Get ebook downloads in full at ebookmeta.

com

Data Driven Smart Manufacturing Technologies and

Applications 1st Edition Weidong Li

https://ebookmeta.com/product/data-driven-smart-
manufacturing-technologies-and-applications-1st-edition-
weidong-li/

OR CLICK BUTTON

DOWNLOAD NOW

Explore and download more ebook at https://ebookmeta.com

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

Data-Driven Mining, Learning and Analytics for Secured

Smart Cities: Trends and Advances (Advanced Sciences and
Technologies for Security Applications) Chinmay
Chakraborty
https://ebookmeta.com/product/data-driven-mining-learning-and-
analytics-for-secured-smart-cities-trends-and-advances-advanced-
sciences-and-technologies-for-security-applications-chinmay-
chakraborty/
ebookmeta.com

Smart Manufacturing Integrating Transformational

Technologies for Competitiveness and Sustainability 1st
Edition Scott M. Shemwell
https://ebookmeta.com/product/smart-manufacturing-integrating-
transformational-technologies-for-competitiveness-and-
sustainability-1st-edition-scott-m-shemwell/
ebookmeta.com

Thermal Energy Waste Recovery Technologies and Systems 1st

Edition Weidong Chen

https://ebookmeta.com/product/thermal-energy-waste-recovery-
technologies-and-systems-1st-edition-weidong-chen/

ebookmeta.com

Wolf Nurse (Shifter Medical Center #2) 1st Edition Sansa

Moon

https://ebookmeta.com/product/wolf-nurse-shifter-medical-center-2-1st-
edition-sansa-moon/

ebookmeta.com
Petroleum Refining: Technology, Economics, and Markets 6th
Edition Mark J. Kaiser

https://ebookmeta.com/product/petroleum-refining-technology-economics-
and-markets-6th-edition-mark-j-kaiser/

ebookmeta.com

Maya Imagery Architecture and Activity Space and Spatial

Analysis in Art History Maline D Werness Rude Kaylee R
Spencer
https://ebookmeta.com/product/maya-imagery-architecture-and-activity-
space-and-spatial-analysis-in-art-history-maline-d-werness-rude-
kaylee-r-spencer/
ebookmeta.com

Propalladia and Other Works of Bartolome de Torres

Naharro, Volume 4: Torres Haharro and the Drama of the
Rensaissance Joseph E. Gillet (Editor)
https://ebookmeta.com/product/propalladia-and-other-works-of-
bartolome-de-torres-naharro-volume-4-torres-haharro-and-the-drama-of-
the-rensaissance-joseph-e-gillet-editor/
ebookmeta.com

Nothing Ever Happens Here 1st Edition Sarah Hagger Holt

https://ebookmeta.com/product/nothing-ever-happens-here-1st-edition-
sarah-hagger-holt/

ebookmeta.com

The Mathematics of Patterns, Symmetries, and Beauties in

Nature

https://ebookmeta.com/product/the-mathematics-of-patterns-symmetries-
and-beauties-in-nature/

ebookmeta.com
Transnational Islam and the Integration of Turks in Great
Britain 1st Edition Erdem Dikici

https://ebookmeta.com/product/transnational-islam-and-the-integration-
of-turks-in-great-britain-1st-edition-erdem-dikici/

ebookmeta.com
Springer Series in Advanced Manufacturing

Weidong Li
Yuchen Liang
Sheng Wang Editors

Data Driven
Smart
Manufacturing
Technologies
and Applications
Springer Series in Advanced Manufacturing

Series Editor
Duc Truong Pham, University of Birmingham, Birmingham, UK
The Springer Series in Advanced Manufacturing includes advanced textbooks,
research monographs, edited works and conference proceedings covering all major
subjects in the field of advanced manufacturing.
The following is a non-exclusive list of subjects relevant to the series:
1. Manufacturing processes and operations (material processing; assembly; test and
inspection; packaging and shipping).
2. Manufacturing product and process design (product design; product data
management; product development; manufacturing system planning).
3. Enterprise management (product life cycle management; production planning
and control; quality management).
Emphasis will be placed on novel material of topical interest (for example, books
on nanomanufacturing) as well as new treatments of more traditional areas.
As advanced manufacturing usually involves extensive use of information and
communication technology (ICT), books dealing with advanced ICT tools for
advanced manufacturing are also of interest to the Series.

