. × 20 ft.
Posts 18 „ 8× 8 „ × 11 „
Braces 16 „ 6× 4 „ × 7 „

Centerings:
Ribs 27 „ 2 × 18 „ × 7 „
Bracing 12 „ 2 × 8 „ × 7 „
Support to crown lagging 2 „ 6 × 6 „ × 10 „
Crown lagging 20 „ 3 × 6 „ × 2 „
Side lagging 30 „ 3 × 6 „ × 10 „
Side strips 2 „ 2 × 12 „ × 9 „
Blocking for rollers 1 „ 5 × 8 „ × 12 „
6 screw and roller castings complete with bolts and lever; 114 bolts 3⁄4-ins.
in diameter; 71⁄2 U. H. hexagonal nut and 2 cast washers each.

With this arrangement the progress made per day varied from 2
lin. ft. to 3 lin. ft. of lining complete. By work complete is meant the
entire lining, including stone packing between the brickwork and the
rock. On Feb. 23, 1900, 363 ft. of lining had been completed, at a
cost of $33.50 per lin. ft. This cost includes the cost of removing the
old timber, the loose rock above it, and all other work whatsoever.
 CHAPTER XXIV.

THE VENTILATION AND LIGHTING OF TUNNELS DURING

CONSTRUCTION.

VENTILATION.

