International Journal of Computer Integrated

Manufacturing

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tcim20

Application of machine learning algorithm in the

sheet metal industry: an exploratory case study

Ahm Shamsuzzoha, Timo Kankaanpaa, Huy Nguyen & Hoang Nguyen

To cite this article: Ahm Shamsuzzoha, Timo Kankaanpaa, Huy Nguyen & Hoang
Nguyen (2021): Application of machine learning algorithm in the sheet metal industry: an
exploratory case study, International Journal of Computer Integrated Manufacturing, DOI:
10.1080/0951192X.2021.1972469

To link to this article: https://doi.org/10.1080/0951192X.2021.1972469

© 2021 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group.

Published online: 03 Sep 2021.

Submit your article to this journal

Article views: 1024

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tcim20
INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING
https://doi.org/10.1080/0951192X.2021.1972469

Application of machine learning algorithm in the sheet metal industry: an

exploratory case study
a
Ahm Shamsuzzoha , Timo Kankaanpaab, Huy Nguyenc and Hoang Nguyenb
a
School of Technology and Innovations, University of Vaasa Po Box 700, Vaasa, Finland; bDepartment of Information Technology Vaasa,
University of Applied Sciences Vaasa, Finland; cDepartment of Information Technology Vaasa , University of Applied Sciences Vaasa, Finland

ABSTRACT ARTICLE HISTORY

This study solved a practical problem in a case in the sheet metal industry using machine learning Received 30 December 2020
and deep learning algorithms. The problem in the case company was related to detecting the Accepted 03 August 2021
minimum gaps between components, which were produced after the punching operation of KEYWORDS
a metal sheet. Due to the narrow gaps between the components, an automated sheer machine Machine learning; deep
could not grip the rest of the sheet skeleton properly after the punching operation. This resulted in learning; detection of gaps;
some of the scraped sheet on the worktable being left behind, which needed a human operator to sheet metal industry; case
intervene. This caused an extra trigger to the production line that resulted in a break in production. study
To solve this critical problem, the relevant images of the components and the gaps between them
were analyzed using machine learning and deep learning techniques. The outcome of this study
contributed to eliminating the production bottleneck by optimizing the gaps between the
punched components. This optimization process facilitated the easy and safe movement of the
gripper machine and contributed to minimizing the sheet waste.

1. Introduction learning in the area of manufacturing can be very

The advance of technological trends has driven the fruitful (Sung 2018; Sharp, Ak, and Hedberg 2018;
rapid diffusion of machine learning across industries Dutta, Kaliannan, and Subramaniam 2021). In today’s
(Sarkar et al. 2016; O’Donovan et al. 2019; Gottsegan, business, the application of machine learning is con­
2020; Cakir, Guvenc, and Mistikoglu 2021). Today’s sidered to be one of the most disruptive innovations
growth of technological, societal, and competitive and can be a strong enabler of competitive advantage
pressures is pushing companies to transform and (Ren 2021; Cakir, Guvenc, and Mistikoglu 2021). The
innovate. Due to the rapid progress of machine- application of machine learning has been around for
learning technologies with their added benefits, orga­ more than 60 years; however, only in recent time has
nizational managers at all levels need to familiarize it showed significant potential for enhancing econo­
themselves with these techniques. Such organiza­ mies and societies (Lee and Shin 2020).
tional changes contribute to ensure creating maxi­ There are various applications of machine learning
mum value for companies. However, there is methods that have been applied in different industrial
a common challenge to the usability of machine- sectors in recent years. These methods have been
learning methods in companies, where a trade-off applied in various industrial segments such as health­
between interpretability and accuracy is needed care, banking, manufacturing, transportation, etc (Li,
(Srdoc et al., 2007; Sarkar et al. 2016). It is important Wang, and Wang 2019; Xie 2020; Gupta et al. 2020;
to understand the decisions made by machine- Danton et al. 2020). Studies have shown that the
learning-based techniques, which should also be adoption of machine learning algorithms can contri­
transparent and interpretable. bute to reducing costs by between 20%–25% across
Machine learning is a technique that is concerned banking, IT operations, infrastructure, and mainte­
with enabling computer programs automatically with nance, improve customer retention and acquisition,
the objective of improving the performance of some etc. (Accenture, 2018). By transforming business pro­
tasks through experience. The application of machine cesses into intelligent and automated processes,

CONTACT Ahm Shamsuzzoha ahsh@uva.fi School of Technology and InnovationsUniversity of Vaasa Po Box 700, FI-65101 Vaasa, Finland
© 2021 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-
nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built
upon in any way.
2 A. SHAMSUZZOHA ET AL.

companies can more effectively utilize their resources manufacturing industries, it is essential nowadays to
for higher-value activities and be able to offer better introduce digital tools and technologies into this seg­
products and services to their customers. ment. This digitalization process results in an enor­
This study is focused on maintaining the punching mous amount of valuable data, which needs to be
operation of a sheet metal company in Finland. This analyzed in order to improve production processes
company operates on a global scale to deliver custo­ and to detect the causes of problems. In this environ­
mized components produced from sheets. It currently ment, machine learning and deep learning can be
has problems in sheet material handling after the useful in saving energy, time, and resources, and
necessary punching operations. The company has avoiding waste (Tsai and Chang 2018; Weichert et al.
noticed that frequently after punching operations 2019). Machine learning, which is a branch of artificial
waste sheets are stuck over the worktable because intelligence, is used to progressively enhance the
the automated grippers fail to grip the sheets com­ performance of tasks based on big data collected in
pletely and human operators are needed to remove the digitalized world (Bianconi et al. 2014; Li, Wang,
the remaining parts of the sheets and put them into and Wang 2019; Carbo-Valverde, Cuadros-Solas, and
trash bins. This often creates a halt in production and Rodríguez-Fernández 2020).
thus a loss in production for the company. This pro­ It is a new go-to strategy to add value in business
blem occurs mostly due to too narrow gaps being processes through the digitalization of workflows
produced between the components following the using machine learning and advanced analytics (Li,
punching operation. To overcome this problem, Wang, and Wang 2019; Hajizadeh 2019; Kang, Catal,
machine learning and deep learning algorithms and Tekinerdogan 2020). Companies are striving to
were used to study the gaps between components embrace the new digital technologies, but find it
with the objective of optimizing the gaps. challenging to put their models into production in
The rest of the article is organized as follows: order to achieve tangible outputs and gain favorable
Section 2 comprises a literature review on machine benefits. Due to the rapid growth in digitalization,
learning, and the methodology of the study is pre­ machine-learning algorithms offer high potential for
sented in Section 3. A description of the case com­ the business process optimization domain.
pany follows in in Section 4, and the overall study Nowadays, machine-learning algorithms are widely
results are analyzed and presented in Section 5. implemented in the production environment
Various practical outcomes are elaborated in terms (Zhang, Wang, and Gao 2019). The application of
of managerial implications at Section 6. The study machine learning is encouraged by industrial commu­
concludes in Section 7 by presenting its contributions nities due to its additional capabilities to save on
and limitations its and suggestions for future resources, machining time and energy, and increase
research. yield in areas where traditional methods such as six
sigma strategies have reached their limits (Köksal,
Batmaz, and Testik 2011; Golkarnareji et al., 2019;
2. Literature review Ren 2021; Xu et al., 2021).
The emergence of digitalization in manufacturing
2.1 Digitalization of industries and machine
processes has attracted industries and evolved over
learning
recent decades. The traditional concept of computer-
Digitalization means the incorporation of digital tech­ integrated manufacturing is transferred to digital
nology into all aspects of everyday life of industrial manufacturing, where the technologies and tools
communities. Today’s digital era emphasizes the facilitate the integration of product and process
importance of integrating digital transformation in design before starting actual production and support
industries in order to improve the overall efficiency the ramp-up phases . In digital manufacturing, parti­
of the business processes. The digital transformation cular emphasis is given to the optimization of net­
of business processes provides an unprecedented worked production facilities, where real time data is
flow of high-quality information to industries by necessary for the decision-making process. Machine
using automated real-time data analytics and process learning, which is a diverse field of artificial intelli­
analysis. Because of the digitalization of gence, has the ability to automatically learn from
 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 3