Springer and Professor Pham welcome book ideas from authors. Potential
authors who wish to submit a book proposal should contact Anthony Doyle,
Executive Editor, Springer, e-mail: [email protected].

More information about this series at http://www.springer.com/series/7113

Weidong Li · Yuchen Liang · Sheng Wang
Editors

Data Driven Smart

Manufacturing Technologies
and Applications
Editors
Weidong Li Yuchen Liang
Faculty of Engineering, Environment Faculty of Engineering, Environment
and Computing and Computing
Coventry University Coventry University
Coventry, UK Coventry, UK

Sheng Wang
Faculty of Engineering, Environment
and Computing
Coventry University
Coventry, UK

ISSN 1860-5168 ISSN 2196-1735 (electronic)

Springer Series in Advanced Manufacturing
ISBN 978-3-030-66848-8 ISBN 978-3-030-66849-5 (eBook)
https://doi.org/10.1007/978-3-030-66849-5

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2021
This work is subject to copyright. All rights are solely and exclusively licensed 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, expressed 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

The advance in the information and communication technologies, such as the

Internet of Things, industrial sensors, sensor networks, etc., have impacted
manufacturing profoundly. With the technologies, data generated from modern
manufacturing systems is experiencing explosive growth, which has reached over
100 EB annually. Manufacturing data contains rich knowledge and offers a
tremendous opportunity in the transformation of the conventional manufacturing
paradigm to the smart manufacturing paradigm.
In the past, physics-based methodologies have been actively investigated for
resolving manufacturing problems. However, it is challenging to developing
accurate physical mechanisms for analyzing and predicting complex physical
phenomena in manufacturing systems preciously in an efficient and cost-effective
means. With the proliferation of data acquired from manufacturing and the
significant increase of computational power, there is a need of developing artificial
intelligence and advanced big data analytics to complement physical science to
handle manufacturing data in supporting smart decision-making. For instance, in
recent years, the developments of innovative machine learning algorithms
especially deep learning algorithms are dramatic. The algorithms have been widely
applied to mine knowledge from manufacturing data to facilitate various smart
manufacturing applications. That is, smart manufacturing aims to take advantage of
the state-of-the-art artificial intelligent technologies to enhance the adaptability,
flexibility, robustness, and sustainability of physical manufacturing processes.
Meanwhile, with the technologies, effectiveness and intelligence in manufacturing
decision-making will be significantly improved. These methodologies and
algorithms can be summarized as data driven manufacturing technologies, which
will reinforce the competitiveness of manufacturing companies to address
increasingly dynamic markets and sophisticated customer needs.
Owing to the importance and rapid progress of data driven smart manufacturing,
it is essential to make a timely update on this subject. With this aim, this book
reports the relevant applied research conducted by the authors in recent years in
some important manufacturing applications, including intelligent diagnostics and
prognostics on manufacturing processes, sustainable manufacturing, prediction of
cutting tool life, heavy-duty machining applications, prediction of strength of
v
vi Preface

adhesive bonded joints, human–robot cooperation, etc. Majority of the first authors
in chapters are Ph.D. students, whose works have been supported by some
important research projects sponsored by the European Commission, the Innovate
UK (the UK industrial funding agency), the Institute of Digital Engineering of the
UK, and manufacturing companies from the UK and China. Meanwhile, some
works have been supported by the National Natural Science Foundation of China
(project no. 51975444) and international cooperative projects between the UK and
China. In order to make this book to be a systematic reference, some brief
theoretical foundations of deep learning algorithms, as well as the state-of-the-art
intelligent algorithms and practical cases on the aforementioned manufacturing
applications, are covered in the book. Case studies in this book illustrate design
details of intelligent algorithms and methodologies for various manufacturing
applications, in a bid to provide a useful reference to readers.
This book supplements the Springer books “Cloud Manufacturing” and
“Research on Intelligent Manufacturing” by providing technical discussions and
implementation details of data driven manufacturing technologies and applications.
We believe that the book offers a valuable resource for researchers in the smart
manufacturing communities, as well as practitioners, engineers, decision-makers in
industry and all those interested in smart manufacturing and Industry 4.0.