In long tunnels, especially when excavated in hard rock, proper

ventilation is of great importance, because the air cannot be easily
renewed, and the amount of oxygen consumed by miners horses
and lamps during construction is very large. The gases produced by
blasting also tend to fill the head of excavation with foul air. Pure
atmospheric air contains about 21% of oxygen and only 0.04% of
carbonic acid; when the latter gas reaches 0.1% the fact is indicated
by the bad odor; at 0.3% the air is considered foul, and when it
reaches 0.5% it is dangerous. It is generally admitted that the
standard of purity of the air is when it contains 0.08% of carbonic
acid.
A large quantity of carbonic acid in the air is easily detected by
observing the lamps, which then give out a dim red light and smoke
perceptibly; the workmen also suffer from headache and pains in the
eyes, and breathe with difficulty. Naturally, miners cannot easily
work in foul air and, therefore, make very slow progress. It is,
therefore, to the interest of the engineer to afford good ventilation,
not only because of his duty to care for the safety and health of his
men, but also for reasons of economy, so that the men may work
with the greatest possible ease, thus assuring the rapid progress of
the work.
 It would be impossible to change completely the atmosphere
inside a tunnel, as the gases developed from blasting will penetrate
into all the cavities and gather there, but the fresh air carried inside
by ventilation has a very small percentage of carbonic acid, mixes
with that which contains a greater quantity, and dilutes it until the
air reaches the standard of purity. We have not here considered the
gases developed from the decomposition of carboniferous and
sulphuric rocks, which may be met with in some tunnels, and which
render ventilation still more necessary. Tunnels may be ventilated
either by natural or artificial means.
Natural Ventilation.—It is well known that if two rooms of
different temperatures are put in communication with each other,
e.g., by opening a door, a draft from the colder room will enter the
other from the bottom, and a similar draft at the top, but with a
contrary direction, will carry the hot air into the colder room, thus
producing perfect ventilation, until the two rooms have the same
temperature. Now, during the construction of tunnels the
temperature inside may be considered as constant, or independent
of the outside atmospheric variations; hence during summer and
winter, there will always be a draft affording ventilation, owing to the
difference of temperature inside and outside the tunnel. In winter
time the cold air outside will enter at the bottom of the entrances
and headings, or along the sides of the shafts, and the hot air will
pass out near the top of the headings or entrances or the center of
the shafts; in summer the air currents will take the contrary
direction.
Natural ventilation in tunnels is improved when the excavation of
the heading reaches a shaft, because the interior air can then
communicate with the exterior at two points, at different levels. In
such cases a force equal to the difference in weight between a
column of air in the shaft and a similar one of different density at
the entrance of the tunnel, will act upon the mass of air in the
tunnel and keep it in movement, thus producing ventilation.
Consequently, during winter, when the outside air has greater weight
than that inside, the air will come in by the headings and go out by
the shaft, and in the summer it will enter at the shaft and pass out
at the entrance. Sometimes to afford better ventilation shafts 8 or 12
in. in diameter are sunk exclusively for the purpose of changing the
air. When the inside temperature is equal to that outside, as often
happens during the spring and autumn, there are no drafts, and
consequently the air in the excavation is not renewed and becomes
foul; then fires are lighted under the shaft and a draft is artificially
produced. The hot air going out through the shaft, as through a
chimney, allows the fresh air to come in as in ordinary ventilation.
When the head of the excavation is very far from the entrances,
or when the mountain is too high to allow excavation by shafts, it is
quite impossible to secure good natural ventilation, especially during
the spring and autumn months, and the engineer has to resort to
some artificial means by which to supply fresh air to the workmen.
Artificial Ventilation.—Artificial ventilation in tunnels may be
obtained in two different ways, known as the vacuum and plenum
methods. Their characteristic difference consists in this, that in the
vacuum method the air is drawn from the inside and the vacuum
thus produced causes the fresh air from the outside to rush in, while
the plenum method consists in forcing in the fresh air which dilutes
the carbonic air produced inside the tunnel by workingmen and
explosives. In the vacuum method the pressure of the atmosphere
inside the tunnel is always less than the pressure outside, while in
the plenum method the pressure within is always greater than that
outside. Ventilation is the result of this difference of pressure, as the
tendency of the air toward equilibrium produces continuous drafts.
Both these methods have their advantages and disadvantages; but
in the presence of hard rock, when explosives are continually
required, the vacuum method is considered the best, because the
gases attracted to the exhaust pipes are expelled without passing
through the whole length of the tunnel, thus avoiding the trouble
that a draft of foul air will give to the workmen who are within the
tunnel. In both these methods it is necessary to separate the fresh
air from the foul one; and this is done by means of pipes which will
exhaust and expel the foul air in the vacuum method, or force to the
front a current of fresh air when the plenum method is used.
Artificial ventilation may also be obtained by compressed air which is
set free after it has driven the machines, especially in tunnels
excavated through rock, when rock drilling machines moved by
compressed air are employed.
Vacuum Method Contrivances.—The most common of the