data and make predictions based on data. This tech­ 2017). Problems associated with machine learning
nique has been widely used in digital manufacturing methodology need to be addressed at regular
for various purposes, such as predictive maintenance intervals.
(Joseph et al. 2014; Dalzochio et al., 2020), demand
forecasting (Huber and Stuckenschmidt 2020; Jayant,
2.2 Application of machine learning in various
Agarwal, and Gupta 2021) to process monitoring
industrial segments
(Cakir, Guvenc, and Mistikoglu 2021), and optimiza­
tion (Weichert et al. 2019; Min et al. 2019). Machine learning methodology is applied in various
The term machine learning describes algorithms fields of research work. Yeh and Deng (Yeh and Deng
used to identify and extract valuable data patterns 2012) applied machine-learning methodology in cost
to improve quality. The implementation of machine estimation of the product life cycle. They used two
learning is not a new field. However, with the recent machine-learning methods, namely back-
growth of computational power, many industries propagation neural networks (BPNs) and least squares
have been considering employing artificial intelli­ support vector machines (LS-SVMs) in order to solve
gence solutions to improve production processes product life cycle cost estimation problems.
(Weichert et al. 2019; Ademujimi, Brundage, and Carbonneau, Laframboise, and Vahidov (2008) used
Prabhu 2017). The development of machine learn­ advanced machine learning to investigate the applic­
ing has been very successful at tackling complex ability of forecasting distorted demand at the end of
tasks (Gottsegan, 2020). It is also noticeable that a supply chain (bullwhip effect). The authors com­
although machine learning and many deep- pared this method with other available methods,
learning methods are known to create highly accu­ such as naïve forecasting, trend, moving average,
rate outcomes, they often lack interpretability and linear regression, and showed that the machine
owing to their black-box nature (Lee and Shin learning method produced a better performance than
2020). It is therefore critical, and often challenging, the regression model. Machine learning-based algo­
to identify the most appropriate machine-learning rithms have also been applied to manage the human
method for specific applications. There are other resource systems of intelligent manufacturing indus­
challenges also for organizational managers when try. The application of machine learning in human
deploying machine learning, such as ethical issues, resource management in enterprises has been inves­
a shortage of machine-learning engineers, and tigated by matching, screening and filtering the user’s
issues of data-quality cost-benefit (Lee and Shin job characteristics and the employee’s requirements
2020). with the objective of improving the effective pursuit
of the most suitable goals for both users and enter­
Although machine learning provides new opportu­
prises (Xie 2020; Garg et al. 2021).
nities in terms of supporting a data-rich digital man­
The application of machine learning algorithms can
ufacturing environment, it often gives rise to
significantly improve the operation and development
additional challenges. The machine learning algo­
of manufacturing industries (Sung 2018). The automo­
rithm requires compliance with defined standards,
tive industry is a leading user of machine learning to
which is required for data science and optimization
assess and minimize risk. Such risk assessments are
techniques (Parkes, Özcan, and Karapetyan 2015). In
conducted by identifying the risks first, which is suc­
addition, there is also a challenge in finding expertise
cessfully performed by adopting machine-learning
related to data science and optimization areas, which
algorithms (Hegde and Rokseth 2020). Financial busi­
requires an inherently different skill set (Baumers and
nesses such as banks adopt machine learning in order
Özcan 2016). Moreover, there are challenges to imple­
to analyze various transaction scenarios, to enable the
menting machine learning such as the selection of the
processing of large data sets faster, and to mak accu­
right algorithm, the right set of data, data processing,
rate forecasts for various automated trading function­
data labelling, etc. Furthermore, there are several
alities (Lee 2017: Rizzi, Wang, and Zielinski 2020).
other challenges during the implementation process
Weichert et al. (2019) reviewed the application of
such as managing the model version, redeveloping
machine learning in industries to optimize methods
the models, managing data versions, etc. (Gupta
on the shop floor in order to improve production
4 A. SHAMSUZZOHA ET AL.

processes in the manufacturing industry. Additionally, strategy for trading orders (McDowell 2018). Table 1
they also discussed how the machine-learning highlights the application of machine learning in
approach used in industries saves energy, time, and some important industrial sectors.
resources, and avoids waste.
In the case of service industry, machine-learning algo­
2.3 Application of machine learning in sheet metal
rithms have also been successfully applied. applied
industries
machine learning in the healthcare industry to identify
ethical considerations, while Gupta and Sedamkar There are only a few applications of machine learning
(2020) deployed machine learning to correctly identify in sheet metal industries. Machine learning and deep
the percentage of sick versus healthy people. For the learning algorithms are successfully applied to sheet
tourism and hospitality industry, machine learning has metal industries to select a suitable manufacturing
shown significant potential. Sun et al. (2019) used process and to achieve the final geometry of a metal
a machine learning algorithm to forecast the number part that is unstructured and heavily reliant on human
of tourists arriving in a certain period. They proposed expertise (Hamouche and Loukaides 2018; Chiu, Tsai,
a forecasting framework that applies a combination of and T-l 2020). Stoerke et al. (2016) proposed a new
machine learning and internet search, with the aim of methodology to increase the geometric accuracy
improving the overall forecasting performance with forming of sheet metal work pieces. They proposed
respect to both forecasting accuracy and robustness the use of a machine learning model, which applies
analysis. Sanchez-Medina and C-Sanchez (2020) used reinforcement learning as a flexible and promising
machine learning to successfully forecast hotel booking solution to increase the geometric accuracy in incre­
cancellations. Other areas of machine learning applica­ mental sheet forming processes.
tions can be seen in transportation (Huang and Zhu Machine learning is also used to predict the defects of
2021), flood prediction (Motta et al., 2021), telecommu­ sheet metal forming processes (Deb, Ribeiro, and Prates
nications (de Andres, 2020), smart city (Zekić-Sušac, 2018; Tsai and Chang 2018). Kwitek (2016) demon­
Mitrović, and Has 2021), amongst others. strated how sheet metal fabrication machines can use
Machine learning algorithms are used in industries machine learning in order to gain competitive advan­
for various purposes such as clustering, classification, tage. He used machine learning to focus on predictive
forecasting, etc. In the case of clustering, objects are maintenance, which contributed to improving opera­
grouped in order to reveal patterns, such as feedback tions management. Hamouche and Loukaides (2018)
from groups of customers to improve customer satis­ applied machine learning to sheet metal forming,
faction. Unsupervised machine learning is used in which is considered a critical component of modern
clustering due to the class labels of some objects manufacturing. In their study, a machine learning
not being known beforehand (Najafabadi, 2017). For approach was used for the first time with the objective
classification processes, all the observations are of identifying the manufacturing process, which formed
already categorized or classed for purposes of training a part solely from the final geometry. By implementing
and testing. Machine learning is helpful in classifying a mapping of the mean and Gaussian curvatures
the business operations. Chandran (2018) studied through machine learning, a high accuracy rate was
a brewing company using a machine-learning classi­ established that automated the operational processes
fier to determine better routes for its drivers. Musani from the initial stage as design to manufacture, thus
(2018) deployed a machine learning based freshness eliminating the requirement for human experts in
algorithm to prioritize the flow of perishable goods matching each product with a suitable forming method.
worldwide. In addition to clustering and classification, Kashid and Kumar 2012) performed a review of the
machine learning is also deployed to identify data applications of artificial neural networks to sheet
patterns that are used to predict future events. For metal work. Zwierzycki, Nicholas, and Thomsen
instance, machine learning can be used to scrutinize (2018) applied pre-process supervised machine learn­
data to detect market signals that will affect future ing to predict and improve sheet-forming tolerance
market performance (Chandran 2018). The JP Morgan and generate corrected fabrication models. Lin and
trading team has developed a proprietary machine- Chang 1995) proposed a model using machine learn­
learning model in order to find the best execution ing from neural networks in an expert system of sheet
 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 5

Table 1. Application of machine learning algorithms in various industrial sectors.