Dr. Weidong Li
Professor
Wuhan University of Technology
Wuhan, China
Professor
Coventry University
Coventry, UK
Dr. Yuchen Liang
Lecturer
Coventry University
Coventry, UK
Dr. Sheng Wang
Researcher
Coventry University
Coventry, UK
Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
W. D. Li, Y. C. Liang, and S. Wang
Fog Computing and Convolutional Neural Network Enabled
Prognosis for Machining Process Optimization . . . . . . . . . . . . . . . . . . . . . . . 13
Y. C. Liang, W. D. Li, X. Lu, and S. Wang
Big Data Enabled Intelligent Immune System for Energy Efficient
Manufacturing Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
S. Wang, Y. C. Liang, and W. D. Li
Adaptive Diagnostics on Machining Processes Enabled by Transfer
Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Y. C. Liang, W. D. Li, S. Wang, and X. Lu
CNN-LSTM Enabled Prediction of Remaining Useful Life
of Cutting Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
X. Y. Zhang, X. Lu, W. D. Li, and S. Wang
Thermal Error Prediction for Heavy-Duty CNC Machines
Enabled by Long Short-Term Memory Networks and Fog-Cloud
Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Y. C. Liang, W. D. Li, P. Lou, and J. M. Hu
Deep Transfer Learning Enabled Estimation of Health State
of Cutting Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
M. Marei, S. El Zaataria, and W. D. Li
Prediction of Strength of Adhesive Bonded Joints Based
on Machine Learning Algorithm and Finite Element Analysis . . . . . . . . . 173
Y. C. Liang, Y. D. Liu, and W. D. Li
Enhancing Ant Colony Optimization by Adaptive Gradient Descent . . . . 191
Y. Zhou, W. D. Li, X. Wang, and Q. Qiu

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
vii
About the Editors

Prof. Weidong Li is a full professor in manufacturing, Coventry University (UK).

Professor Li has more than 20 years of experience in computer-aided design,
manufacturing informatics, smart manufacturing, and sustainable manufacturing.
His research has been sponsored by a number of research and development projects
from the UK EPSRC, European Commission, and European industries such as
Jaguar Land Rover, Airbus, Rolls-Royce, Sandvik, etc. In the research areas, he has
published four books and around 200 research papers in international journals and
conferences.

Dr. Yuchen Liang is a lecturer from Coventry University. Dr. Liang has gotten his
Ph.D. degree from Coventry University from Automotive and Mechanical
Engineering. His research areas are data driven smart manufacturing and
cyber-physical manufacturing systems. His research works have been sponsored by
the European Commission and the Innovate UK.

Dr. Sheng Wang is a senior researcher in manufacturing, Coventry University,

UK. Dr. Wang got her Ph.D. degree from Queen Mary University of London from
Computer Science and Electronic Engineering. In the past five years, Dr. Wang has
participated in a number of European Commission-sponsored projects in smart
manufacturing.

