International Journal of

Advances in Scientific Research and Engineering (ijasre) E-ISSN : 2454-8006

DOI: 10.31695/IJASRE.2021.34027 Volume 7 (6) 2021

June - 2021

Deep Learning in Manufacturing

Matthew N. O. Sadiku, Guddi K. Suman, and Sarhan M. Musa
Roy G. Perry College of Engineering
Prairie View A&M University
Prairie View, TX 77446
U.S.A
________________________________________________________________________________________________________

ABSTRACT
Deep learning is presently receiving a lot of attention. It is a subset of machine learning, based on multi-layer neural networks or
deep neural networks. It is a novel, data-hungry, and high-accuracy analytics approach. This paper discusses deep learning
algorithms and their applications in manufacturing.

Key Words: Deep Learning, Deep Machine Learning, Hierarchical Learning, Manufacturing.
________________________________________________________________________________________________________

1. INTRODUCTION
One of the most vital industries in the world is manufacturing. The heart of any manufacturing system is the movement of raw
materials, goods, and parts. The search for high efficiency and enhanced productivity have always been at the heart of
manufacturing. In today’s manufacturing, technology is the determining factor for securing competitive advantage. Innovative
ideas in the field of artificial intelligence in general and deep machine learning in particular can significantly improve the quality
control tasks in a large assembly line. These technologies have come to rescue organizations such manufacturing drowned in the
deluge of data, data explosion, and information overload. In manufacturing, they are business critical.

Artificial intelligence (AI) has seen a recent surge in use across a wide variety of industries. One type of AI currently being used
in image processing is called deep learning (DL). As illustrated in Figure 1, deep learning is a subset of machine learning which,
in turn, is a subset of artificial intelligence [1]. DL is a new area emerging from machine learning and multilayer neural network.
It refers to a family of approaches that have taken machine learning to a new level, helping computers make sense out of vast
amounts of data in the form of text, images, and sound. It has come to the forefront as a way of overcoming the limitations of
neural networks when they are employed to real-world challenges [2].

2. OVERVIEW OF DEEP LEANRING

Machine learning refers to algorithms that learn from data. The learning process of these algorithms can either be supervised or
unsupervised, depending on the data being used to feed the algorithms. Deep learning algorithms can be regarded as an evolution
of machine learning algorithms. Thus, deep learning (DL), or also known as deep neural networks (DNN), refers to a set of
machine learning techniques that utilize neural networks with many hidden layers for tasks, where the goal of each layer is to
learn to transform input data into a non-linear and more abstract representation. It has become an analytical tool that has attracted
more and more attention from researchers in different areas of research in recent years.

The term “deep learning” was coined in 2006 by Geoffrey Hinton and Ruslan Salakhutdinov [3]. Deep learning has given a new
dimension to the machine learning research. Deep learning algorithms use neural networking to process massive data sets in order
to execute a specified task. They have been used to predict the quality attribute (output) with the process parameters (input). Data
is the key in deep learning’s effectiveness. Deep-learning techniques can automatically learn complex high-level data features
from a large amount of data. Traditionally, a deep learning model has to be retrained every time a new rule is included.

Deep learning relies on a layered structure of algorithms called an artificial neural network (ANN), which is inspired by the
functionality of our brain cells called neurons. Consider the example ANN in the Figure 2 [4]. The leftmost layer is called the

www.ijasre.net Page 36
Licensed Under Creative Commons Attribution CC BY-NC.4
International Journal of Advances in Scientific Research and Engineering (ijasre), Volume 7 (6), 2021

input layer, the rightmost layer is the output layer. Each layer contains one or more neurons. These multilevel layers allow models
to detect and train from their own mistakes. The term “deep” in deep learning refers to the number of layers in ANN.

The method that brought significant success to DL is called the Convolutional Neural Network (CNN),proposed by Yann
LeCun et al. in 1998. CNN is poised to take over from human operators to spot and detect visual clues.

Advancing DL systems often involves three kinds of engineers: (1) data scientists with domain experts in medical, economic, or
climate; (2) hardware designers that develop specialized hardware; and (3) performance and compiler engineers that optimize
software [5].

DL is powerful in discovering complex structures in high-dimensional data. It is widely used in many applications such as
computer vision, natural language processing, and big data analysis. In addition, DL has several applications in the manufacturing
domain. Examples of deep learning in manufacturing include sales forecasting and advanced analytics. Deep Learning is bringing
benefits to the design phase, evaluation, production, operation control, maintenance, and sustainment.