Serial
number Field of application Specific contribution Reference
1 Architectural design Machine learning algorithm is used to assess two cases of architectural practices with Tamke et al. (2018)
performance-based design, fabrication and learning.
2 Steel industry Artificial intelligence and machine learning are applied to develop plant coordination, Backman, Kyllönen, and
control, raw materials, energy optimization and quality management. Helaakoski (2019)
3 Process industry Review of future research on data mining and analytics conducted for the benefit of Ge et al. (2017)
process industry.
4 Food industry Machine learning was implemented to predict deviations in food production, Garre, Ruiz, and Hontoria (2020)
reducing uncertainties and minimizing the amount of waste of food. It also works
as a tool to help to identify possible production anomalies.
5 Construction industry Machine learning algorithm was applied to execute a successful contract between Valpeters et al. (2018)
stakeholders. It also revealed the factors influencing the possibility of contractual
default and tried to define corrective actions for a customer.
6 Industrial pumps Anomalies among industrial pumps are detected by machine learning algorithm that Dutta, Kaliannan, and
supports condition monitoring and prevents pump failure. Subramaniam (2021)
7 Machine industry The possibility of utilizing machine learning in machining processes is reviewed in Kim et al. (2018)
order to improve product quality levels and productivity rates and to optimize
design and process parameters.
8 Petrochemical industry A framework based on digital twin and machine-learning approach is presented to Min et al. (2019)
support petrochemical and other process manufacturing industries to dynamically
adapt to the changing environment and to respond in a timely manner to improve
economic benefits.
9 Mineral processing A thorough review on the applicability of machine learning in mineral processing McCoy and Auret (2019)
with respect to success level, area of application and major problems is presented
with a view to equipping researchers and industrial practitioners with structured
knowledge.
10 Construction industry This research study applies machine-learning technique to analyze 16 critical factors Zhu et al. (2020)
with the aim of assessing the impact of the identified factors in predicting the
severity of construction accidents.
11 Oil and gas industry A machine-learning algorithm was applied to cluster oilfields which are very similar, Patel and Patwardhan (2018)
which contributed to significantly reducing the engineering effort and operator
involvement in developing the models for each well.
12 Textile industry In addition to machine learning algorithms, data mining studies, including Yildirim, Birant, and Alpyildiz
classification and clustering techniques, are implemented in the textile industry to (2018)
deal with different problems where traditional methods are not useful.
13 Power industry Machine learning methods are applied to select the most effective management Massel et al. (2019)
actions (preventive measures) and to prevent a critical situation from developing
into an emergency in a power system.
14 Healthcare industry In order to reduce healthcare cost and to improve patient safety and healthcare Nithya and Ilango (2017)
quality, machine learning can be used as an important tool or technique.
15 Pottery industry A number of unsupervised and supervised machine learning methods are applied to Anglisano et al. (2020)
test the applicability of defining geochemical fingerprints to track or attest the
provenance of samples.

metal bending tooling. In this study, the authors shortage of experts, cost benefit analysis etc. In
developed a learning model with conditional attri­ terms of ethical challenges, the application of
butes to develop an expert system of sheet metal machine learning is concerned with data privacy
bending tooling selection. Wu et al. (1999) deployed and protection rules (e.g. personal medical record
a combination of artificial neural network and data). During algorithm development for machine
machine learning methods for investigating surface learning, training data is needed, which may exhi­
defects in sheet metal forming. bit an inevitable bias if there are any latent biases
in that data (e.g. bias accompanied with gender,
age, race) (Sharp, Ak, and Hedberg 2018). In order
2.4 Related challenges to implementing
to develop an ethical machine learning algorithm,
machine-learning methods in industries
cooperation between researchers, developers and
There are several challenges in deploying policy makers is needed. Turilli 2020)recom­
a machine-learning approach in industries, such mended that machine-learning algorithms should
as data privacy, data accessibility, and data shar­ reflect the same ethical principles as human work­
ing. In addition, there are other related challenges ers to ensure consistency with an organization’s
such as the quality of data, ethical issues, ethical standards.
6 A. SHAMSUZZOHA ET AL.

In addition to the ethical challenge, there is 3. Study methodology

a shortage of machine learning engineers. The
This study was conducted in a case company,
demand for artificial intelligence (AI) talents has
engaged in sheet metal work, where the problem
increased fast in recent years, resulting in a high num­
was to identify the optimal gaps between compo­
ber of AI-related vacancies (Culbertson, 2018). In the
nents in order to more easily handle the waste sheet
American job market, machine-learning engineers are
after the punching operation. To identify and visualize
listed as one of the top emerging professions
the gaps, both machine learning and deep learning
(LinkedIn, 2017). It will take years to adequately
algorithms were adopted to analyze images of the
meet the demand for machine learning engineers in
sheets provided by the company. These algorithms
industry (Woolf and McIntyre 2018).
were deployed as part of the image processing tech­
The challenge related to data quality refers to the
nique to visualize the gaps clearly. The provided
fitness of data for specific applications of machine
images were converted to suitable ones, which
learning. It is important to have high-quality data for
could easily be analyzed by using machine and deep
the successful application of machine learning to
learning algorithms.
solve problems. The performance of machine learning
In order to study the gaps, all images of the
or deep learning highly depends on the data quality
components over the sheet were analyzed. The
(Greenspan, van Ginneken, and Summers 2016). If the
study of these images helps to identify which
study data are collected from various sources in an
parts of the images are actual components, and
unstructured format, the quality of the data deterio­
which parts are waste materials to trash. This gap
rates quickly. Machine learning algorithms demand
optimization process also helps the case company
structured data, although the majority of data are
to maintain its operational schedule between
unstructured (e.g. voice, text, image data, etc.),
the day and night production lines by employing
which are often difficult to process (Lee 2017). To
the most critical, and non-critical components
get fruitful outcomes from machine learning algo­
based on the gaps. This study analyzes the gaps
rithms, companies need to establish a data-quality
between components by converting the compo­
control process to develop quality metrics, collect
nents’ images from vector to pixel images
new data, evaluate data quality, remove inaccurate
through a machine learning algorithm. This
data from the training data set, and assess the trade-
image conversion process facilitates the study
off between quality-assurance costs and gains (Lee
and contributes to the company by analyzing
and Shin 2020).
the design of the components and the gaps
The final challenge of applying machine learning is to
necessary for a stable and optimal production
analyze the relevant costs and benefits. Although the
process.
relevant benefits for using machine learning are huge, it
The input of the machine-learning algorithm was
can still be difficult to proving the value of the invest­
an image of a metal sheet, which contains the com­
ment to stakeholders owing to the delay between
ponents’ drawings and their corresponding thresh­
investment and reward (Deloitte Technology, 2018).
olds. This image is used to identify the acceptable
Organizational managers need to keep in mind that
distance between any two components. In addition,
machine learning cannot solve all the business pro­
the given image is a .svg file and a gray scale image.
blems; therefore, it is necessary to investigate the invest­
The output of the algorithm is the same image as the
ment analysis with all stakeholders, including top
input with the location of a too narrow gap (less than
management and users, before its deployment. In the
the given threshold) between components, and it is
case of a risky and irreversible project, traditional invest­
marked by a circle. The output image should show all
ment-evaluation techniques may not be suitable to
the possible narrow areas between components.
capture the value of machine learning projects. It is
This study uses Python 3.7.0 to program the neces­
therefore advisable to adopt a real-option approach to
sary stages of the machine-learning algorithm. During
investment justification, which offers managers the
this project, there were several library files used,
opportunity to optimize the planning of machine learn­
which are listed and briefly described in Table 2.
ing projects (Lee and Lee 2015).
These library files were used to achieve better images
 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 7