ix
Introduction

W. D. Li, Y. C. Liang, and S. Wang

1 Smart Manufacturing Trend

In the past century, the manufacturing industry has undergone significant paradigm
shifts, from mass production (Ford assembly line) (1900s) to lean manufac-
turing (Toyota production system) (1960s), flexible manufacturing (mass customiza-
tion) (1980s), reconfigurable manufacturing (1990s), collaborative manufacturing
(2000s), and smart manufacturing (2010s) [1–3]. The development trend of the global
manufacturing industry is shown in Fig. 1.
In 2010, Germany proposed the framework of Industry 4.0 in transforming facto-
ries towards smart manufacturing [4]. Industry 4.0 interprets major manufacturing
paradigm shifts from the perspective of prominent breakthrough of enabling tech-
nologies. The 1st industrial revolution was signified by the invention of the steam
engine, which empowered mechanical machines to expand manufacturing capabil-
ities. The emergence of electrical energy indicated the 2nd industrial revolution,
through which mass production lines were electrified contributing to significantly
improved productivity, product quality repeatability and process control. Comput-
erized numerical controlling (CNC) machines and programming logic controllers
were some important technologies deployed in factories. The 3rd industrial revolution
witnessed the wide adoption of electronics, computers, programming languages, and
various industrial software systems (computer-aided design (CAD), computer-aided
manufacturing (CAM), computer-aided engineering (CAE), manufacturing execu-
tion systems (MES), etc.). Design and manufacturing processes were digitalized, and
close-loop control and programable/re-programable automation in manufacturing

W. D. Li (B)
School of Logistics Engineering, Wuhan University of Technology, Wuhan, China
e-mail: [email protected]
W. D. Li · Y. C. Liang · S. Wang
Faculty of Engineering, Environment and Computing, Coventry University, Coventry, UK

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 1

W. Li et al. (eds.), Data Driven Smart Manufacturing Technologies
and Applications, Springer Series in Advanced Manufacturing,
https://doi.org/10.1007/978-3-030-66849-5_1
2 W. D. Li et al.

1900s 1960s 1980s 1990s 2000s 2010s

Cost Quality Variety Responsiveness Competitiveness Smartness
Volume
per
product Lean manu-
variant
Mass pro- facturing
duction Flexible man-
ufacturing
Reconfigurable
manufacturing
Smart manu-
facturing
Collaborative
manufacturing Product variety

Fig. 1 Significant paradigm shifts of the global manufacturing industry

were realized. Lean, flexible, reconfigurable, customable and collaborative produc-

tion lines were subsequently implemented in factories. The 4th industrial revolution
was characterized by the proliferate use of industrial sensors, the Internet of Things,
machine learning, deep learning, big data analytics and cloud computing to establish
smart manufacturing. During the era, design of cyber-physical systems for smart
factories and processing of the vast amount of manufacturing data for smart decision
making were central topics. The review of Industry 4.0 and relevant technological
development can be found in [5,6] (illustrated in Fig. 2).
Apart from Industry 4.0, in recent years, strategic roadmaps of smart manufac-
turing have been recently developed by some major industrialized countries to take
advantage of the state-of-the-art information technologies. In 2011, the Smart Manu-
facturing Leadership Coalition (SMLC) in the U.S. created a systematic framework
of smart manufacturing to be implemented in American factories [6]. In 2014, the

1st Industrial Mechani-

revolution Power
zation
(1770s-)
Data acquisition
Control Human
2nd Industrial
revolution Electrifi-
cation Simulation Data presentation
(1900s-)
Machine Data
3rd Industrial
Digitali- Monitoring Data analysis
revolution
(1970s-) zation
Diagnostics
th Product Data security
4 Industrial Smart-
revolution ness Prognostics
(2010s-)

Fig. 2 Industry 4.0 and key technologies supporting manufacturing

Introduction 3

Manufacturing Industry Innovation 3.0 strategy was launched in Korea to realize

the Korean strategy of smart manufacturing. The plan “China Manufacturing 2025”,
initialized in China in 2015, aims to promote advanced manufacturing in China [7].
The National Institute of Standards and Technology (NIST) defined smart manufac-
turing as “fully-integrated, collaborative manufacturing systems that respond in real
time to meet changing demands and conditions in the factory, in the supply network
and in customer needs” [6]. According to the factories of the future roadmap of Euro-
pean Union, European factories should focus on key priorities under a common head-
line “co-creation through manufacturing eco-systems” [8], including: (1) respon-
sible and smart manufacturing systems; (2) manufacturing with zero-environmental
impact; (3) customer-driven value networks; (4) parallel product and manufacturing
engineering; (5) human-driven innovation.