vacuum appliances consists in the simple arrangement of a pipe
leading from the head of the tunnel out through the fire of a
furnace. The air in the pipe is rarefied by the heat of the furnace and
then set free from the other end of the pipe, thus creating a partial
vacuum in the pipe, into which the foul air of the head rushes, the
fresh air from the entrance taking its place, and thus ventilating the
tunnel. A similar arrangement may be used with shafts, and the foul
air may be driven out by a furnace which is placed either at the top
or bottom of the shaft. Such furnaces act the same as those
commonly used for heating purposes in the houses, with this
difference, that, instead of fresh air being forced in, foul air is
expelled. Another simple arrangement for producing a vacuum is by
means of a steam jet which is thrown into the pipe, and which helps
the expulsion of the air by heating it, thus producing a different
density which originates a draft besides that mechanically originated
by the force of the steam jet, which tends to carry out the foul air of
the pipes.
Foul air may also be expelled by means of exhaust fans which
are connected with pipes near the entrance of the tunnel. The fan
consists of a box containing a kind of a paddle wheel turned by
steam or water power and arranged so as to revolve at a high
speed. The air inside the pipe is forced out by blades attached to the
wheel, and thus the foul air of the front is driven away and fresh air
from the entrance rushes in to take its place, and perfect ventilation
is obtained.
The best manner of expelling foul air from tunnels, according to
the vacuum method, is by means of bell exhausters. This consists of
two sets of bells connected by an oscillating beam and balancing
each other. Each set consists of a movable bell, which covers and
surrounds a fixed bell with a water joint. In the central part of the
fixed bell there are valves which open upwards, and on the bottom
of each movable bell there are valves which open from the outside.
When one bell ascends, the valves at the bottom are closed, the air
beneath is then rarefied, and a vacuum is produced; the valves in
the central part of the fixed bell filled with water are opened, and
there is an aspiratory action from the pipe leading to the headings,
and the foul air is thus carried away. The apparatus makes about ten
oscillations per minute, and the dimensions of the bells depend upon
the quantity of air to be exhausted in a minute. In the St. Gothard
tunnel, where these bell exhausters were used, they exhausted
16,500 cu. ft. of air per minute.
Plenum Method Contrivances.—Fresh air may be driven into
tunnels to dilute the carbonic acid by two different ways, viz., by
water blast and by fans. Water when running at a great velocity
produces a movement in the air which may be sometimes usefully
and economically employed for ventilating tunnels. Water falling
vertically is let run into a large horizontal zinc pipe having a funnel at
the outer end; into this the air attracted by the velocity of the water
is forced. By an opening at the bottom the water is afterward
withdrawn from the pipe, and there remains only the air which is
pushed forward by the air which is being continually sucked in by
the velocity of the water.
The best and most common means of ventilation by the plenum
method is by fans. There are numerous varieties of these fans in the
market, but they all consist of a kind of fan wheel which by rapid
revolution forces the fresh air into the pipe leading to the headings
of the tunnel or to the working places. Instead of a large single fan,
such as is used for mining purposes, it is better to have a number of
small fans acting independently of each other, conveying the fresh
air where it is needed through independent pipes.
Saccardo’s System.—A new method of ventilating tunnels was
devised by Mr. Saccardo for the ventilation of the Pracchia tunnel
along the Bologna and Lucca Railway in Italy. At the highest end of
the tunnel the mouth was contracted inward in a funnel shaped form
so as to just admit a train. Immediately at this contraction, a lateral
tunnel, 50 feet long, branched off from one side of the main tunnel.
At the mouth of this lateral tunnel was installed a fan which forced
air into the tunnel and with 70 revolutions per minute delivered
3.532 cu. ft. of air per second at a water pressure of 1 in. This air
current was directed inward through a second contraction or funnel,
parallel to the one at the entrance and 23 ft. beyond it. In operation
the action of the artificial air current was to suck in a considerable
volume of outside air, while the air pressure was sufficient to
counterbalance the movement of air produced by a train moving at a
velocity of 16.1 ft. per second. Mr. Saccardo’s method was employed
in ventilating a tunnel on the Norfolk and Western Railway with
satisfactory results.
Compressed Air.—In the excavation of tunnels in hard rock a
number of rock drilling machines are employed which are moved by
compressed air at a pressure of not less than five atmospheres. At
each stroke about 100 cu. ins. of compressed air are set free, and at
an average of 10 strokes per minute there would be 5000 cu. ins. of
air at five atmospheres or 25,000 cu. ins., or a little more than 175
cu. ft. of fresh air at normal pressure set free every minute by each
of the machines employed. But the air exhausted from the drilling
machine is foul.
Regarding ventilation by compressed air, Mr. Adolph Sutro, in a
lecture delivered to the mining students of the University of
California, said:
“I will note a curious fact which I have never seen explained, and which is
worthy of close investigation by means of experiments. In the Sutro tunnel we
found that the compressed air used for driving the machine drills, after having
been compressed and expanded and discharged from the drills, was not
wholesome to breathe, and the men and mules would all crowd around the end of
the blower pipe to get fresh air. Whether the air in being compressed has parted
with some of its oxygen or because vitiated from some other cause, I do not
know, and I hope that this subject will at some future day be carefully examined
into.”