of the gaps between components. In order to solve Table 2. List of library files used during the study.
the problems related to the gaps, it is essential to Serial Description of the library file and its
no. Library file functionality
obtain better visibility of the images. 1 requirements.txt: Contains all the needed libraries. In
The most direct way to measure the distances of addition, this file is also used for installing
these libraries.
components in an image is by measuring the coordi­ 2 setup.sh: Contain the commands for setting up
nates of these in a pixel image. Since the input is not a virtual environment as well as installing
the needed libraries.
a pixel image (it is an svg image), it should be con­ 3 run.sh: Contains the commands for activating the
verted into pixel. By doing that, the input can now be virtual environment and running the
application.
handled as a 2D matrix W with m rows and n columns 4 Config.py: Contains all the default parameter values;
(m, n are known as the size of the converted image, you can modify it directly in this file.
5 Preprocessing.py: This file is responsible for pre-processing the
which we can adjust in the converting process). Each input images so that we can pass these as
input to our application. Preprocessing.py
entry in row ith (1 � i � m1) and column jth (1 � j � n) contains some functions; each function
of this matrix, Wij , is proportional to the brightness of does the specific task.
6 Component.py: We are working with the image input. In
the image at that point. Because the input is a gray particular, we measure the distance of
scale image, each entry Wij takes a value between 0 to any 2 of the components on this image,
so it is better for us if we get the
255 (0 is taken to be black, 255 is taken to be white). components from the image and treat
them as objects. In this class Component,
we define some needed attributes and
methods.
7 Image.py: This file represents the Image object. In this
3.1 Justification for using machine-learning class, we also define some needed
attributes and methods for Image object.
algorithm in this study 8 Processing.py: This file performs the main purpose of this
application. It plots the circles in every
Sheet metal companies demand a high level of knowl­ single pixel in the components, and after
edge and expertise from the competent designers. To that these circles are responsible for
measuring the distance.
achieve such expertise in this industrial sector, various
artificial intelligence techniques are being deployed
with the objective of reducing complexity, minimizing considered a powerful and lasting solution (Ghatnekar
the human work force as well as improving opera­ 2018). Using computer vision and machine learning
tional excellence. The machine learning algorithm as can provide a robust and effective approach to over­
part of artificial intelligence is one of the most power­ come the challenge of detecting and classifying com­
ful techniques for solving engineering problems and ponent defects. Image recognition is considered an
reducing complexity and minimizing the use of human important strategy, where an image is used to detect
expertise and time taken for production processes. which type of defect the components or parts have. In
There are several successful applications of ML in real-time applications, the image recognition techni­
manufacturing industries. The major application of que is widely used in industries (Ghatnekar 2018). The
ML is monitoring (Chinnam 2002), especially in the machine-learning algorithm also helps to predict the
areas of quality monitoring, machine condition mon­ occurrence of defects in sheet metal forming processes
itoring, fault diagnosis, tool wear, optimization, etc. (Dib, Oliveira, and Marques et al. 2020), and in the
(Wuest eal.2016).In addition, ML is also used in man­ identificationof parts (Sheu et al. 2020). These unique
ufacturing in image recognition, where the images characteristics of the ML algorithm influenced and jus­
are used to identify damaged products. These appli­ tified the authors in deploy this methodology in the
cations of ML in different manufacturing and optimi­ study.
zation problems demonstrate the vast adaptability Other than machine learning algorithm, other
and applicability of the ML algorithm. algorithms are used in sheet metal industries for
Component quality is an essential parameter for the various purposes. For instance, linear program­
sheet metal industry, where it is a tedious job to manu­ ming can be used to minimize the cost of sheet
ally inspect for component defects caused by the metal punching when batching orders (Herrmann
punching operation. In order to avoid this, it is useful and Delalio 2001). Kakandikar, Darade, and
to use an image-based component investigation strat­ Nandedkar (2009) deployed genetic algorithm to
egy. This vision-based investigation system is optimize the geometry parameters (e.g. die
8 A. SHAMSUZZOHA ET AL.

Figure 1. Function to obtain the coordinates of components.

design, hammering sequence, blank holder pres­ Let us assume that the threshold (minimum
sure, etc.) in the sheet metal industry. acceptable distance) is d (pixels). On every pixel
Ashokkumar et al. (2020) proposed the use of p in the boundary, we plot a circle center p and
ranking algorithm to optimize the quality of radius d/2. Therefore, if any 2 circles which come
parts produced in sheet metal forming. Jiao and from different components are overlapping each
Xing (2018) used heuristic algorithm in analyzing other, then this is the location we need to mark
the assembly deformation of parts, clamps and (because at this location, the distance of the 2
supporting locators in the sheet metal industry. components is less than d). Figure 3 shows various
circles over the components needed to find the
gaps between them.
3.2 Description of the machine learning algorithm
• Step 1: Convert input image from .svg file (vector
3.3 Coordinates of components
image) to .jpg or .png file (pixel image)
• Step 2: From the pixel image input, obtain the As mentioned earlier, the most direct way to measure
coordinates of the components. Figure 1 displays the the distance dab between point a and point b in 2D
necessary program for conversion of vector image to dimension is from its coordinates: (xa, ya) and (xb, yb)
pixel image. Figure 2 visualizes the converted image respectively, using the formula:
from vector to pixel image. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
• Step 3: Get rid of the body of the component. dab ¼ ðxa xb Þ2 þ ðya yb Þ2 (1)
The main purpose of this algorithm is computing
the distance between components. In particular, the However, how do we obtain the coordinates of
distance between the boundaries of those compo­ the components in an image? The image basically
nents needs to be measured. Therefore, there is no contains two parts: components and background.
need to take care of the body. Furthermore, getting The color of the background is completely differ­
rid of the body also helps to reduce the running time ent from the components. In our case, the back­
of the algorithm. ground is black, while the component is not,
• Step 4: Perform the detection which means in the matrix that if any entry
 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 9

Figure 2. Transformation of vector image to pixel image.

Figure 3. Various boundary circles over a single component.(a) The sample input image with (b) The corresponding graph white
components.

takes the value of 0, it must belong to the back­ Separate the components from each other. For
ground. Based on that, we have two approaches each component, detect whether the gap between
to solve this problem. it and the others is larger than d or not.
Mark the area where the gap is less than d.

3.4 Algorithm used for the detection of narrow

3.4.1 Segmentation and separation of the
gaps
components
To detect the narrow gaps (less than a given threshold An input image has two parts: components (white)
d), the authors performed the following steps: and background (black). It can be represented as
Segment and identify the components in images. an m × n matrix W, whose entries range from 0
10 A. SHAMSUZZOHA ET AL.

(black) to 255 (white). Since the background is black, algorithm, the narrow gaps between two circles OA and
O
the components can be extracted by removing all the B are not overlapping if the distance between points
‘0ʹ entries in the matrix. Furthermore, it is easy to see A and B is satisfied as:
that the entries, which represent a component, can be
formed in a set of vertices in an undirected graph G as d d
dAB � rA þ rB ¼ þ ¼d
shown in Figure 4. Any vertex (also an entry of matrix 2 2
W) ‘u’ in G has at most 8 adjacent nodes, which are Algorithm:
entries around ‘u’. To detect a component in the
image, firstly iterating over entries in W, if there is an entry Wij Function Main (W):
6 = 0, consider it as a root, then reach other nodes. If it is reaching for each entry u in W do
a node such that at least one of its adjacent nodes has if u has value/= 0 and u has not been visited, then
non-zero value, then this node would be a part of the Get – components (W, u)
components. The process terminates if we reach the Function Get- components (W, u):
background i.e. entry that has ‘0ʹ value. Algorithm 1 Declare B as a set of boundary points
below shows the pseudocode of the implementation Declare P as a set of points of a component, includ­
of the above algorithm in order to obtain the coordi­ ing the boundary points. Initialize an empty queue Q
nates of the components as well as their boundary. Insert u to B and Q
Mark u as visited entry while Q is not empty do
3.4.2 Detection of narrow gap v = Q.back()
The information in Figure 5 is used to explain the for each entry a around v do
detection of narrow gaps between two components if a has value = 0 and a has not been visited, then
in a worksheet. The gaps between components P and Mark a as visited point
Q, as shown in Figure 5, need to be identified. In order Insert a to Q
to identify the gaps, let us assume two arbitrary If a is a boundary point, then
points, A and B, on components P and Insert a to B
Q respectively. The distance between point A and
point B is assumed as ‘d’. In order to detect the gaps The pseudocode of this algorithm is described as
between components P and Q, let us consider two below.
circles, OA and OB, on components P and Q with
centers at points A and B respectively with equal Function Detection(W):
radius r = d/2. Based on this idea, the authors suggest for all components P in image do
an algorithm to identify the narrow gaps between for all points A in boundary of P do
components P and Q. According to the proposed Draw a circle OA, center in A, radius d

Figure 4. The sample input image and its corresponding graph.

 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 11

Figure 5. Two overlapping circles of different components.

if there exists a point B in another component with P, company has a line of machines/solutions with
in which the circle OB overlaps with OA various levels of automation available. One of the
then most sophisticated ones is fully automated and
Output A and B as error points requires little human interaction. It has several
product lines such as TheBEND – sheet metal
bending, TheCOMBI – multifunctional systems,
3.4.3 Polygonization e.g. punching and laser cutting, TheLASER – sheet
To generate a polygon to visualize the error ”zones”, the metal cutting with a laser with some products
authors applied Grahm’s Scan algorithm to find the providing also welding and drilling capabilities,
convex hull of finite set of error points in the plane. ThePUNCH – sheet metal punching, TheSYSTEM –
Given the error points, this algorithm allows for the a versatile range of solutions, which combines
construction of a convex hull in O(n log n) time using functionalities of the case company’s machines in
only comparison, addition and multiplication opera­ one automated production line, TheSOFTWARE –
tions. The details of Grahm’s Scan algorithm can be a number of additional software solutions, which
found in (Kong, Everett, and Toussaint 1990). further optimize machine operations. In addition,
With the Tulus® software family the company is
capable of parts order and inventory handling,
4. Study of a sheet metal company: an work scheduling and machine capacity monitoring,
exploratory case example control and monitoring of machines’ tasks, control
of material storage, production reports, integration
4.1 Description of the case company with ERP (enterprise resource planning), and acting
The case company operates in the sheet metal as MES (manufacturing execution system). The case
forming industry and is a manufacturer of sophis­ company’s products are used in many industries:
ticated sheet metal forming machines. The aerospace, agricultural, automotive, domestic
12 A. SHAMSUZZOHA ET AL.