2 Data Driven Smart Manufacturing

Physics-based methodologies have been actively investigated for manufacturing

problems. For example, physical mechanisms of manufacturing elements are
analyzed based on the Finite Element Method (FEM), Finite Difference Method
(FDM), or Finite Difference Element Method (FDEM). However, it is challenging
to developing accurate physical mechanisms for analyzing and predicting complex
physical phenomena in manufacturing systems preciously. With the proliferation
of data acquired from manufacturing and the significant increase of computational
power, there is a need of developing artificial intelligence and advanced data analytics
to complement physical science to handle manufacturing data for smart decision
making. In recent Gartner’s reports for strategic technology trends in the world,
artificial intelligence has been listed as a top strategic technology or the foremost
enabling technology in supporting smart manufacturing application [9].
The artificial neural network (ANN) is one of the most important progresses
in the artificial intelligence development. The initial design of ANN can be traced
back in 1940s, when simple ANN models were proposed to simulate how neurons
work in the human brain for linear optimization [10]. The models were implemented
using electrical circuits. In 1970s, the back propagation (BP) algorithm was designed
for ANN to solve non-linear problems in complex neural networks [11]. As more
sensors have been installed in factories’ shop floors, manufacturing has been getting
smarter, and data captured throughout manufacturing life cycles have been growing
exponentially. Accumulated data in manufacturing are often considered as big data
since they can be described in four Vs: Volume (there are a large amount of collected
data), Variety (collected data are in different forms or types), Velocity (data change
or are updated frequently), and Veracity (it refers to the biases, noise and abnormality
in collected data). ANN, however, could not converge in resolving such big data. To
tackle this issue, in the 1990s, deep learning models, which are extended versions of
ANN, were proposed.
4 W. D. Li et al.

Manual feature
Input Output
defintions

(a) Coventional machine learning processes

Input Output

(b) Deep learning processes

Fig. 3 Comparison between conventional machine learning and deep learning [12]

Deep learning models were considered as breakthrough techniques and demon-

strated outstanding performance in large-volume data processing in the areas of
speech recognition, image recondition, natural language processing, multimodal
image-text, games (e.g., Alphago), and various engineering and manufacturing appli-
cations. In comparison with ANN, support vector machine (SVM) and Boltzmann
machine, deep learning models allow processing highly non-linear and complex
data towards feature abstraction via a cascade of multiple layers and leaning mecha-
nisms. It overcomes the laborious process of a conventional machine learning model
to handcraft optimum feature representations and extract features from data. The
different processes between deep learning models and the conventional machine
learning models is illustrated in Fig. 3.
Deep learning offers a great potential to boost data driven smart manufacturing,
from product quality inspection to manufacturing process health management, and
manufacturing optimization in terms of improved energy saving, productivity and
quality performance, etc. For example, manufacturing systems are subject to various
failures or defects caused by excessive load, fracture, overheating, corrosion and
wear. Diagnostics and prognostics are core functions of manufacturing process
health management that allow manufacturers to optimize the conditions and perfor-
mance of manufacturing systems. A schematic diagram for some potential smart
manufacturing applications based on deep learning models is shown in Fig. 4.

3 Typical Deep Learning Models

3.1 Some Development of Deep Learning Models

Deep learning is a branch of machine learning that is an extended version of ANN in

mimicking the human brain. A deep learning model processes data through multiple
Introduction 5

Cost, Time, Efficiency, Energy use, Performance, etc.

Knowledge
Inspection Diagnosis Prognosis Optimization
Knowledge of Analysis of Prediction of Improvement
product, root causes of quality, per- of quality,
Deep Learning models

equipment and faults and formance and productivity

process failures patterns and energy
Features

Network Technologies (Industry Internet, Wireless, etc.)