In the December, 1901, number of “Compressed Air,” a magazine

especially devoted to the useful application of compressed air, is
read:
Compressed air wasted from power drills is so contaminated with oil from the
cylinders that it cannot be taken into consideration as ventilation. It is as
important to displace it with pure air as it is to drive out or draw off other vitiated
air. The ventilation should be an independent supply provided by fan or blower,
delivering by pipe at the point where miners are working.

Quantity of Air.—The quantity of air to be introduced into

tunnels must be in proportion to the oxygen consumed by the men,
the animals, and the explosions. It is allowed that the quantity of air
required for breathing purpose and explosions is as follows:
1 workman with lamp needs 240 cu. yds. of fresh air in 24 hours.
1 horse „ 850 „ „ „ „
1 lb. gunpowder 100 „ „ „
1 lb. dynamite 150 „ „ „

In a long tunnel excavated through hard rock the number of

workmen all together may be assumed at 400 at each end, and each
workman is supposed to be furnished with a lamp. No less than ten
horses are employed, and the average quantity of dynamite
consumed is 600 lbs. per day. From the data given the consumption
of air by workmen and lamps would be: 240 × 400 = 96,000 cu.
yds.; the consumption of air by horses would be 850 × 10 = 8500
cu. yds.; the consumption of air by dynamite would be 150 × 600 =
90,000 cu. yds.; making a total consumption of air per day of
194,500 cu. yds., or about 8000 cu. yds. per hour.
To obtain good ventilation, then, it will be necessary to furnish
every hour a quantity of fresh air amounting to not less than 8000
cu. yds. Since, however, a large quantity of pure air is expelled with
the foul air, it is necessary greatly to increase this quantity.
It may be observed, in closing, that the water having its particles
divided, as in a fog or mist, rapidly precipitates the gases produced
by explosions. Now, when hydraulic machines are used, there is a
hollow ball pierced by holes that are almost imperceptible, from
which the compressed water spreads in very subtile particles, and
this causes the fall of the gases from explosions. Such a method of
precipitating gases is very good, but does not have the advantage of
supplying new oxygen to replace that consumed by the men,
animals, lamps, and explosions; besides, it has the defect of
increasing the quantity of water to be removed. In tunnels the pipes
used either for conveying the fresh air or for carrying away the foul
air, are of iron, having a diameter of about 8 in.; they are fixed along
the side walls about 3 ft. above the inverted arch.

LIGHTING.

The object and necessity of a perfect lighting of the tunnel-

workings during construction are so obvious that they need not be
enlarged upon. Comparatively few tunnels require lighting after
completion; and these are generally tunnels for passenger traffic
under city streets, of which the Boston Subway is a representative
American example. Considering the methods of lighting tunnels
during construction, we may, for sake of convenience, chiefly, divide
the means of supplying light into (1) lamps and lanterns usually
burning oil; (2) coal-gas lighting; (3) acetylene gas lighting; and (4)
electric lighting.
Lamps and Lanterns.—Lamps and lanterns are commonly
employed by engineers for making surveys inside the tunnel, and to
light the instrument. For ranging in the center line, a convenient
form of lamp consists of an oil light inclosed in glass chimney
covered with sheet metal, except for a slit at the front and back
through which the light shines, and on which the observer sights his
instrument. To direct the operations of his rodmen the engineer
usually employs a lantern, either with white or colored glass, much
like the ordinary railway trainman’s lantern, which he swings
according to some prearranged code of signals.
Lamps and lanterns are used by the workmen both for signaling
and for lighting the workings. For signaling purposes red lanterns are
usually placed to denote the presence of unexploded blasts or other
points of possible danger; and colored or white lights are usually
placed on the front and rear of spoil and material trains. For lighting
purposes, two forms of lamps are employed, which may be
somewhat crudely designated as lamps for individual use and lamps
for general lighting. Individual lamps are usually of small size, and
burn oil; they may be carried in front of the miner’s helmet, or be
fixed to standards, which can be set up close to the work being done
by each man. Miners’ safety lamps should be employed where there
is danger from gas. A great variety of lamps for mining and
tunneling purposes are on the market, for descriptions of which the
reader is referred to the catalogues of their manufacturers.
Lamps for general lighting are always of larger size than lamps
for individual use. A common form consists of a cylinder ten or
twelve inches in diameter, provided with a hook or bail for
suspension, and filled with benzine, gasolene, or other similar oil.
Connected with this cylinder is a pipe of considerable length and
small diameter through which the benzine or gasolene vapor runs,
and burns when lighted with a brilliant flame. Lamps of this type
burning gasolene were extensively employed in building the Croton
Aqueduct tunnel. Various patented forms of lamps for burning coal-
oil products are on the market, for descriptions of which the
manufacturers’ catalogues may be consulted.
Coal-gas Lighting.—A common method of lighting tunnel
workings is by piping coal-gas into the headings and drifts from
some nearby permanent gas plant, or from a special gas works
constructed especially for the work. Gas lighting has the great
advantage over lamps and lanterns of giving a light which is more
brilliant and steady. Its great objection is the danger of explosion
caused by leaks in the pipes, by breaks caused by flying fragments
of rock, and by the carelessness of workmen who neglect to turn off
completely the burners when they extinguish the lights. In nearly
every tunnel where gas has been used for lighting, the records of
the work show the occurrence of accidents which have sometimes
been very serious, particularly when fire has been communicated to
the tunnel timbering.
Acetylene Gas Lighting.—The comparatively recent
development of acetylene gas manufactured from carbide of calcium
has given little opportunity for its use in tunnel lighting, and the only
instance of its use in the United States, so far as the author knows,
is the water-works tunnel conduit for the city of Washington, D. C.
Col. A. M. Miller, U. S. Engineer Corps, who is in charge of this work,
describes the method adopted in his annual report for 1899 as
follows:—
“It had been the practice to do all work underground by the light of miners’
lamps and torches. This means of illumination is very poor for mechanical work.
The fumes and smoke from blasting, added to the smoke from torches and lamps,
render the atmosphere underground, especially when the barometer conditions
were unfavorable to ventilation, very offensive and discomforting to the workmen.
An investigation of the subject of lighting the tunnel by other means, more
especially at the locality where the mechanics were at work,—brick and stone
masons, and the workmen on the iron lining,—resulted in the selection of
acetylene gas as the most available and economical in this special emergency.
Accordingly, an acetylene gas plant for 300 burners was erected at Champlain-
Avenue shaft, and one for 60 lights at Foundry Branch. The engine-houses at the
shafts, the head-houses, and localities in the tunnel, when required, are lighted by
these plants.
“Gas pipes were carried down the Champlain-Avenue shaft and along the
tunnel both in an easterly and westerly direction, with cocks for burners at proper
intervals every 30 feet; and this system sufficed for illumination from Hock Creek
to Harvard University, a distance of over two miles. The plant erected at Foundry
Branch was in like manner utilized for the illumination from that point in both
directions.
“By connecting with the stopcocks by means of a rubber hose, a movable light,
chandelier, or ‘Christmas-tree’ of any required number of burners is used, thus
concentrating the light in the immediate vicinity of the work, and also enabling the
illumination to be carried into the cavities or ‘crow-nests,’ so called, behind the
defective old lining.
“This method of illuminating has proved very satisfactory and quite economical.
It is especially valuable as enabling good work to be done, and facilitating a
thorough inspection of the same.”