appliances, elevators, HVAC, hospital and lab from the nest and organize the job queue in order
equipment, to list just a few. The case company to optimize the throughput during the unmanned
is an innovative company that always searches for working shift.
and is open to new ideas and close cooperation at
different levels.
4.2 Case definition and assumptions
There are many facilities of the company in Italy,
Finland, the USA and China, from which the company This research focused on the problem of estimating
delivers machines and systems all over the world. The the gaps between components for the sheet metal
company’s sales and service network is active in over industry. We suppose all structural parts of the sheets
80 countries, with direct presence or through are made with the same kinds of materials and fabri­
a network of specialized dealers. The company’s pro­ cated by conventional machining processes. After
duct family offers highly advanced servo-electric solu­ analyzing the component punching problem, the
tions for punching, bending and integrated processes gaps between the components are decided. These
and is the widest in the world. All the products from include the image of the components, weight of the
the company are developed according to the ‘green sheets from which components were produced,
means’ concept, combining sustainability and punching rate, number of sheets punched, and cut­
productivity. ting tool changes in the punching operation. To sim­
Nesting applications are very advanced in the plify the punching operation and increase its
sheet metal industry. The initial target for the nesting production rate in general, we considered the speed
algorithms is to reduce waste and optimize the utili­ of punching rate, number of sheets in the process,
zation of the metal sheet process in the case com­ and the number of changes of cutting tools in the
pany. However, once the components to be nested process. In this case, we divided the punching com­
are organized too close to each other there is a risk plexity to four different levels (simple, medium, com­
that parts or the metal sheet frame that finally will be plex, and very complex). The study suggests that
waste material will drop to the table. In the case of punched components can be categorized into two
fully automated nesting like Libellula for SI different levels for every geometric complexity of
Engineering (https://www.libellula.eu/media/sicam- the structural components. The new punching pro­
tutorial/?lang=en), there is a risk that the nesting is blem contains two input variables (gap and complex­
too tight and the scrap that is supposed to stay on the ity of the sheets) and one output (production rates).
frame of the metal sheet scrap will drop out. In case of We try to find the relationship between inputs and
manual operations, this is not a big issue – the opera­ outputs through machine learning methods.
tor simply removes the extra scrap from the worktable This research study was conducted using machine
between loading a new metal sheet to the machine. and deep learning techniques to identify the weak
However, in the case of fully automated work, like points from the metal sheet. The process was as lean
running the machine station unmanned during the as possible and the images of the nesting results were
night shift, the scrap material will cause a much big­ used as input. The result ended with two job queues:
ger problem. one that needs operator presence and one that can
Once the scrap component drops on the workta­ be run unmanned. Besides the input images, the
ble, the whole production process stops and waits for operator should be able to fine tune the settings of
the morning for the operator to fix the error trigger­ the validation process of the nests by specifying how
ing situation. The simplest solution for this kind of wide the gap can be with various materials to keep
problem is to organize the work queue so that the the metal sheet scrap frame in one piece. The source
nesting that historically has not caused this kind of images were in vector graphic format, which were
problem will be entered to a separate job queue to be converted to pixel graphics. After being able to oper­
manufactured during the unmanned working shift. In ate the images in pixel graphic format, the machine
addition, a human operator can pre-check the nesting learning algorithm was developed in order to identify
and estimate whether the nest will cause extra scrap. the boundaries of the components on the sheet. After
The most optimal solution would be a machine learn­ identifying the components, the actual gaps between
ing solution to automatically find the weak points the parts were measured.
 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 13

5. Analysis of results for moving away the waste sheet to the trash bin. This
also causes a complete shutdown of the punching
After applying input image as shown in Figure 6 (a)
operation and consequent production loss. The target
with the parameter threshold of 1.5 pixels, the output
of the case company was therefore to optimize the
image is obtained as in Figure 6 (b). From the output,
gaps between the components so that waste metal
it can be seen that the algorithm works quiet well. It
sheet does not stay on the worktable.
detects exactly the narrow locations of less than 1.5
In order to optimize the gaps between the
pixels (the circles in Figure 6(b)). In addition, it also
punched components, some pictures attached with
returns the boundary exactly and removes the body
the design of the components over the sheet metal
of components.
were used to analyze the gaps. These pictures were in
vector image format, and were converted to pixel
5.1 Sensitivity to noise images. This conversion makes the pictures more visi­
ble in order to optimize the gaps between the com­
The proposed algorithm goes through every single ponents. In addition, the authors emphasize here that
pixel in the input; therefore, it is very sensitive to the input image shows the components in white color
noise, which could be the variation of brightness or and the background in black.
color information in the image input. This problem Figure 8(a) visualizes the gaps between several
can be solved by preprocessing the input image. components which are more than 5 pixels. They
satisfy the design requirement, and as can be
5.2 Running time seen, no error is noticed. On the other hand,
Figure 8(b) displays the gaps between compo­
As mentioned earlier, this algorithm takes O (m n). nents of more than 5 pixels which are not accep­
In general, this is quite good for an application. For table in terms of the design requirement and
more complicated input (more noise, more compo­ create problems on the worktable. Due to such
nents), this algorithm takes a little more time time to inconsistencies with the design requirements, sev­
run. This problem can be solved by vectorizing the eral errors are noticed after analysis, as are also
operations in the algorithm. This can produce a better shown in Figure 8(b).
running time by forgetting to use for-loop and using
vectorized arithmetic instead, as shown in Figure 7.
5.4 Analysis of the gaps between the components
In order to optimize the gaps between punched
5.3 Gap detection
components, some pictures attached with the
The gaps between each of the components are critical design of the components over the sheet metal
in order to punch the components after punching on were taken into consideration to analyze the
the metal sheet. If the gaps between the components gaps. These pictures were in vector image format,
are too narrow, this creates a problem over the work­ which were converted to pixel image by machine
table by leaving some of the sections of the sheet on learning algorithm. This conversion makes the pic­
it. The rest of the parts stick on the worktable and tures more visible to optimize the gaps between
trigger a stop of the automated grippers responsible the components.

Figure 6. Input and output images of the worksheet.(a) Input Image (b) Output Image.
14 A. SHAMSUZZOHA ET AL.

Figure 7. Comparative speeds of execution for conditional loop-based code block.

Figure 9 (a) visualizes the gaps between two adja­ 6. Managerial implications
cent components; it is 1.5 pixels and satisfies the
Operational excellence is a key criterion for organiza­
design requirement with no error being noticed, as
tions to be profitable with higher customer satisfac­
displayed in Figure 9 (b).
tion. Organizations are striving to achieve operational
On the other hand, Figure 10(a) displays the gaps
excellence within their limited resources. Any com­
between components of more than 1.5 pixels: it is not
plexities or bottlenecks represent a heavy cost to
acceptable in terms of design requirement and cre­
organizational success and negatively affect customer
ates a problem on the worktable. Due to such incon­
goodwill. Organizational managers therefore always
sistencies with the design requirements, several errors
need to be careful to avoid such negative impacts on
are noticed after analysis and visualized in
their organizations. Managers need to be capable of
Figure 10(b).
developing their situational awareness, which might
Such errors continue to increase as long as the
evolve from their operational processes.
optimum gaps between the components are not
In this research, the overall objective was to bring
achieved. This optimization is done by computer
about operational excellence by overcoming pro­
vision algorithm, which is essentially a part of
blems existing in an organization. The studied case
machine learning, as explained earlier. This optimiza­
company was suffering an operational problem,
tion process facilitates the waste sheet management
which was solved by applying an up-to-date techni­
process after the necessary operations. Figure 11
que known as computer vision. This computer vision,
visualizes the unacceptable gaps between the com­
which is a branch of machine learning, helped to solve
ponents as red polygons.
the existing problem in the company. This technique

Figure 8. Image and its pixel matrix.(a) Acceptable Gaps (b) Unacceptable Gaps.
 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 15

Figure 9. Visualization of acceptable gap between components and associated no-errors.(a) Acceptable gaps between components (b)
No-errors due to acceptable gaps between components.

Figure 10. Visualization of non-tolerable gap between components and associated errors.(a) Non-tolerable gap between components
(b) Errors due to non-tolerable gap between components.