Data
Sensors, IoTs

Human Equipment ERP

MES
Product Process SCADA

Fig. 4 Smart manufacturing and deep learning models

(deeper) layers of neural network, each of which passes features recognized from data
into the next layer. Based on deep learning models, features from a vast amount of data
can be extracted and represented autonomously rather than the process of a conven-
tional machine learning model by which explicit engineered features are used to
perform for this process. That is, deep learning is an end-to-end learning model to inte-
grate feature learning and model construction with the minimum human inference.
The architecture of a deep learning model is composed of multiple hidden layers to
conduct multi-level and non-linear computational operations. In each layer, features
are extracted from the original input and transferred into more abstracted features
in higher layers to identify complicated inherent relationships in the original data.
These abstracted features are then input into a final layer to perform classification,
clustering or regression tasks.
From the mid of 1990s, there are important progresses in deep learning develop-
ment, and some typical deep learning models were proposed. In 1995, the recurrent
neural network (RNN) was designed [13]. In RNN, the outputs from neurons are
used as feedback to the neurons of the previous layer recurrently. The architecture
of RNN is naturally suited to processing time-series data and other sequential data
for prediction or control tasks.
In 1997, an improved version of RNN, namely the long short-term memory
(LSTM), was developed to tackle the vanishing and exploding gradient problems
in RNN to process complex time-series data [14].
In 1998, the convolutional neural network (CNN) was devised to facilitate images
or other two-dimensional information, in which feature learning was achieved by
stacking convolutional layers and pooling layers [15]. The success of a deep convo-
lutional architecture called AlexNet in the 2012 ImageNet competition (the annual
6 W. D. Li et al.

Olympics of computer vision) was the shot heard round the world. Using the AlexNet,
the classification error was dropped from 26 to 15%, an astounding improvement at
the time [16]. Ever since then, a host of companies have been using deep learning at
the core of their services. Facebook uses deep learning for their automatic tagging
algorithms, Google for their photo search, Amazon for their product recommenda-
tions, Pinterest for their home feed personalization, and Instagram for their search
infrastructure. CNN has powered major advances in computer vision, which has wide
applications, such as self-driving cars, robotics, drones, security, medical diagnoses,
and treatments for the visually impaired.
From 2000s, deep learning has gained wider popularity, and more new models
were designed. In 2006, the deep belief network (DBN) was developed [17]. DBN
uses probabilities and unsupervised learning to produce outputs. It contains both
undirected layers and directed layers. Unlike other deep learning models, each layer
in DBN learns the entire input, while in CNN, the first layers only filter inputs for
basic features, such as edges of an image, and the later layers recombine all the simple
features identified by the previous layers. DBN, on the other hand, works globally
and regulates each layer in order. Greedy learning algorithms are used to pre-train
DBN. This is a problem-solving approach that involves making the optimal choice at
each layer in the sequence, fine-tuning the model parameters and eventually finding
a global optimum.
In 2009, the deep Boltzmann machine was designed to have a bidirectional struc-
ture to learn ambiguous input, and the model parameters were optimized using
layer-wise pre-training efficiently [18].
In 2014, the generative adversarial network (GAN) was devised to learn from
a set of training data and generate new data with the same characteristics as the
training data [19]. GAN consists of two neural networks, i.e., the generator and the
discriminator, which compete against each other. The generator is trained to produce
fake data, and the discriminator is trained to distinguish the generator’s fake data from
real examples. If the generator produces fake data that the discriminator can easily
recognize as implausible, such as an image that is clearly not a face, the generator
is penalized. Over time, the generator learns to produce more plausible examples.
GAN has been used for some applications, such as image editing, animation model
generation, prognostics and heath management in manufacturing.
Deep learning models are still in development. For example, to address issues
of CNN, improved models were developed, including regions with CNN (R-CNN)
(2014) [20], fast R-CNN (2015) [21], capsule networks (CAPSNet) [22], etc.

3.2 Typical Deep Learning Algorithms

A deep learning model is generally classified as a supervised and an unsuper-

vised learning model. A supervised learning model is trained to classify data or
predict outcomes accurately based on pre-defined (supervised) labeled datasets. In
Other documents randomly have
different content
 The Illustrations.

J. L. SIMPSON, 37. THE GIRLS’ CLUB.

Architect.