3. APPLICATIONS
Many fields of research have used deep learning tools to facilitate the processing of massive data. Applications of deep learning
are almost limitless, with different applications employing different deep learning models. Deep learning has matured from being
a special purpose machine learning technique to a general purpose machine learning tool. It has enjoyed success in various
applications such as automatic speech recognition, image recognition, computer vision, object detection, bioinformatics, drug
discovery and information retrieval, industrial machines, and manufacturing [6]. In manufacturing, deep learning has the
following applications.

 Smart manufacturing: This refers to using advanced data analytics to complement physical science for improving
system performance and decision making. It is a fully integrated collaborative system that uses sensors, Internet-
connected machines, and big data to monitor the production process and to improve manufacturing efficiencies. Smart
manufacturing envisions systems that provide us with insightful information about every step of a product lifecycle.
Information-system-enabled smart manufacturing has increased productivity and quality of industrial organizations.
Smart manufacturing will enrich the lives of consumers by providing goods and services with high quality and at an
affordable cost. Deep learning is applied in the predictive analytics for defect prognosis, i.e. maintenance and service
prediction. Deep learning provides advanced analytics tools for processing and analyzing big data in manufacturing and
make manufacturing “smart,” as shown in Figure 3 [7].

 Automotive industry: The automotive industry covers a wide range of vehicles. Deep learning has many potential
applications in the automotive industry during development, manufacturing, and sales. It is also useful in advanced
driving assistance systems, autonomous driving, and advanced detection controls [8].

 Predictive Maintenance: In predictive maintenance, data is collected over time to monitor and find patterns to predict
failures. Although predictive maintenance is applied in many industries, it thrives in manufacturing. AI in manufacturing
has come a long way with technologies like predictive maintenance. The adoption of machine learning and deep learning
in manufacturing will only improve predictive maintenance. Deep learning can aid in the predictive maintenance of
complex machinery and connected systems.

 Automation: In the hope of decreasing costs and increasing quality and the push for higher volumes of output with
lower investments, manufacturing industry is making efforts to achieve a higher degree of automation. For years,
machine builders in the automation field of assembly equipment have been focusing on the mechanics of machines.

 Robotics: Deep learning architectures enable robots to learn on their own. Companies use industrial robots to handle
complicated and dangerous processes. A robot can train itself for new tasks by object and pattern recognition capabilities
of deep learning models. Deep learning models have proven to be compelling in areas such as financial aspects, time-
setting data management and money related.

www.ijasre.net Page 37

DOI: 10.31695/IJASRE.2021.34027
International Journal of Advances in Scientific Research and Engineering (ijasre), Volume 7 (6), 2021

 Additive manufacturing (AM): This is a new paradigm in design-driven build of customized products. Machine
learning and deep learning can be adopted for both analysis and design of microlattices, which can be fabricated using
additive manufacturing techniques [9]. Figure 4 shows an example of additive manufacturing [10].

Other applications of deep learning in manufacturing include machining process, process monitoring, continuous manufacturing,
pharmaceutical manufacturing, and medical device manufacturing.

4. BENEFITS

Deep learning techniques are emerging approaches from machine learning. Manufacturing organizations that adopt DL models
derive knowledge, understanding, insight, and foresight to improve their products, processes, and decisions, thereby giving them a
competitive advantage. Other benefits include:

 Deep learning techniques are achieving unprecedented levels of accuracy. It shows trends and behaviors that influence
the direction of technology.
 They advance current manufacturing process by efficiently handling big data in a rapid pace and with autonomy.
 Deep learning algorithms can outperform humans at recognizing and classifying images.
 Deep learning requires little human intervention to function properly.
 Deep learning can integrate seamlessly with the ambitious goals of Industry 4.0.

5. CHALLENGES

A major challenge facing deep learning in various applications include:

 Deep learning requires vast amounts of data. There is the need to store and process large volumes of data as well as
dealing with unstructured data.
 Deep learning requires substantial computing power.

Despite that deep learning is emerging as a very powerful technique, it is attracting some criticism. Techniques based on deep
learning can learn complex.