Figure 11. Red polygons show the too narrow gaps between components.

involved image processing which is deployed to The study results highlighted in this research
detect the gaps between punched components. This provide a guideline for organizational managers
image processing can be applied to other operational to deploy state-of the-art technologies and tools
processes within organizations to solve other related for mutual benefit. In deploying such technologies
problems. and tools, managers also need to trade off their
16 A. SHAMSUZZOHA ET AL.

benefits with the associated costs involved. In components. Moreover, this study experienced many
using the deployed image processing technique coordinate points during the error recognition process
within machine learning algorithm, managers that can be minimized through optimization of the gaps
would be able to avoid limitations to their opera­ between components as much as possible. In addition,
tional processes and optimize the processes for a future study might continue to investigate and com­
competitive advantage. pare the study findings with other available algorithms
and techniques such as genetic algorithm, heuristic algo­
rithm, computer vision, simulation, etc. with the objec­
7. Discussion and conclusions
tive of solving the operational bottlenecks of the case
During the study, a component analysis technique was company. Moreover, further study also can be initiated
used to create and narrow gap detection between com­ to investigate the component gaps through partial least
ponents from images. It was observed from this study square regression and popular feature extraction
that the proposed algorithm works well, except for the method.
fact that there are some problems that still need to be
solved to obtain better results such as sensitivity to noise
and speeding up the running time. The component Disclosure statement
analysis technique was adopted through image proces­
No potential conflict of interest was reported by the author(s).
sing, where the computer vision principle was used. This
image processing technique, which is an integral part of
machine learning, is used to make the gaps between
ORCID
components clearly visible. It is critical to maintain the
gaps between components during the drawing stage Ahm Shamsuzzoha http://orcid.org/0000-0002-4219-0688
that supports the punching operation.
This study clearly identifies and contributes to
improving the operation limitations of a case sheet References
metal company in Finland. The application of the com­
Acceure. 2018. “Redefine Banking with Artificial Intelligence.”
ponent analysis technique, which was formulated Retrieved 2 July 2020, from https://www.accenture.com/_
through machine learning, helps the company’s opera­ acnmedia/pdf-68/accenture-redefine-banking.pdf
tional flexibility and minimizes the production time. It Ademujimi T.T., M.P. Brundage, and V.V. Prabhu.“A Review of
ensures the maintaining of the shifts at the company Current Machine Learning Techniques Used in
during both day and night. For instance, in the case of Manufacturing Diagnosis.„ In: Lödding H., Riedel R.,
Thoben KD., von Cieminski G., Kiritsis D. (eds) Advances in
complex components, where the gaps between them
Production Management Systems. The Path to Intelligent,
was critical, they were separated and allocated to the day Collaborative and Sustainable Manufacturing. APMS 2017.
shift when human operators can remove the waste sheet IFIP Advances in Information and Communication
after the punching operation if blocked by the auto­ Technology, vol 513. Springer, Cham. https://doi.org/10.
mated gripper. On the other hand, less complicated 1007/978-3-319-66923-6_48 2017.
components, where there are sufficient gaps between Anglisano, A., L. Casas, M. Anglisano, and I. Queralt. 2020.
“Application of Supervised Machine-learning
the components, are allocated to the night shift, where
Methods for Attesting Provenance in Catalan
automated grippers comfortably remove the waste Traditional Pottery Industry.” Minerals 10: 8.
sheets after operation. In this way, this gap optimization doi:10.3390/min10010008.
technique through machine learning application contri­ Ashokkumar, S. P., S. S. Sivam, S. Balasubramanian, and
butes to maintaining production stability and quality at R. V. Nanditta. 2020. “Effects of Process Variables
the same time. Optimization on the Quality of Parts Processed in High
This study includes some limitations that can be Speed Single Point Incremental Sheet Metal Forming by
Ranking Algorithm.” Materials Today: Proceedings. 10.1016/
improved in a future study. For instance, the runtime
j.matpr.2020.08.614.
was not fast enough to execute the system within Backman, J., V. Kyllönen, and H. Helaakoski. 2019. “Methods
a reasonable time. The image processing can also be and Tools of Improving Steel Manufacturing Processes:
made noise free to produce faster image conversion, Current State and Future Methods.” IFAC-PapersOnLine 52:
which is necessary to detect the gaps between 1174–1179. doi:10.1016/j.ifacol.2019.11.355.
 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 17

Baumers, M., and E. Özcan. 2016. “Scope for Machine Learning Dib, M. A., N. J. Oliveira, A. E. Marques, et al. 2020. “Single and
in Digital Manufacturing, Short Position Paper.” Royal Society Ensemble Classifiers for Defect Prediction in Sheet Metal
Workshop, July 2016: “Realising the Benefits of Machine Forming under Variability.” Neural Computing &
Learning in Manufacturing”, University of Nottingham, UK Applications 32:12335–12349. 10.1007/s00521-019-04651-6.
Bianconi, F., L. Ceccarelli, A. Fernández, and S. A. Saetta. 2014. Dutta, N., P. Kaliannan, and U. Subramaniam. 2021.
“A Sequential Machine Vision Procedure for Assessing Paper “Application of Machine Learning Algorithm for
Impurities.” Computers in Industry 65: 325–332. doi:10.1016/j. Anomaly Detection for Industrial Pumps.” In Machine
compind.2013.12.001. Learning Algorithms for Industrial Applications. Studies in
Cakir, M., M. A. Guvenc, and S. Mistikoglu. 2021. “The Computational Intelligence, edited by S. Das, S. Das,
Experimental Application of Popular Machine Learning N. Dey, and A. E. Hassanien, 237–263. Cham, 907:
Algorithms on Predictive Maintenance and the Design of Springer.
IIoT Based Condition Monitoring System.” Computers & Garg, S., S. Sinha, A. K. Kar, and M. Mani. 2021. “A Review of
Industrial Engineering 151: 106948. doi:10.1016/j. Machine Learning Applications in Human Resource
cie.2020.106948. Management.” International Journal of Productivity and
Carbonneau, R., K. Laframboise, and R. Vahidov. 2008. “Application Performance Management. ahead-of-print No. 10.1108/
of Machine Learning Techniques for Supply Chain Demand IJPPM-08-2020-0427.
Forecasting.” European Journal of Operations Research 184: Garre, A., M. C. Ruiz, and E. Hontoria. 2020. “Application of
1140–1154. doi:10.1016/j.ejor.2006.12.004. Machine Learning to Support Production Planning of
Carbo-Valverde, S., P. Cuadros-Solas, and F. Rodríguez- a Food Industry in the Context of Waste Generation under
Fernández. 2020. “A Machine Learning Approach to the Uncertainty.” Operations Research Perspectives 7: 100147.
Digitalization of Bank Customers: Evidence from Random doi:10.1016/j.orp.2020.100147.
and Causal Forests.” PLoS ONE 15 (10): e0240362. Ge, Z., Z. Song, S. X. Ding, and B. Huang. 2017. “”Data Mining
doi:10.1371/journal.pone.0240362. and Analytics in the Process Industry: The Role of Machine
Chandran, P. 2018. “Disruption in Retail: AI, Machine Learning, Learning.”.” IEEE Access 5: 20590–20616. doi:10.1109/
and Big Data”, Medium (2018, July 23), Retrieved 20 June ACCESS.2017.2756872.
2020, from https://towardsdatascience.com/disruption-in- Ghatnekar, S. 2018. “Use Machine Learning to Detect Defects
retail-ai-machine-learning-big-data-7e9687f69b8f on the Steel Surface.” Available at: https://software.intel.
Chinnam, R. B. 2002. “Support Vector Machines for Recognizing com/content/www/us/en/develop/articles/use-machine-
Shifts in Correlated and Other Manufacturing Processes.” learning-to-detect-defects-on-the-steel-surface.html,
International Journal of Production Research 40: 4449–4466. accessed on 11 April 2021
doi:10.1080/00207540210152920. Golkarnarenji, G., M. Naebe, K. Badii, A. S. Milani, R. N. Jazar, and
Chiu, M.-C., C.-D. Tsai, and L. T-l T-L. 2020. “An Integrative Machine H. Khayyam. 2019. “A Machine Learning Case Study with
Learning Method to Improve Fault Detection and Productivity Limited Data for Prediction of Carbon Fiber Mechanical
Performance in a Cyber-physical System.” Journal of Computer Properties.” Computers in Industry 105: 123–132.
Information Science and Engineering 20(2): 12. pages. https://doi. doi:10.1016/j.compind.2018.11.004.
org/10.1115/1.4045663 Gottsegen, G. 2020. “Machine Learning Is Changing the Way
Danton, S., M. D. Char, A. Feudtner, and C. Feudtner. 2020. Retailers Do Business.” Built In. Retrieved 3 July 2020, from
“Identifying Ethical Considerations for Machine Learning https://builtin.com/artificial-intelligence/machine-learning-
Healthcare Applications.” The American Journal of Bioethics ecommerce-retail
20 (11): 7–17. doi:10.1080/15265161.2020.1819469. Greenspan, H., B. van Ginneken, and R. M. Summers. 2016. “Deep
de Andrés, M. L. O., A. Poniszewska-Marańda, and L. A. Hernández Learning in Medical Imaging: Overview and Future Promise of
Gómez. 2020. “Towards the Machine Learning Algorithms in an Exciting New Technique.” IEEE Transactions, Medical
Telecommunications Business Environment.” In Information Imaging 35: 1153–1159. doi:10.1109/TMI.2016.2553401.
Systems. EMCIS 2020. Lecture Notes in Business Information Gupta, A. 2017. “Machine Learning Challenges in the
Processing, 402, edited by M. Themistocleous, M. Papadaki, Implementation of Industrial Internet of Things.” available at:
and M. M. Kamal. Cham: Springer. 79-96. https://iiot-world.com/artificial-intelligence-ml/artificial-
“Deloitte Technology, Media, and Telecommunications intelligence/machine-learning-challenges-in-the-
Predictions.” 2018. Retrieved 23 June 2020, from https:// implementation-of-industrial-internet-of-things/, accessed on
www2.deloitte.com/content/dam/Deloitte/global/Images/ 13 April 2021.
infographics/technologymediatelecommunications/gx- Gupta, S., and R. R. Sedamkar. 2020. “Machine Learning for
deloitte-tmt-2018-predictions-full-report.pdf Healthcare: Introduction.” In Machine Learning with
Dib, M., B. Ribeiro, and P. Prates. 2018. “Model Prediction of Health Care Perspective. Learning and Analytics in
Defects in Sheet Metal Forming Processes.” In: Pimenidis E., Intelligent Systems, 13, edited by V. Jain and
Jayne C. (eds) Engineering Applications of Neural Networks. J. Chatterjee. Cham: Springer.1-25.
EANN 2018. Communications in Computer and Information Hajizadeh, Y. 2019. “Machine Learning in Oil and Gas: SWOT
Science, vol 893. Springer, Cham. https://doi.org/10.1007/ Analyses Approach.” Journal of Petroleum Science and
978-3-319-98204-5_14 Engineering 176: 661–663. doi:10.1016/j.petrol.2019.01.113.
18 A. SHAMSUZZOHA ET AL.