The full-page Plates 3 to 27 give an idea of the road views and the
relation of the houses to them, with their perspective effects in
straight or winding lines of frontage, with quadrangular recesses as
in Pls. 11 and 20, or where the L-shaped blocks of cottages leave a
good open space as in Pl. 9. The sturdy tower of Christ Church is a
telling feature at the end of the Causeway (see Pl. 13), and the
picturesque pavilion roof with its clever tiling makes a telling feature
at the junction of the Causeway with Greendale Road. It is hard to
imagine anything more delightful in the early spring or autumn than
the Greendale Road where it approaches the post office, with the
peeps of the buildings through the tall poplars. The view towards the
post office in both directions (Pls. 3 and 7) are equally pleasing.
Nothing shows better the good qualities of an Old English half-timber
building than such a setting. There is hardly anything in the village
which comes back on one with such recurring charm as the row of
five gables in the Park Road cottages shown in Pls. 5 and 14. In Pl.
5 we see something of the delightful result of the continuous sloping
banks from the road up to the cottage, and a certain picturesque
irregularity where the old hawthorn bushes which formerly existed
have been left at intervals. This vertical timber framing has a simple
breadth of effect which is well shown in Pls. 4, 5, and 14. It would be
difficult to do justice to Greendale Road with its continuous line of 97
cottages, which form a picture of great variety and interest as viewed
from the passing trains, and give us a long perspective of trees and
houses, broken at the point of view of our sketch (Pl. 15) by the half-
timber group of cottages which is an exact replica of the design of
Kenyon Old Hall. This delightful group is also shown in Pl. 20. There
is nothing more satisfactory in proportion and colour than the
recessed group of cottages which fronts the Diamond in the Queen
Mary’s Drive. The yellow-grey stone slates, the red brick chimneys,
the white rough cast, dark boarding, and robust half-timber work in
the flanking gables, make up a picture of colour and texture which is
most satisfactory (Pl. 2). The cottages here have the advantage of a
raised terrace bounded by a stone wall. One of the nearest
approaches to the charm of an old English village is probably the L-
shaped group of cottages in Bath Street (see Pl. 9). A photograph of
the Dell has been taken to show one of the natural features which
has been turned to so good an account in the village (see Pl. 10).
The cottages looking over the roadways surrounding the Dell have
delightful outlooks over here. Our view was taken in the winter, so as
to show something of the bridge and houses. The stone bridge at the
end of the Dell is an excellent architectural feature (see Nos. 1 and
2), and groups with remarkably good result below the Lyceum
buildings.
No illustrative account of Port Sunlight could be considered
complete without some reference to Christ Church, which is a central
and interesting feature. Its solidly built red stone walls and stone
slated roof, and its finely appointed interior with a wealth of enriched
oak timber work, commands one’s attention whether as architect or
layman. It speaks of strength and endurance and a sincere love and
study of our traditional English Gothic. Its value is sufficiently
apparent both from the social and the artistic point of view. In one
respect, however, this church may claim a special distinction, for at
its western end has been erected a richly detailed narthex, with a
vaulted roof, forming a shrine for a beautiful sculptured memorial to
the late Lady Lever. Both in idea and execution, this forms a striking
and touching memorial to a gracious lady whose kindliness of heart
endeared her to all. Children were her special friends, and this is
reflected in the two charming figures of children at one end of the
sarcophagus. Sir Goscombe John, the sculptor, has never been
more successful than in this tenderly and gracefully modelled
reclining figure of Lady Lever. This vaulted porch, with its richly
carved bosses (on one of which are painted the arms of Sir William),
largely enhances the value of the memorial sarcophagus itself, which
is one of the most satisfactory of recent years. The illustration of the
interior (see Pl. 33) is from a large drawing exhibited by the architect
at the Royal Academy in 1916.
Our illustrations, in a general way, represent what may be taken to
be the best examples of design in the village. They do not, of course,
show all the best. In Pl. 28 we have a very good example of the
quality of detail which lifts the work at Port Sunlight so far above the
level of the ordinary speculative cottage building. Here we find
carved oak beams and posts and brackets and barges, and an
excellent piece of modelled plaster work in the gable. When it is
remembered that this is no isolated example, we see how unusually
liberal has been the hand that directed the outlay. Corners like the
picturesque grouping of chimneys in No. 28, or the carved oak and
modelled plaster in the corner gable (No. 9), would not have existed
in an industrial village had not the founder been imbued with a keen
appreciation of architectural values. One would present a sketch
proposal for such a type of cottages with some trepidation to the
average building owner! One of the noticeable bits of rich detail is to
be found in the Flamboyant and Gothic dormers in Pl. 27. We have
in No. 32 a delightful corner of half-timber building with a sweet little
garden foreground. The old Cheshire type of half-timber work is
tellingly expressed in the corner houses in Park Road (Pl. 16). Other
especially effective corners are seen in Nos. 11 and 27. A contrast
between Queen Anne brick gables and the half-timber house is
effectively shown in No. 21. Contrasting again with the richness of
carved oak and modelled plaster in the more elaborate buildings, we
come across delightfully simple designs, such as Nos. 15 and 17,
which may some day very well pass for ancient buildings.
 Amongst the conspicuously successful of recent groups is that of
the parlour houses in Bolton Road (Pl. 21), which has the advantage
of a good setting on the front of a circular place. This only needs a
terrace wall and some formal planting to make it one of the pictures
of the village. What an enticing prospect opens up in the possibilities
of formal evergreen planting amongst all these cottage homes!
It is with some feeling of regret that more sketches have not been
given showing examples of interesting ornamental detail which lift
the quality of these cottage homes so much above the ordinary level
of industrial homes. But the limits of the volume place an inevitable
check on one’s desires. If the author has been able to convey to his
readers a tithe of the pleasure he has felt in the subject of this little
book he will be amply rewarded.