Electric Lighting.—By far the most perfect, and at present the

most commonly employed means for lighting tunnel workings, is
electricity. The light furnished by electric lamps is steady and
brilliant, and does not consume oxygen or give off offensive gases.
The wires are easily removed and extended, and the lamps are
easily put in place and removed. About the only objection to the
method is the fragility of the lamps, which are easily broken by the
flying stones and the concussion produced by blasting.
 CHAPTER XXV.

THE COST OF TUNNEL EXCAVATION AND THE TIME

REQUIRED FOR THE WORK.

Cost.—The cost of a tunnel will depend upon the cost of the two
principal operations required in its construction, viz., the excavation
of the cross section and the lining of the excavation with masonry,
metal, or timber. These two operations may in turn be subdivided, in
respect to expense, into cost of labor and cost of materials. It is a
comparatively simple matter to calculate the cost of the building
materials required to construct a tunnel; but it is very difficult to
estimate with accuracy what the cost of labor will be. The reason for
this is that it is impossible to foresee exactly what the conditions will
be; the character of the material may change greatly as the work
proceeds, increasing or decreasing the cost of excavation; water
may be encountered in quantities which will materially increase the
difficulties of the work, etc. Nevertheless, while accurate preliminary
estimates of cost are not practicable, it is always desirable to
attempt to obtain some idea of the probable expense of the work
before beginning it, and the more usual means of getting at this
point will be discussed here.
Two methods of estimating the cost of tunnel work are
employed. The first is to calculate the probable expense of the
various items of work, based upon the available data, per unit of
length, and then add to this a margin of at least 10% to allow for
contingencies; the second is to apply to the new work the unit cost
of some previous tunnel built under substantially the same
conditions. In the first method it is usual to consider the strutting
and hauling as constituting a part of the work of excavation. To
estimate the cost of excavation involves the consideration of three
general items, viz., the excavation proper, the strutting of the walls
of the excavation, and the hauling of the excavated materials and
the materials of construction.
The cost of excavating the preliminary headings or drifts is
greater per unit of material removed than that of excavating the
enlargement of the section. The cost of bottom drifts is also always
greater than that of top headings, the material penetrated remaining
the same. Mr. Rziha gives the comparative unit costs of excavating
drifts, headings, and enlargement of the profile as follows:—
Bottom drifts $9.20 per cu. yd.
Top headings 4.80 „ „ „
Enlargement of profile 2.84 „ „ „