6. CONCLUSION

Deep learning may be regarded as a software technology designed by programmers to teach computers to do what humans have
been doing: learning by example or learning from data. Deep learning has attracted attention in several fields of research because
it allows the extraction of features from processing large data. Leading tech companies by the likes of Google, Microsoft, and
Apple have invested heavily in DL.
Deep learning is all around us in the form of Siri, Alexa, and Google Assistant. Although deep learning is still in its infancy in
many fields of study, its breakthroughs are driving the AI boom and providing a bright future for AI. More information about
deep learning in manufacturing can be found in the books in [11,12] and related journal:
Journal of Manufacturing Systems

REFERENCES

[1] A. Wolfewicz, “Deep learning vs. machine learning – What’s the difference?” May 2021, https://levity.ai/blog/difference-
machine-learning-deep-learning
[2] M. N. O. Sadiku, M. Tembely, and S. M. Musa, ”Deep learning,” International Research Journal of Advanced Engineering
and Science, vol. 2, no. 1, 2017, pp. 77,78.
[3] G. E. Hinton and R. R. Salakhutdinov, “Reducing the dimensionality of data with neural networks,” Science, vol. 313, 2006,
pp. 504-507.
[4] S. Singh, “Cousins of artificial intelligence,” May 2018,
https://towardsdatascience.com/cousins-of-artificial-intelligence-dda4edc27b55
[5] A. Rodriguez, “Deep learning systems: Algorithms, compilers, and processors for large-scale production,” Synthesis Lectures
on Computer Architecture, vol. 15, no. 4, October 2020, pp. 1-265.

www.ijasre.net Page 38

DOI: 10.31695/IJASRE.2021.34027
International Journal of Advances in Scientific Research and Engineering (ijasre), Volume 7 (6), 2021

[6] “Deep learning,” Wikipedia, https://en.wikipedia.org/wiki/Deep_learning

[7] J. Wang et al., “Deep learning for smart manufacturing: Methods and applications,” Journal of Manufacturing Systems, 2018.
[8] A. Luckow et al., “Deep learning in the automotive industry: Applications and tools,” Proceedings of 2016 IEEE
International Conference on Big Data (Big Data), December 2016.
[9] N. Després et al., “Deep learning and design for additive manufacturing: A framework for microlattice architecture,” JOM,
vol 72, 2020, pp. 2408–2418.
[10] D. Butcher, “Additive manufacturing in the medical device industry,” November 2020, https://www.mastercontrol.com/gxp-
lifeline/additive-manufacturing-in-the-medical-device-industry/
[11] F. Chollet, Deep Learning with Python. Shelter Island, NY: Manning Publications Co., 2018.
[12] E. Stevens, L. Antiga, and T. Viehmann, Deep Learning with PyTorch. Shelter Island, NY: Manning Publications Co.,
2020.

ABOUT THE AUTHORS

Matthew N.O. Sadiku is a professor emeritus in the Department of Electrical and Computer Engineering at Prairie View A&M
University, Prairie View, Texas. He is the author of several books and papers. His areas of research interest include
computational electromagnetics and computer networks. He is a fellow of IEEE.

Guddi K. Suman is currently working towards a PhD in Electrical and Computer Engineering at Prairie View A&M University,
Prairie View, TX. She is working as a Ph.D. intern at Pacific Northwest National Lab in Richland, WA. Her areas of research
interest include space radiation effects on electronics, nanomaterials, characterization of nanomaterials and semiconductors, thin
film nanofabrication, and nanosensors. She is a student member of IEEE.
Email: [email protected]

Sarhan M. Musa is a professor in the Department of Electrical and Computer Engineering at Prairie View A&M University,
Texas. He has been the director of Prairie View Networking Academy, Texas, since 2004. He is an LTD Sprint and Boeing
Welliver Fellow. His areas of research interest include computational electromagnetics and computer networks.

Figure 1. The relationship between AI, ML, and DL [1].

www.ijasre.net Page 39

DOI: 10.31695/IJASRE.2021.34027
International Journal of Advances in Scientific Research and Engineering (ijasre), Volume 7 (6), 2021

Figure 2. Layers in artificial neural network [4]

Figure 3. The role of Deep learning in smart manufacturing [7].

www.ijasre.net Page 40

DOI: 10.31695/IJASRE.2021.34027
International Journal of Advances in Scientific Research and Engineering (ijasre), Volume 7 (6), 2021

Figure 4. Additive manufacturing [10]

.

www.ijasre.net Page 41

DOI: 10.31695/IJASRE.2021.34027