Hamouche, E., and E. G. Loukaides. 2018. “Classification and Köksal, G., İ. Batmaz, and M. C. Testik. 2011. “A Review of Data
Selection of Sheet Forming Processes with Machine Mining Applications for Quality Improvement in
Learning.” International Journal of Computer Integrated Manufacturing Industry.” Expert Systems and Applications
Manufacturing 31: 921–932. doi:10.1080/ 38: 448–467. doi:10.1016/j.eswa.2011.04.063.
0951192X.2018.1429668. Kong, X., H. Everett, and G. Toussaint. 1990. “The Graham Scan
Hegde, J., and B. Rokseth. 2020. “Applications of machine Triangulates Simple Polygons.” Pattern Recognition Letters
learning methods for engineering risk assessment – A 11: 713–716. doi:10.1016/0167-8655(90)90089-K.
review.” Safety Science, 122, 104492, https://doi.org/10. Kwitek, M. 2016. “A Feasibility Study of Azure Machine Learning
1016/j.ssci.2019.09.015. for Sheet Metal Fabrication.” Master’s thesis, Industrial
Herrmann, J. W., and D. R. Delalio. 2001. “Algorithms for Management Unit, University of Vaasa, Finland.
Sheet Metal Nesting.” IEEE Transactions on Robotics and Lee, I. 2017. “Big Data: Dimensions, Evolution, Impacts, and
Automation 17: 183–190. doi:10.1109/70.928563. Challenges.” Business Horizons 60: 293–303. doi:10.1016/j.
Huang, H., and J. Zhu. 2021. “A Short Review of the bushor.2017.01.004.
Application of Machine Learning Methods in Smart Lee, I., and K. Lee. 2015. “The Internet of Things (Iot):
Airports.” Journal of Physics: Conference Series 1769: Applications, Investments, and Challenges for Enterprises.”
012010. Business Horizons 58: 431–440. doi:10.1016/j.
Huber, J., and H. Stuckenschmidt. 2020. “Daily Retail bushor.2015.03.008.
Demand Forecasting Using Machine Learning with Lee, I., and Y. J. Shin. 2020. “Machine Learning for Enterprises:
Emphasis on Calendric Special Days.” International Applications, Algorithm Selection, and Challenges.” Business
Journal of Forecasting 36: 1420–1438. doi:10.1016/j. Horizons 63: 157–170. doi:10.1016/j.bushor.2019.10.005.
ijforecast.2020.02.005. Li, Z., Y. Wang, and K. Wang. 2019. “A Deep Learning Driven
Jayant, A., A. Agarwal, and V. Gupta. 2021. “Application of Method for Fault Classification and Degradation Assessment
Machine Learning Technique for Demand Forecasting: A in Mechanical Equipment.” Computers in Industry 104: 1–10.
Case Study of the Manufacturing Industry.” In: Pandey P.M., doi:10.1016/j.compind.2018.07.002.
Kumar P., Sharma V. (eds) Advances in Production and Lin, Z. C., and D. Y. Chang. 1995. “Application of a Neural
Industrial Engineering. Lecture Notes in Mechanical Network Machine Learning Model in the Selection System
Engineering 403–421, Springer, Singapore. https://doi.org/ of Sheet Metal Bending Tooling.” Artificial Intelligence and
10.1007/978-981-15-5519-0_31 Engineering 10: 21–37. doi:10.1016/0954-1810(95)00013-5.
Jiao, Z., and Y. Xing. 2018. “Clamping-sequence Optimisation LinkedIn’s. 2017. “U.S. Emerging Jobs Report.” Retrieved 26
Based on Heuristic Algorithm for Sheet-metal Components.” June 2020, from https://economicgraph.linkedin.com/
International Journal of Production Research 56: 7190–7200. research/LinkedIns-2017-US-Emerging-Jobs-Report
doi:10.1080/00207543.2017.1410245. Massel, L. V., O. M. Gerget, A. G. Massel, and T. G. Mamedov.
Joseph, J., O. Sharif, A. Kumar, S. Gadkari, and A. Mohan. 2019. “The Use of Machine Learning in Situational
2014. “Using Big Data for Machine Learning, Analytics in Management in Relation to the Tasks of the Power
Manufacturing, Manufacturing Innovation and Industry.” EPJ Web of Conferences 217, 01010. Irkutsk, Baikal
Transformation Group.” White Paper available at: http:// Lake, Russia.
tinyurl.com/zfpvzau, Accessed on 07 April 2021. McCoy, J. T., and L. Auret. 2019. “Machine-learning Applications
Kakandikar, G. M., P. D. Darade, and V. M. Nandedkar. 2009. in Minerals Processing: A Review.” Minerals Engineering 132:
“Applications of Evolutionary Algorithms to Sheet Metal 95–109. doi:10.1016/j.mineng.2018.12.004.
Forming Processes: A Review.” International Journal of McDowell, H. 2018. “JP Morgan Trading Team Develops
Machine Intelligence 1: 47–49. doi:10.9735/0975- Machine Learning Model for Equity Desk.” The Trade (2018,
2927.1.2.47-49. April 12), Retrieved 20 June 2020, from https://www.thetra
Kang, Z., C. Catal, and B. Tekinerdogan. 2020. “Machine denews.com/jp-morgans-trading-team-develop-machine-
Learning Applications in Production Lines: A Systematic learning-model-equity-desk/
Literature Review.” Computers & Industrial Engineering 149: Min, Q., Y. Lu, Z. Liu, C. Su, and B. Wang. 2019. “Machine
106773. doi:10.1016/j.cie.2020.106773. Learning Based Digital Twin Framework for Production
Kashid, S., and S. Kumar. 2012. “Applications of Artificial Neural Optimization in Petrochemical Industry.” International
Network to Sheet Metal Work - A Review.” American Journal of Journal of Information Management 49: 502–519.
Intelligent System 2: 168–176. doi:10.5923/j.ajis.20120207.03. doi:10.1016/j.ijinfomgt.2019.05.020.
Motta, M., M. C. Neto, and P. Sarmento. 2021. “A Mixed
Kim, D.-H., T. J. Y. J. Y. Kim, X. Wang, M. Kim, Y.-J. Quan, J. W. Oh, Approach for Urban Flood Prediction Using Machine
J. S.-H. Min, H. Kim, B. Bhandari, I. Yang, and S.-H. Ahn. 2018. Learning and GIS.” International Journal of Disaster Risk
“Smart Machining Process Using Machine Learning: Reduction 56: 102154. doi:10.1016/j.ijdrr.2021.102154.
A Review and Perspective on Machining Industry.” Musani, P. 2018. “Eden: The Tech That’s Bringing Fresher
International Journal of Precision Engineering and Groceries to You.” Walmart. Available at https://corporate.
Manufacturing-Green Technology 5: 555–568. doi:10.1007/ walmart.com/newsroom/community/20180301/eden-the-
s40684-018-0057-y. tech-thats-bringing-fresher-groceries-to-you, 1 March 2018.
 INTERNATIONAL JOURNAL OF COMPUTER INTEGRATED MANUFACTURING 19