The cost of hauling increases with the length of the tunnel. This
fact and amount of this increase are indicated by the following actual
prices for the Arlberg tunnel:—
Top heading $6.76 per cu. yd., increasing 37 cts. per mile
Bottom drift 7.40 „ „ „ „ 26 „ „ „
Enlargement of profile 2.70 „ „ „ „ 10 „ „ „

In all the prices given above, the cost of strutting and hauling is
included in the cost of excavation.
The cost of excavation is not always the same for the same
character of materials in different tunnels. The following figures
show the prices paid for the excavation of calcareous rock in four
different German tunnels:—
Berliner Nordhausen Wetzler R.R. $1.24 per cu. yd.
Ofen 1.30 „ „ „
Stafflach 2.76 „ „ „
Gries 1.92 „ „ „

The method of tunneling has little influence upon the cost of the
work, as shown by the following figures from tunnels excavated
through calcareous rock by different methods:—
Ofen tunnel Austrian method $93.19 per lin. ft.
Dorremberg tunnel Belgian method 86.08 „ „ „
Stafflach tunnel English method 91.69 „ „ „

The Martha and Merten tunnels, excavated through soft ground

by the Austrian and German methods respectively, cost $87.95 and
$87.55 per lin. ft. respectively. In the excavation of the various
sections of the tunnel for the new Croton Aqueduct in America, the
following prices were paid:—
Excavation of heading $8 to $10.00 per cu. yd.
Tunnel in soft ground 8 to 9.00 „ „ „
Tunnel in rock 7 to 8.50 „ „ „
Brick masonry 10.00 „ „ „
Timber in place $40 per M. ft. B. M.

It is the practice in America to include the work of hauling under

excavation, but not to include the strutting, which is paid for
separately. In some cases only the market price of the timber is paid
for separately, the cost of setting up being included in the price of
excavation. The writer prefers the European practice of including the
total cost of timbering under excavation, since the two operations
are so closely connected, and since the contractor employs the same
timber over and over again. Knowing the dimensions of the several
members of the strutting, it is a simple, although somewhat tedious,
process to calculate the total quantity required. An idea of the
quantity of timber required for strutting in soft ground may be had
from the data given on page 55. The quantity will decrease as the
cohesion of the material penetrated increases, until it becomes so
small in hard rock-tunnels as to cut very little figure in the total cost.
The cost of hoisting excavated materials through shafts depends
upon the depth from which it is hoisted, and upon the character of
hoisting apparatus employed. The following table, showing the cost
of hoisting for different lifts and by different methods, is given by
Rziha, the cost being in francs per cubic meter:—
 Steam
Windlass. Horse Gins.
Hoists.
Height One Two
in Metres. Francs Horse. Horses. Francs
per Cu. M. Francs Francs per Cu. M.
per Cu. M. per Cu. M.
15 0.172 0.077 0.062 0.035
30 0.212 0.087 0.070 0.045
45 0.257 0.100 0.080 0.050
60 0.305 0.112 0.092 0.082
90 0.410 0.152 0.110 0.087
120 0.535 0.195 0.135 0.092
150 0.722 0.240 0.157 0.112

Mr. Séjourné, a French engineer, who has been connected with

the construction of numerous tunnels by the Belgian method where
he was in position to secure comparative figures, has given the
following rules for calculating the cost of tunnels. Assuming A to
represent the cost of excavating a cu. yd. in the open air, the cost of
excavating the same quantity underground in driving headings will
be from 9 A to 11 A, and in enlarging the profile it will be about 5 A.
The cost of constructing single-track tunnels varies with the
thickness of the lining, and may be calculated by the following
formulas:
Without lining, C = 5.5 A.
With roof arch only, C = 6.4 + 6.4 A.
With lining 18 in. thick, C = 9.4 + 7 A.
With lining 2 ft. thick, C = 11 + 8 A.