38. AN EXAMPLE OF SIMPLE TREATMENT.

Enough has perhaps been said in appreciation of some points

which mark out the qualities of Port Sunlight. Much more might be
written, but for the rest we leave the illustrations to tell their own tale.
We have shown nothing of the great auditorium which seats 3,000
people, or of the detailed appointments of the new Art Gallery, but
we have a sketch of the splendid open air swimming pool and our
birds-eye view of the Diamond gives some notion of the pleasant
grouping of the homes amongst pleasant open spaces and tree-lined
avenues. A great deal of what we do not show is left in reserve for
those who can find opportunity to go and see for themselves what
this wonderful scheme of industrial housing can teach.

PLAN OF THE VILLAGE OF PORT SUNLIGHT 1914.

39. A GENERAL PLAN OF THE VILLAGE.
 PLATE 2.

THE DIAMOND, LOOKING TOWARDS ART GALLERY.

 PLATE 3.

GREENDALE ROAD, LOOKING TOWARDS POST OFFICE.

POST OFFICE. LEVER FREE LIBRARY.
 PLATE 4.

WESTWARD VIEW OF PARK ROAD, SHOWING KITCHEN

COTTAGES.
 PLATE 5.

VIEW IN PARK ROAD TOWARDS THE LYCEUM.

 PLATE 6.

SOUTH SIDE OF PARK ROAD.

 PLATE 7.

GREENDALE ROAD AND CO-PARTNERS’ CLUB ANNEXE.

 PLATE 8.

NEW CHESTER ROAD.

 PLATE 9.

COTTAGES IN BATH STREET.

 PLATE 10.

THE DELL.
 PLATE 11.

SEMI-QUAD OF COTTAGES, QUEEN MARY’S DRIVE, THE

DIAMOND.
 PLATE 12.

PARK ROAD. BRIDGE COTTAGE IN FOREGROUND.

 PLATE 13.

THE CAUSEWAY, LOOKING TOWARDS CHRIST CHURCH.

 PLATE 14.

PARK ROAD COTTAGES. LYCEUM IN DISTANCE.

 PLATE 15.

GREENDALE ROAD. COTTAGE GROUP IN FOREGROUND,

REPRODUCING KENYON OLD HALL.
 PLATE 16.

HOUSES IN PARK ROAD. SOUTH SIDE.