Najafabadi, M. K., M. N. Mahrin, S. Chuprat, and H. M. Sarkan. Sun, S., Y. Wei, K.-L. Tsui, and S. Wang. 2019. “Forecasting
2017. “Improving the Accuracy of Collaborative Filtering Tourist Arrivals with Machine Learning and Internet Search
Recommendations Using Clustering and Association Rules Index.” Tourism Management 70: 1–10. doi:10.1016/j.
Mining on Implicit Data.” Computer and Human Behavior 67: tourman.2018.07.010.
113–128. doi:10.1016/j.chb.2016.11.010. Sung, T. K. 2018. “Industry 4.0: A Korea Perspective.” Technology
Nithya, B., and V. Ilango. 2017. “Predictive Analytics in Forecasting and Social Change 132: 40–45. doi:10.1016/j.
Health Care Using Machine Learning Tools and techfore.2017.11.005.
Techniques.” International Conference on Intelligent Tsai, S.-Y., and J.-Y. Chang. 2018. “Parametric Study and Design
Computing and Control Systems (ICICCS), Madurai, India, of Deep Learning on Leveling System for
492–499. Smart Manufacturing.” 2018 IEEE International Conference
O’Donovan, P., C. Gallagher, K. Leahy, and D. T. J. O’Sullivan. on Smart Manufacturing, Hsinchu, Taiwan, 8–9 February.
2019. “A Comparison of Fog and Cloud Computing Turilli, M. 2020 . “Ethical Protocols Design.” Ethics in
Cyber-physical Interfaces for Industry 4.0 Real-time Information Technology 9: 49–62. doi:10.1007/s10676-
Embedded Machine Learning Engineering Applications.” 006-9128-9.
Computers in Industry 110: 12–35. doi:10.1016/j. Weichert, D., P. Link, A. Stoll, S. Ruping, S. Ihlenfeldt,
compind.2019.04.016. and S. Wrobel. 2019. “A Review of Machine Learning
Parkes, A. J., E. Özcan, and D. Karapetyan. 2015. “A Software for the Optimization of Production Processes.” The
Interface for Supporting the Application of Data Science to International Journal of Advanced Manufacturing
Optimisation.” LION 9, LNCS 8994: 306–311. Technology 104: 1889–1902. doi:10.1007/s00170-019-
Ren, Y. 2021. “Optimizing Predictive Maintenance with 03988-5.
Machine Learning for Reliability Improvement.” ASME J . Woolf, A., and A. McIntyre. 2018. “Realizing the Full Value of AI
Risk Uncertainty Part B . 10.1115/1.4049525. in Banking.” Accenture (2018, April 4), Retrieved 26 June
Rizzi, W., Z. M. Wang, and K. Zielinski. 2020. “How Machine 2020, from https://www.accenture.com/us-en/insights/bank
Learning Can Improve Pricing Performance.” McKinsey & ing/future-workforce-banking-survey
Company (2018, September), Retrieved 3 July 2020, from Wu, X., J. Wang, A. Flitman, and P. Thomson. 1999. “Neural and
https://www.mckinsey.com/industries/financial-services Machine Learning to the Surface Defect Investigation in
/our-insights/how-machine-learning-can-improve-pricing- Sheet Metal Forming.” Proceedings of 6th International
performance Conference on Neural Information Process Perth, Australia,
Sánchez-Medina, A. J., and E. C-Sánchez. 2020. “Using November 16-20, 1088–1093.
Machine Learning and Big Data for Efficient Forecasting Wuest, T., D. Weimer, C. Irgens, and K.-D. Thoben. 2016. “Machine
of Hotel Booking Cancellations.” International Journal of Learning in Manufacturing: Advantages, Challenges, and
Hospitality Management 89: 102546. doi:10.1016/j. Applications.” Production & Manufacturing Research: An Open
ijhm.2020.102546. Access Journal 23–45. doi:10.1080/21693277.2016.1192517.
Sarkar, S., T. Weyde, A. d’Avila, G. Garcez, S. Slabaugh, and Xie, Q. 2020. “Machine learning in human resource system of
C. P. Dragicevic. 2016. “Accuracy and Interpretability intelligent manufacturing industry.” Enterprise Information
Trade-offs in Machine Learning Applied to Safer Systems, Taylor and Francis, London, UK.10.1080/
Gambling.” Retrieved 3 July 2020, from https://pdfs.seman 17517575.2019.1710862
ticscholar.org/8f67/7a206e647c1a0bddf2d54175801633 Xu, Z., and J. H. Saleh. 2021. “Machine Learning for Reliability
d5a5ec.pdf Engineering and Safety Applications: Review of Current
Sharp, M., R. Ak, and T. Hedberg. 2018. “A Survey of the Status and Future Opportunities.” Reliability Engineering &
Advancing Use and Development of Machine Learning in System Safety 211: 107530. doi:10.1016/j.ress.2021.107530.
Smart Manufacturing.” Journal of Manufacturing Systems? Yeh, T.-H., and S. Deng. 2012. “Application of Machine Learning
48: 170–179. doi:10.1016/j.jmsy.2018.02.004. Methods to Cost Estimation of Product Life Cycle.” International
Sheu, R., Y. Lin, C. Huang, L. Chen, M. S. Pardeshi, and Journal of Computer Integrated Manufacturing 25: 340–352.
H. Tseng. 2020. “IDS-DLA: Sheet Metal Part Identification doi:10.1080/0951192X.2011.645381.
System for Process Automation Using Deep Learning Yildirim, P., D. Birant, and T. Alpyildiz. 2018. “Data Mining and
Algorithms.” IEEE Access 8: 127329–127342. doi:10.1109/ Machine Learning in Textile Industry.” WIREs Data Mining and
ACCESS.2020.3007257. Knowledge Discovery 8 (1): e1228. doi:10.1002/widm.1228.
Srdoč, A., I. Bratko, and A. Sluga. 2007. “Machine Learning Zekić-Sušac, M., S. Mitrović, and A. Has. 2021. ““Machine
Applied to Quality Management - A Study in Ship Repair Learning Based System for Managing Energy Efficiency of
Domain.” Computers in Industry 58: 464–473. doi:10.1016/j. Public Sector as an Approach Towards Smart Cities”.
compind.2006.09.013. International.” Journal of Information Management 58:
Stoerkle, D. D., P. Seim, L. Thyssen, and B. Kuhlenkoetter. 102074.
2016. “Machine Learning in Incremental Sheet Forming.” Zhang, J., P. Wang, and R. X. Gao. 2019. “Deep Learning-based
Proceedings of 47th International Symposium on Tensile Strength Prediction in Fused Deposition Modeling.”
Robotics, Munich, Germany, Germany, 667–673, 21–22 Computers in Industry 107: 11–21. doi:10.1016/j.
June. compind.2019.01.011.
20 A. SHAMSUZZOHA ET AL.

Zhu, R., X. Hu, J. Hou, and X. Li. 2020. “Application of Machine Zwierzycki, M., P. Nicholas, and M. Ramsgaard Thomsen. 2018.
Learning Techniques for Predicting the Consequences of “Localised and Learnt Applications of Machine Learning for
Construction Accidents in China.” Process Safety and Robotic Incremental Sheet Forming.” In: De Rycke K. et al.
Environmental Protection 145: 293–302. doi:10.1016/j. (eds) Humanizing Digital Reality. Springer, Singapore. 382,
psep.2020.08.006. https://doi.org/10.1007/978-981-10-6611-5_32