In these formulas C is the cost per cu. yd. of excavation,

including the masonry. For double-track tunnels the amounts given
by the above formulas may be used by reducing them about 71⁄2%
or 8%.
The second method of estimating the cost of tunnel work
consists in assuming as a unit the unit cost of tunnels previously
excavated under similar conditions. Mr. La Dame gives the following
unit prices for a number of tunnels driven through different
materials:

Tunnels, Excav. Cost per Max. and Min.

Nature of Soil.
No. of per Cu. Yd. Lin. Ft. per Lin. Ft.
Granite-gneiss 56 $3.07 @ $3.85 $100. $61.46 @ $190.40
Schist 39 1.38 @ 1.53 75.42 43.11 @ 70.68
Triassic 3 ... 90.85 84.75 @ 93.33
Jurassic 69 1.23 @ 1.38 77.86 35.24 @ 157.2
Cretaceous 34 0.61 @ 0.77 59.60 27.37 @ 92.25
Tertiary and modern 39 0.33 @ 0.61 105.80 51.52 @ 188.36

In the following table is given a list of tunnels excavated through

different soils, from the most compact to very loose materials, and
driven according to the various methods which have been illustrated.
DOUBLE-TRACK TUNNELS.

Cost per Method of

Name of Tunnels. Quality of Soil.
Lin. Ft. Tunneling.
Mt. Cenis Granitic, $273.73 Drift.
St. Gothard ... 193.63 Heading.
Stammerich Granitic, 157.90 English.
Stalle Broken schist, 290.58 Austrian.
Bothenfels Dolomite, 115.64 English.
Dorremberg Calcareous, 86.08 Belgian.
Stafflach Calcareous, 91.69 English.
Ofen Calcareous, 93.19 Austrian.
Wartha Grewack, 87.95 Austrian.
Mertin Grewack, 87.55 German.
Schloss Matrei Clay schist, 94.25 English.
Trietbitte Clay and sand, 229.0 German.
Canaan Clay-slate, 69.50 Wide heading.
Church-Hill Clay with shells, 178.0 ...
Bergen No. 1 Trap rock, 182.31 ...

SINGLE-TRACK TUNNELS.
 Cost per Method of
Name of Tunnels. Quality of Soil.
Lin. Ft. Tunneling.
Mt. Cenis Gneiss, $82.27 Heading.
Stalletti Granite and quartz, 62.75 Austrian.
Marein Clay schist, 64.36 English.
Welsberg Gravel, 165.07 Austrian.
Sancina Clay of 1st variety, 129.40 Belgian.
Starre Clay of 2d variety, 191.61 Belgian.
Cristina Clay of 3d variety, 307.42 Italian.
Burk ... 83.90 Wide heading.
Brafford Ridge ... 85.33 Wide heading.
Dunbeithe Limestone, 70.47 Wide heading.
Fergusson Sandstone, 37.46[16] Wide heading.
Port Henry Limestone, 80.00[17] Wide heading.
Points Granite, 72.00[16] Wide heading.

[16] Are unlined.

[17] Lined with timber.
The Habas tunnel through quicksand, between Dax and Ramoux,
France, cost $118.50 per lin. ft. The cost of the Boston subway was
$342.40 per lin. ft. The Severn and Mersey tunnels, constructed
through rock under water, cost respectively $208.38 and $263 per
lin. ft. The First Thames Tunnel, driven by Brunel’s shield, cost
$1661.66 per lin. ft. The Hudson River and St. Clair River tunnels,
excavated through soft ground by means of shields and compressed
air, cost respectively $305 and $315 per lin. ft. The Blackwall double-
track tunnel under the River Thames, which is the largest tunnel
ever built by the shield system, cost $600 per lin. ft.
In making estimates of the cost of projected tunnel work based
on the cost of tunnels previously constructed through similar
materials, it is important to keep in mind the date and location of the
work used as the basis for calculations. For example, a tunnel
excavated in Italy, where labor is very cheap, will cost less than one
excavated in America, where labor is dear, all other conditions being
the same. Other reasons for variation in cost due to difference of
date and location of construction will suggest themselves, and